AI Email Management

AI Email Management — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  Fake nude photography

  Fake nude photography is the creation of nude photographs designed to appear as genuine nudes of an individual. The motivations for the creation of these modified photographs include curiosity, sexual gratification, the stigmatization or embarrassment of the subject, and commercial gain, such as through the sale of the photographs via pornographic websites. Fakes can be created using image editing software or through machine learning. Fake images created using the latter method are called deepfakes. == History == Magazines such as Celebrity Skin published non-fake paparazzi shots and illicitly obtained nude photos, showing there was a market for such images. Subsequently, some websites hosted fake nude or pornographic photos of celebrities, which are sometimes referred to as celebrity fakes. In the 1990s and 2000s, fake nude images of celebrities proliferated on Usenet and on websites, leading to campaigns to take legal action against the creators of the images and websites dedicated to determining the veracity of nude photos. "Deepfakes", which use artificial neural networks to superimpose one person's face into an image or video of someone else, were popularized in the late 2010s, leading to concerns about the technology's use in fake news and revenge porn. Fake nude photography is sometimes confused with Deepfake pornography, but the two are distinct. Fake nude photography typically starts with human-made non-sexual images, and merely makes it appear that the people in them are nude (but not having sex). Deepfake pornography typically starts with human-made sexual (pornographic) images or videos, and alters the actors' facial features to make the participants in the sexual act look like someone else. === DeepNude === In June 2019, a downloadable Windows and Linux application called DeepNude was released which used a Generative Adversarial Network to remove clothing from images of women. The images it produced were typically not pornographic, merely nude. Because there were more images of nude women than men available to its creator, the images it produced were all female, even when the original was male. The app had both a paid and unpaid version. A few days later, on June 27, the creators removed the application and refunded consumers, although various copies of the app, both free and for charge, continue to exist. On GitHub, the open-source version of this program called "open-deepnude" was deleted. The open-source version had the advantage of allowing it to be trained on a larger dataset of nude images to increase the resulting nude image's accuracy level. A successor free software application, Dreamtime, was later released, and some copies of it remain available, though some have been suppressed. === Deepfake Telegram Bot === In July 2019 a deepfake bot service was launched on messaging app Telegram that used AI technology to create nude images of women. The service was free and enabled users to submit photos and receive manipulated nude images within minutes. The service was connected to seven Telegram channels, including the main channel that hosts the bot, technical support, and image sharing channels. While the total number of users was unknown, the main channel had over 45,000 members. As of July 2020, it is estimated that approximately 24,000 manipulated images had been shared across the image sharing channels. === Nudify websites === By late 2024, most ways to produce nude images from photographs of clothed people were accessible at websites rather than in apps, and required payment. == Purposes == The reasons for the creation of nude photos may range from a need to discredit the target publicly, personal hatred for the target, or the promise of pecuniary gains for such work on the part of the creator of such photos. Fake nude photos often target prominent figures such as businesspeople or politicians. == Notable cases == In 2010, 97 people were arrested in Korea after spreading fake nude pictures of the group Girls' Generation on the internet. In 2011, a 53-year-old Incheon man was arrested after spreading more fake pictures of the same group. In 2012, South Korean police identified 157 Korean artists of whom fake nudes were circulating. In 2012, when Liu Yifei's fake nude photography released on the network, Liu Yifei Red Star Land Company declared a legal search to find out who created and released the photos. In the same year, Chinese actor Huang Xiaoming released nude photos that sparked public controversy, but they were ultimately proven to be real pictures. In 2014, supermodel Kate Upton threatened to sue a website for posting her fake nude photos. Previously, in 2011, this page was threatened by Taylor Swift. In November 2014, singer Rain was angry because of a fake nude photo that spread throughout the internet. Information reveals that: "Rain's nude photo was released from Kim Tae-hee's lost phone." Rain's label, Cube Entertainment, stated that the person in the nude photo is not Rain and the company has since stated that it will take strict legal action against those who post photos together with false comments. In July 2018, Seoul police launched an investigation after a fake nude photo of President Moon Jae-in was posted on the website of the Korean radical feminist group WOMAD. In early 2019, Alexandria Ocasio-Cortez, a Democratic politician, was berated by other political parties over a fake nude photo of her in the bathroom. The picture created a huge wave of media controversy in the United States. == Methods == Fake nude images can be created using image editing software or neural network applications. There are two basic methods: Combine and superimpose existing images onto source images, adding the face of the subject onto a nude model. Remove clothes from the source image to make it look like a nude photo. == Impact == Images of this type may have a negative psychological impact on the victims and may be used for extortion purposes.

  Read more →
• 

  Serge Belamant

  Serge Belamant (born 1953) is a French-born South African entrepreneur best known for designing the Universal Electronic Payment System (UEPS) and the Chip Offline Pre-authorised Card (COPAC). He founded the cash-payments company Net1 UEPS Technologies in 1989, led it through dual listings on the NASDAQ and the Johannesburg Stock Exchange, and oversaw the contentious welfare-payments contract with the South African Social Security Agency (SASSA) until his retirement in 2017. Since 2018 he has been non-executive chair of London-based buy-now-pay-later fintech Zilch. == Early life and education == Belamant moved from France to South Africa with his family in 1967 and matriculated from Highlands North Boys' High School, Johannesburg. In 1972 he entered the University of the Witwatersrand to study civil engineering but switched to computer science and applied mathematics in his second year. He left the university without a degree and later took short courses in information systems at the University of South Africa (UNISA). == Early career and SASWITCH (1981–1989) == Belamant worked for Control Data Corporation as a systems analyst for a decade before joining SASWITCH Ltd in 1985. Economic sanctions had left the consortium's national ATM network dependent on unsupported Christian Rovsing computers. Belamant led a rebuild on fault-tolerant Stratus hardware and wrote protocol-translation software that allowed fourteen banks to connect without altering their host systems. By 1988 SASWITCH was handling about three million ATM transactions a month, according to the Competition Commission. The switch—now run by BankservAfrica—remains the backbone of South Africa's shared ATM network. == Net1 UEPS Technologies (1989–2017) == === Founding and UEPS === In 1989, Serge Belamant developed the Universal Electronic Payment System (UEPS), enabling secure, real-time transactions even in areas with limited connectivity. In the same year, he founded NET1 UEPS Technologies Inc., serving as its CEO and Director. === COPAC for VISA === In 1995, VISA tasked Belamant with designing the Chip Offline Pre-authorized Card (COPAC), a technology still widely used in chip-enabled credit and debit cards. A year later, he listed his company APLITEC (Applied Technology Holdings Limited) on the Johannesburg Stock Exchange. === Listings and acquisitions === In 1999, Belamant acquired Cash Payment Services (CPS) from First National Bank of South Africa, modernizing its welfare payment system to serve millions in rural areas. In 2005, he led NET1 Technologies to an IPO, listing it as NET1 UEPS Technologies Inc. on the Nasdaq. A secondary listing on the Johannesburg Stock Exchange (JSE) followed in 2008. === SASSA contract === Under Belamant's leadership, NET1 managed welfare payments for the South African Social Security Agency (SASSA), handling payments for over 10 million beneficiaries monthly. Despite criticism over handling the SASSA contract, investigations by the U.S. Department of Justice and the South African Constitutional Court found no wrongdoing. == Zilch (2018–present) == Belamant co-founded London-based "buy-now-pay-later" firm Zilch Technology in 2018 and serves as non-executive chair. Zilch reported £145 million in annual-recurring revenue and 4.5 million customers in January 2025. == Patents == Belamant is listed as inventor on more than a dozen payment-security patents, including: "Funds transfer system" (US RE36,788, 2000) – the basis for UEPS. "Financial transactions with a varying PIN" (WO 2014/037869, 2014).

  Read more →
• 

  Multiple encryption

  Multiple encryption is the process of encrypting an already encrypted message one or more times, either using the same or a different algorithm. It is also known as cascade encryption, cascade ciphering, cipher stacking, multiple encryption, and superencipherment. Superencryption refers to the outer-level encryption of a multiple encryption. Some cryptographers, like Matthew Green of Johns Hopkins University, say multiple encryption addresses a problem that mostly doesn't exist: Modern ciphers rarely get broken... You’re far more likely to get hit by malware or an implementation bug than you are to suffer a catastrophic attack on Advanced Encryption Standard (AES). However, from the previous quote an argument for multiple encryption can be made, namely poor implementation. Using two different cryptomodules and keying processes from two different vendors requires both vendors' wares to be compromised for security to fail completely. == Independent keys == Picking any two ciphers, if the key used is the same for both, the second cipher could possibly undo the first cipher, partly or entirely. This is true of ciphers where the decryption process is exactly the same as the encryption process (a reciprocal cipher) – the second cipher would completely undo the first. If an attacker were to recover the key through cryptanalysis of the first encryption layer, the attacker could possibly decrypt all the remaining layers, assuming the same key is used for all layers. To prevent that risk, one can use keys that are statistically independent for each layer (e.g. independent RNGs). Ideally each key should have separate and different generation, sharing, and management processes. == Independent Initialization Vectors == For en/decryption processes that require sharing an Initialization Vector (IV) / nonce these are typically, openly shared or made known to the recipient (and everyone else). Its good security policy never to provide the same data in both plaintext and ciphertext when using the same key and IV. Therefore, its recommended (although at this moment without specific evidence) to use separate IVs for each layer of encryption. == Importance of the first layer == With the exception of the one-time pad, no cipher has been theoretically proven to be unbreakable. Furthermore, some recurring properties may be found in the ciphertexts generated by the first cipher. Since those ciphertexts are the plaintexts used by the second cipher, the second cipher may be rendered vulnerable to attacks based on known plaintext properties (see references below). This is the case when the first layer is a program P that always adds the same string S of characters at the beginning (or end) of all ciphertexts (commonly known as a magic number). When found in a file, the string S allows an operating system to know that the program P has to be launched in order to decrypt the file. This string should be removed before adding a second layer. To prevent this kind of attack, one can use the method provided by Bruce Schneier: Generate a random pad R of the same size as the plaintext. Encrypt R using the first cipher and key. XOR the plaintext with the pad, then encrypt the result using the second cipher and a different (!) key. Concatenate both ciphertexts in order to build the final ciphertext. A cryptanalyst must break both ciphers to get any information. This will, however, have the drawback of making the ciphertext twice as long as the original plaintext. Note, however, that a weak first cipher may merely make a second cipher that is vulnerable to a chosen plaintext attack also vulnerable to a known plaintext attack. However, a block cipher must not be vulnerable to a chosen plaintext attack to be considered secure. Therefore, the second cipher described above is not secure under that definition, either. Consequently, both ciphers still need to be broken. The attack illustrates why strong assumptions are made about secure block ciphers and ciphers that are even partially broken should never be used. == The Rule of Two == The Rule of Two is a data security principle from the NSA's Commercial Solutions for Classified Program (CSfC). It specifies two completely independent layers of cryptography to protect data. For example, data could be protected by both hardware encryption at its lowest level and software encryption at the application layer. It could mean using two FIPS-validated software cryptomodules from different vendors to en/decrypt data. The importance of vendor and/or model diversity between the layers of components centers around removing the possibility that the manufacturers or models will share a vulnerability. This way if one components is compromised there is still an entire layer of encryption protecting the information at rest or in transit. The CSfC Program offers solutions to achieve diversity in two ways. "The first is to implement each layer using components produced by different manufacturers. The second is to use components from the same manufacturer, where that manufacturer has provided NSA with sufficient evidence that the implementations of the two components are independent of one another." The principle is practiced in the NSA's secure mobile phone called Fishbowl. The phones use two layers of encryption protocols, IPsec and Secure Real-time Transport Protocol (SRTP), to protect voice communications. The Samsung Galaxy S9 Tactical Edition is also an approved CSfC Component.

  Read more →
• 

  Data storage

  Data storage is the recording (storing) of information (data) in a storage medium. Handwriting, phonographic recording, magnetic tape, and optical discs are all examples of storage media. Biological molecules such as RNA and DNA are considered by some as data storage. Recording may be accomplished with virtually any form of energy. Electronic data storage requires electrical power to store and retrieve data. Data stored in a digital, machine-readable medium is called digital data. Computer data storage is one of the core functions of a general-purpose computer. Electronic documents can be stored in much less space than paper documents. Barcodes and magnetic ink character recognition (MICR) are two ways of recording machine-readable data on paper. == Recording media == A recording medium is physical material that holds information. Newly created information is distributed and can be stored in four storage media–print, film, magnetic, and optical–and seen or heard in four information flows–telephone, radio, TV, and the Internet as well as being observed directly. Digital information is stored on electronic media in many different recording formats. With electronic media, the data and the recording media are sometimes referred to as "software" despite the more common use of the word to describe computer software. With (traditional art) static media, art materials such as crayons may be considered both equipment and medium as the wax, charcoal or chalk material from the equipment becomes part of the surface of the medium. Some recording media may be temporary, either by design or by nature. Volatile organic compounds may be used to purposely make data expire over time or to reduce environmental impact. Data such as smoke signals or skywriting are temporary by nature. Depending on the volatility, a gas (e.g., atmosphere, smoke) or a liquid surface such as a lake would be considered a temporary recording medium, if it could be considered a recording medium at all. == Global capacity, digitization, and trends == A 2003 UC Berkeley report estimated that about five exabytes of new information were produced in 2002 and that 92% of this data was stored on magnetic media (primarily hard disk drives). This was about twice the data produced in 1999. The amount of data transmitted over telecommunications systems in 2002 was nearly 18 exabytes—three and a half times more than was recorded on non-volatile storage. Telephone calls constituted 98% of the telecommunicated information in 2002. The researchers' highest estimate for the growth rate of newly stored information (uncompressed) was more than 30% per year. In a more limited study, the International Data Corporation estimated that the total amount of digital data in 2007 was 281 exabytes and that the total amount of digital data produced exceeded the global storage capacity for the first time. A 2011 article in Science estimated that the year 2002 was the beginning of the digital age for information storage: an age in which more information is stored on digital storage devices than on analog storage devices. In 1986, approximately 1% of the world's capacity to store information was in digital format; this grew to 3% by 1993, to 25% by 2000, and to 94% by 2007. These figures correspond to less than three compressed exabytes in 1986, and 295 compressed exabytes in 2007. The quantity of digital storage doubled roughly every three to four years. It is estimated that around 120 zettabytes of data will be generated in 2023, an increase of 60x from 2010, and that it will increase to 181 zettabytes generated in 2025. == Mass storage ==

  Read more →
• 

  International Conference on Language Resources and Evaluation

  The International Conference on Language Resources and Evaluation is an international conference organised by the ELRA Language Resources Association every other year (on even years) with the support of institutions and organisations involved in Natural language processing. The series of LREC conferences was launched in Granada in 1998. == History of conferences == The survey of the LREC conferences over the period 1998-2013 was presented during the 2014 conference in Reykjavik as a closing session. It appears that the number of papers and signatures is increasing over time. The average number of authors per paper is higher as well. The percentage of new authors is between 68% and 78%. The distribution between male (65%) and female (35%) authors is stable over time. The most frequent technical term is "annotation", then comes "part-of-speech". == The LRE Map == The LRE Map was introduced at LREC 2010 and is now a regular feature of the LREC submission process for both the conference papers and the workshop papers. At the submission stage, the authors are asked to provide some basic information about all the resources (in a broad sense, i.e. including tools, standards and evaluation packages), either used or created, described in their papers. All these descriptors are then gathered in a global matrix called the LRE Map. This feature has been extended to several other conferences.

  Read more →
• 

  Viber

  Rakuten Viber, commonly known as Viber, is a cross-platform voice over IP (VoIP) and instant messaging (IM) software application owned by the Japanese technology company Rakuten Group. The service is available as freeware for Android, iOS, Microsoft Windows, macOS and Linux. Users are registered and identified through a mobile phone number, although the service can also be accessed on desktop platforms without mobile connectivity. In addition to instant messaging, the platform allows users to exchange media such as images, videos and files, and provides a paid international calling service called Viber Out. The software was launched in 2010 by the company Viber Media, founded by Talmon Marco and Igor Magazinnik. Rakuten acquired Viber Media in 2014 and later renamed the company Rakuten Viber. The company is headquartered in Cyprus and maintains offices in London, Manila, Paris, San Francisco, Singapore, Tokyo and Beijing. == History == === Founding (2010) === Viber Media was founded in Tel Aviv, Israel, in 2010 by Talmon Marco and Igor Magazinnik. Marco and Magazinnik are also co-founders of the peer-to-peer media and file-sharing client iMesh. The company was run from Israel and was registered in Cyprus. Sani Maroli and Ofer Smocha soon joined the company as well. Marco said Viber allows instant calling and synchronization with contacts because the ID is the user's cell number. In its early days, Viber relied on a patchwork of outsourcing partners from different countries, commissioning specific solutions from external vendors — including teams based in Cyprus and Belarus. According to the company's statements, development of Viber's core functionality historically originated from its Tel Aviv office — a testament to its roots — even though the legal entity was registered elsewhere. === Early monetisation (2011) === In its first two years of availability, Viber did not generate revenues. It began doing so in 2013, via user payments for Viber Out voice calling and the Viber graphical messaging "sticker market". The company was originally funded by individual investors, described by Marco as "friends and family". They invested $20 million in the company, which had 120 employees as of May 2013. On 24 July 2013, Viber's support system was defaced by the Syrian Electronic Army. According to Viber, no sensitive user information was accessed. By the time Rakuten came forward with its acquisition deal in 2014, Viber had already stopped working with external vendors, choosing instead to consolidate development under its own offices. === Rakuten acquires Viber (2014) === On 13 February 2014, Rakuten announced they had acquired Viber Media for $900 million, and since then Viber has been owned by Rakuten, Inc., an e-commerce conglomerate headquartered in Tokyo. The sale of Viber earned the Shabtai family (Benny, his brother Gilad, and Gilad's son Ofer) some $500 million from their 55.2% stake in the company. At that sale price, the founders each realized over 30 times return on their investments. Later that year, the company established a UK presence with the incorporation of Viber UK Limited in London. Djamel Agaoua became Viber Media CEO in February 2017, replacing co-founder Marco who left in 2015. In July 2017 the corporate name of Viber Media was changed to Rakuten Viber and a new wordmark logo was introduced. Its legal name remains Viber Media, S.à r.l. based in Luxembourg. === Post-acquisition === In August 2015 Viber opened a regional office for Central and Eastern Europe in Sofia to support growth in the region. In 2017, Rakuten Viber and the World Wildlife Fund engaged in a commercial transaction aimed at raising awareness and protecting wildlife. After first using Viber to spread its message in June 2020, the International Federation of the Red Cross launched an official chatbot and community on the messaging app to combat the spread of false information, which they termed an infodemic, about COVID-19. The chatbot is still active as of June 2022, with over 1.4 million subscribers. In 2020, Rakuten Viber and the World Health Organization (the WHO) engaged in a commercial transaction for a chatbot to inform users of issues such as women's health. and an anti-smoking campaign. In the wake of the July–August 2020 Belarusian election protests, to avoid sanctions and harassment from monopolies the company closed its office in Minsk. In 2022, Ofir Eyal became Viber CEO, replacing Djamel Agaoua. Eyal is a Viber veteran; he worked as Vice President of Product in 2014 before his promotion to Chief Operating Officer in 2019. Shortly after the appointment of a new CEO, Viber continued its international expansion. In March 2022, Rakuten announced the opening of a development center in Tbilisi, Georgia, intended to support work on mobile applications and technology projects in the region. In July 2022, Rakuten Viber partnered with Rapyd to launch instant cross-border P2P payments. The company launched payments on the Viber app first in Greece and Germany, and then in other countries. In August, Mineski teamed up with Viber to develop a social minigame platform that can play off Viber's application. In May 2022, Rakuten Viber launched the premium chat service Viber Plus that offers exclusive features, including sticker market privileges, ad-free use, priority Viber support, exclusive badge, unique Viber icon, large file sharing, and more. In 2022, Viber joined the European Union’s Code of Conduct on countering illegal hate speech online. As part of this framework, the company undertook to review reported content and remove material identified as hate speech in accordance with the Code and its platform rules. In January 2024 Rakuten (the company behind Viber) established an office in Kyiv to bring together engineering and marketing departments. Alongside launching its Kyiv office the company joined Diia.City as a resident. Subsequently in October 2024 Rakuten Viber inaugurated an office in Manila to broaden its operations, in the Philippines. The company’s legal entity remains Viber Media S.à r.l., registered in Luxembourg. Viber’s engineering work has been carried out across multiple countries and through external partners, including outsourcing and near-shore vendors. As a result, its development operations are distributed internationally rather than concentrated in a single location. In December 2024, Viber was blocked in Russia. Roskomnadzor announced the nationwide blocking of the messaging app due to non-compliance with local legal requirements. == Security audit == On 4 November 2014, Viber scored 1 out of 7 points on the Electronic Frontier Foundation's "Secure Messaging Scorecard". Viber received a point for encryption during transit but lost points because communications were not encrypted with keys that the provider did not have access to (i.e. the communications were not end-to-end encrypted), users could not verify contacts' identities, past messages were not secure if the encryption keys were stolen (i.e. the service did not provide forward secrecy), the code was not open to independent review (i.e. the code was not open-source), the security design was not properly documented, and there had not been a recent independent security audit. On 14 November 2014, the EFF changed Viber's score to 2 out of 7 after it had received an external security audit from Ernst & Young's Advanced Security Centre. On 19 April 2016, with the announcement of Viber version 6.0, Rakuten added end-to-end encryption to their service. The company said that the encryption protocol had only been audited internally, and promised to commission external audits "in the coming weeks". In May 2016, Viber published an overview of their encryption protocol, saying that it is a custom implementation that "uses the same concepts" as the Signal Protocol. In 2022, Rakuten Viber won a Security Award, by test.de, a tech firm based in Germany where there are over 3 million Viber users. In 2024, Rakuten Viber received SOC certification following an audit conducted by Ernst & Young. The certification relates to the company’s controls for data protection and information security. == Market share == As of December 2016, Viber had 800 million registered users. According to Statista, there are 260 million monthly active users as of January 2019. The Viber messenger is very popular in the Philippines, Greece, Eastern Europe, Russia, the Middle East, and some Asian markets. India was the largest market for Viber as of December 2014 with 33 million registered users, the fifth most popular instant messenger in the country. At the same time there were 30 million users in the United States, 28 million in Russia and 18 million in Brazil. Viber is particularly popular in Eastern Europe, being the most downloaded messaging app on Android in Belarus, Moldova and Ukraine as of 2016. It is also popular in Iraq, Libya and Nepal. Viber is translated in 44 languages and used in more than 190 co

  Read more →
• 

  Social commerce

  Social commerce is a subset of electronic commerce that involves social media and online media that supports social interaction, and user contributions to assist online buying and selling of products and services. More succinctly, social commerce is the use of social network(s), and user-generated content in the context of e-commerce transactions. The term social commerce was introduced by Yahoo! in November 2005 which describes a set of online collaborative shopping tools such as shared pick lists, user ratings and other user-generated content of online product information and advice. The concept of social commerce was developed by David Beisel to denote user-generated advertorial content on e-commerce sites, and by Steve Rubel to include collaborative e-commerce tools that enable shoppers "to get advice from trusted individuals, find goods and services and then purchase them". The social networks that spread this advice have been found to increase the customer's trust in one retailer over another. Social commerce may assist companies in achieving the following purposes: Firstly, social commerce helps companies engage customers with their brands according to the customers' social behaviors. Secondly, it provides an incentive for customers to return to their website. Thirdly, it provides customers with a platform to talk about their brand on their website. Fourthly, it provides all the information customers need to research, compare, and ultimately choose you over your competitor, thus purchasing from you and not others. In these days, the range of social commerce has been expanded to include social media tools and content used in the context of e-commerce, especially in the fashion industry. Examples of social commerce include customer ratings and reviews, user recommendations and referrals, social shopping tools (sharing the act of shopping online), forums and communities, social media optimization, social applications and social advertising. Technologies such as augmented reality have also been integrated with social commerce, allowing shoppers to visualize apparel items on themselves and solicit feedback through social media tools. Some academics have sought to distinguish "social commerce" from "social shopping", with the former being referred to as collaborative networks of online vendors; the latter, the collaborative activity of online shoppers. == Timeline == 2005: The term "social commerce" was first introduced on Yahoo! in 2005. 2021: The Global Web Index associated one's use of social media to his/her eagerness to buy. Social media with its entertaining and inspirational content can increase a product's profitability. This explains why Instagram expanded its Checkout feature to similar content like IG Stories, IGTV, and Reels. == Elements == The attraction and effectiveness of Social Commerce can be understood in terms of Robert Cialdini's Principles of InfluenceInfluence: Science and Practice": Reciprocity – When a company gives a person something for free, that person will feel the need to return the favor, whether by buying again or giving good recommendations for the company. Community – When people find an individual or a group that shares the same values, likes, beliefs, etc., they find community. People are more committed to a community that they feel accepted within. When this commitment happens, they tend to follow the same trends as a group and when one member introduces a new idea or product, it is accepted more readily based on the previous trust that has been established. It would be beneficial for companies to develop partnerships with social media sites to engage social communities with their products. Social proof – To receive positive feedback, a company needs to be willing to accept social feedback and to show proof that other people are buying, and like, the same things that I like. This can be seen in a lot of online companies such as eBay and Amazon, that allow public feedback of products and when a purchase is made, they immediately generate a list showing purchases that other people have made in relation to my recent purchase. It is beneficial to encourage open recommendation and feedback. This creates trust for you as a seller. 55% of buyers turn to social media when they're looking for information. Authority – Many people need proof that a product is of good quality. This proof can be based on the recommendations of others who have bought the same product. If there are many user reviews about a product, then a consumer will be more willing to trust their own decision to buy this item. Liking – People trust based on the recommendations of others. If there are a lot of "likes" of a particular product, then the consumer will feel more confident and justified in making this purchase. Scarcity – As part of supply and demand, a greater value is assigned to products that are regarded as either being in high demand or are seen as being in a shortage. Therefore, if a person is convinced that they are purchasing something that is unique, special, or not easy to acquire, they will have more of a willingness to make a purchase. If there is trust established from the seller, they will want to buy these items immediately. This can be seen in the cases of Zara and Apple Inc. who create demand for their products by convincing the public that there is a possibility of missing out on being able to purchase them. == Types == === Onsite === Onsite social commerce refers to retailers including social sharing and other social functionality on their website. Some notable examples include Zazzle which enables users to share their purchases, Macy's which allows users to create a poll to find the right product, and Fab.com which shows a live feed of what other shoppers are buying. Onsite user reviews are also considered a part of social commerce. This approach has been successful in improving customer engagement, conversion and word-of-mouth branding according to several industry sources. === Offsite === Offsite social commerce includes activities that happen outside of the retailers' website. This may include posting products on social networks such as Facebook, X, and TikTok. It may also include advertising on shopping forums such as SlickDeals, Red Flag Deals, and LatestDeals.co.uk. == Measurements == Social commerce can be measured by any of the principle ways to measure social media. Return on Investment: measures the effect or action of social media on sales. Reputation: indices measure the influence of social media investment in terms of changes to online reputation – made up of the volume and valence of social media mentions. Reach: metrics use traditional media advertising metrics to measure the exposure rates and levels of an audience with social media. == Business applications == This category is based on individuals' shopping, selling, recommending behaviors. Social network-driven sales (Soldsie) – Facebook commerce and Twitter commerce belong to this part. Sales take place on established social network sites. Peer-to-peer sales platforms (eBay, Etsy, Amazon) – In these websites, users can directly communicate and sell products to other users. Group buying (Groupon, LivingSocial) – Users can buy products or services at a lower price when enough users agree to make this purchase. Peer recommendations and reviews (Amazon, Yelp, Bazaarvoice) – Users can see recommendations and reviews from other users. User-curated shopping (The Fancy, Lyst) – Users create and share lists of products and services for others to shop from. Participatory commerce (Betabrand, Threadless, Kickstarter) – Users can get involved in the production process. Social shopping (Squadded) – Allowing e-commerce to provide their users live chat sessions and shared shopping lists so they can communicate with their friends or other shoppers for advice. == Business examples == Here are some notable business examples of Social Commerce: Betabrand: an online brand using participatory design to release new, community-created ideas every week. Cafepress: an online retailer of stock and user-customized on demand products. Etsy: an e-commerce website focused on handmade or vintage items and supplies, as well as unique factory-manufactured items under Etsy's new guidelines. Eventbrite: an online ticketing service that allows event organizers to plan, set up ticket sales and promote events (event management) and publish them across Facebook, Twitter and other social-networking tools directly from the site's interface. Groupon: a deal-of-the-day website that features discounted gift certificates usable at local or national companies. Houzz: a web site and online community about architecture, interior design and decorating, landscape design and home improvement. LivingSocial: an online marketplace that allows clients to buy and share things to do in their city. Lockerz: an international social commerce website based in Seattle, Washington. OpenSky: is a r

  Read more →
• 

  MDS matrix

  An MDS matrix (maximum distance separable) is a matrix representing a function with certain diffusion properties that have useful applications in cryptography. Technically, an m × n {\displaystyle m\times n} matrix A {\displaystyle A} over a finite field K {\displaystyle K} is an MDS matrix if it is the transformation matrix of a linear transformation f ( x ) = A x {\displaystyle f(x)=Ax} from K n {\displaystyle K^{n}} to K m {\displaystyle K^{m}} such that no two different ( m + n ) {\displaystyle (m+n)} -tuples of the form ( x , f ( x ) ) {\displaystyle (x,f(x))} coincide in n {\displaystyle n} or more components. Equivalently, the set of all ( m + n ) {\displaystyle (m+n)} -tuples ( x , f ( x ) ) {\displaystyle (x,f(x))} is an MDS code, i.e., a linear code that reaches the Singleton bound. Let A ~ = ( I n A ) {\displaystyle {\tilde {A}}={\begin{pmatrix}\mathrm {I} _{n}\\\hline \mathrm {A} \end{pmatrix}}} be the matrix obtained by joining the identity matrix I n {\displaystyle \mathrm {I} _{n}} to A {\displaystyle A} . Then a necessary and sufficient condition for a matrix A {\displaystyle A} to be MDS is that every possible n × n {\displaystyle n\times n} submatrix obtained by removing m {\displaystyle m} rows from A ~ {\displaystyle {\tilde {A}}} is non-singular. This is also equivalent to the following: all the sub-determinants of the matrix A {\displaystyle A} are non-zero. Then a binary matrix A {\displaystyle A} (namely over the field with two elements) is never MDS unless it has only one row or only one column with all components 1 {\displaystyle 1} . Reed–Solomon codes have the MDS property and are frequently used to obtain the MDS matrices used in cryptographic algorithms. Serge Vaudenay suggested using MDS matrices in cryptographic primitives to produce what he called multipermutations, not-necessarily linear functions with this same property. These functions have what he called perfect diffusion: changing t {\displaystyle t} of the inputs changes at least m − t + 1 {\displaystyle m-t+1} of the outputs. He showed how to exploit imperfect diffusion to cryptanalyze functions that are not multipermutations. MDS matrices are used for diffusion in such block ciphers as AES, SHARK, Square, Twofish, Anubis, KHAZAD, Manta, Hierocrypt, Kalyna, Camellia and HADESMiMC, and in the stream cipher MUGI and the cryptographic hash function Whirlpool, Poseidon.

  Read more →
• 

  Whitelist

  A whitelist or allowlist is a list or register of entities that are being provided a particular privilege, service, mobility, access or recognition. Entities on the list will be accepted, approved and/or recognized. Whitelisting is the reverse of blacklisting, the practice of identifying entities that are denied, unrecognized, or ostracized. == Email whitelists == Spam filters often include the ability to "whitelist" certain sender IP addresses, email addresses or domain names to protect their email from being rejected or sent to a junk mail folder. These can be manually maintained by the user or system administrator - but can also refer to externally maintained whitelist services. === Non-commercial whitelists === Non-commercial whitelists are operated by various non-profit organizations, ISPs, and others interested in blocking spam. Rather than paying fees, the sender must pass a series of tests; for example, their email server must not be an open relay and have a static IP address. The operator of the whitelist may remove a server from the list if complaints are received. === Commercial whitelists === Commercial whitelists are a system by which an Internet service provider allows someone to bypass spam filters when sending email messages to its subscribers, in return for a pre-paid fee, either an annual or a per-message fee. A sender can then be more confident that their messages have reached recipients without being blocked, or having links or images stripped out of them, by spam filters. The purpose of commercial whitelists is to allow companies to reliably reach their customers by email. == Advertising whitelist == Many websites rely on ads as a source of revenue, but the use of ad blockers is increasingly common. Websites that detect an adblocker in use often ask for it to be disabled - or their site to be "added to the whitelist" - a standard feature of most adblockers. == Network whitelists == === LAN whitelists === A use for whitelists is in local area network (LAN) security. Many network admins set up MAC address whitelists, or a MAC address filter, to control who is allowed on their networks. This is used when encryption is not a practical solution or in tandem with encryption. However, it's sometimes ineffective because a MAC address can be faked. === IP whitelist === Firewalls can usually be configured to only allow data-traffic from/to certain (ranges of) IP-addresses. === Application whitelists === One approach in combating viruses and malware is to whitelist software which is considered safe to run, blocking all others. This is particularly attractive in a corporate environment, where there are typically already restrictions on what software is approved. Leading providers of application whitelisting technology include Bit9, Velox, McAfee, Lumension, ThreatLocker, Airlock Digital and SMAC. On Microsoft Windows, recent versions include AppLocker, which allows administrators to control which executable files are denied or allowed to execute. With AppLocker, administrators are able to create rules based on file names, publishers or file location that will allow certain files to execute. Rules can apply to individuals or groups. Policies are used to group users into different enforcement levels. For example, some users can be added to a report-only policy that will allow administrators to understand the impact before moving that user to a higher enforcement level. Linux systems typically have AppArmor and SE Linux features available which can be used to effectively block all applications which are not explicitly whitelisted, and commercial products are also available. On HP-UX introduced a feature called "HP-UX Whitelisting" on 11iv3 version. == Controversy regarding name == In 2018, a journal commentary on a report on predatory publishing was released making claims that "white" and "black" are racially charged terms that need to be avoided in instances such as "whitelist" and "blacklist". The premise of the journal is that "black" and "white" have negative and positive connotations respectively. It states that since "blacklisting" was first referred to during "the time of mass enslavement and forced deportation of Africans to work in European-held colonies in the Americas," the word is therefore related to race. There is no mention of "whitelist" and its origin or relation to race. This issue is most widely disputed in computing industries where "whitelist" and "blacklist" are prevalent (e.g. IP whitelisting). Despite the commentary nature of the journal, some companies and individuals in others have taken to replacing "whitelist" and "blacklist" with new alternatives such as "allow list" and "deny list". Those adopting this change consider using the "whitelist"/"blacklist" names as a code smell. Those that oppose these changes question its attribution to race, citing the same etymology quote that the 2018 journal uses. According to the remark, the term "blacklist" evolved from the term "black book" about a century ago. The term "black book" does not appear to have any etymology or sources that support racial associations, instead originating in the 1400s as a reference to "a list of people who had committed crimes or fallen out of favor with leaders", and popularized by King Henry VIII's literal use of a black book. Others also note the prevalence of positive and negative connotations to "white" and "black" in the Bible, predating attributions to skin tone and slavery. It wasn't until the 1960s Black Power movement that "Black" became a widespread word to refer to one's race as a person of color in America (alternate to African-American) lending itself to the argument that the negative connotation behind "black" and "blacklist" both predate attribution to race.

  Read more →
• 

  Data monetization

  Data monetization, a form of monetization, may refer to the act of generating measurable economic benefits from available data sources (analytics). Less commonly, it may also refer to the act of monetizing data services. In the case of analytics, typically, these benefits accrue as revenue or expense savings, but may also include market share or corporate market value gains. Data monetization leverages data generated through business operations, available exogenous data or content, as well as data associated with individual actors such as that collected via electronic devices and sensors participating in the internet of things. For example, the ubiquity of the internet of things is generating location data and other data from sensors and mobile devices at an ever-increasing rate. When this data is collated against traditional databases, the value and utility of both sources of data increases, leading to tremendous potential to mine data for social good, research and discovery, and achievement of business objectives. Closely associated with data monetization are the emerging data as a service models for transactions involving data by the data item. There are three ethical and regulatory vectors involved in data monetization due to the sometimes conflicting interests of actors involved in the digital supply chain. The individual data creator who generates files and records through his own efforts or owns a device such as a sensor or a mobile phone that generates data has a claim to ownership of data. The business entity that generates data in the course of its operations, such as its transactions with financial institutions or risk factors discovered through feedback from customers also has a claim on data captured through their systems and platforms. However, the person that contributed the data may also have a legitimate claim on the data. Internet platforms and service providers, such as Google or Facebook that require a user to forgo some ownership interest in their data in exchange for use of the platform also have a legitimate claim on the data. Thus the practice of data monetization, although common since 2000, is now getting increasing attention from regulators. The European Union and the United States Congress have begun to address these issues. For instance, in the financial services industry, regulations involving data are included in the Gramm–Leach–Bliley Act and Dodd-Frank. Some individual creators of data are shifting to using personal data vaults and implementing vendor relationship management concepts as a reflection of an increasing resistance to their data being federated or aggregated and resold without compensation. Groups such as the Personal Data Ecosystem Consortium, Patient privacy rights, and others are also challenging corporate cooptation of data without compensation. Financial services companies are a relatively good example of an industry focused on generating revenue by leveraging data. Credit card issuers and retail banks use customer transaction data to improve targeting of cross-sell offers. Partners are increasingly promoting merchant based reward programs which leverage a bank’s data and provide discounts to customers at the same time. == Types of data monetization == Internal data monetization - An organization's data is used internally, resulting in economic benefit. This is commonly the case in organizations using analytics to uncover insights, resulting in improved profit, cost savings or the avoidance of risk. Internal data monetization is currently the most common form of monetization, requiring far fewer security, intellectual property, and legal precautions when compared to other types. The potential economic gains from this type of data monetization are limited by the organization's internal structure and situation. External data monetization - A person or organization makes data they possess available on a for-fee basis to external parties, or as a broker for same. This type of monetization is less common and requires various methods to distribute the data to potential buyers and consumers. However, the economic gain that results from collecting data, packaging and distributing it, can be quite large. == Steps == Identification of available data sources – this includes data currently available for monetization as well as other external data sources that may enhance the value of what’s currently available. Connect, aggregate, attribute, validate, authenticate, and exchange data - this allows data to be converted directly into actionable or revenue generating insight or services. Set terms and prices and facilitate data trading - methods for data vetting, storage, and access. For example, many global corporations have locked and siloed data storage infrastructures, which hinders efficient access to data and cooperative and real-time exchange. Perform Research and analytics – draw predictive insights from existing data as a basis for using data for to reduce risk, enhance product development or performance, or improve customer experience or business outcomes. Action and leveraging – the last phase of monetizing data includes determining alternative or improved data centric products, ideas, or services. Examples may include real-time actionable triggered notifications or enhanced channels such as web or mobile response mechanisms. == Pricing variables and factors == A fee for use of a platform to connect buyers and sellers use of a platform to configure, organize, and otherwise process data included in a data trade connecting or including a device or sensor into a data supply chain connecting and credentialing a creator of a data source and a data buyer – often through a federated identity connecting a data source to other data sources to be included in a data supply chain use of an internet service or other transmission services for uploading and downloading data – sometimes, for an individual, through a personal cloud use of encrypted keys to achieve secure data transfer use of a search algorithm specifically designed to tag data sources that contain data points of value to the data buyer linking a data creator or generator to a data collection protocol or form server actions – such as a notification – triggered by an update to a data item or data source included in a data supply chain A price or exchange or other trade value assigned by a data creator or generator to a data item or a data source offered by a data buyer to a data creator assigned by a data buyer for a data item or a data source formatted according to criteria set by a data buyer An incremental fee assigned by a data buyer for a data item or a data set scaled to the reputation of the data creator == Benefits == Improved decision-making that leads to real time crowd sourced research, improved profits, decreased costs, reduced risk and improved compliance More impactful decisions (e.g., make real-time decisions) More timely (lower latency) decisions (e.g., a vendor making purchase recommendations while the customer is still on the phone or in the store, a customer connecting with multiple vendors to discover the best price, triggered notifications when thresholds are reached for data values) More granular decisions (e.g., localized pricing decisions at an individual or device or sensor level versus larger aggregates). Targeted Marketing (e.g., Vendors with access to big data can make targeted advertisements to specific customers within a set data pool decreasing costs for the advertiser and reaching most interested customers) == Frameworks == There are a wide variety of industries, firms and business models related to data monetization. The following frameworks have been offered to help understand the types of business models that are used: Roger Ehrenberg of IA Ventures, a venture capital firm that invests in this sector, has defined three basic types of data product firms: Contributory databases. The magic of these businesses is that a customer provides their own data in exchange for receiving a more robust set of aggregated data back that provides insight into the broader marketplace, or provides a vehicle for expressing a view. Give a little, get a lot back in return – a pretty compelling value proposition, and one that frequently results in a payment from the data contributor in exchange for receiving enriched, aggregated data. Once these contributory databases are developed and customers become reliant on their insights, they become extremely valuable and persistent data assets. Data processing platforms. These businesses create barriers through a combination of complex data architectures, proprietary algorithms, and rich analytics to help customers consume data in whatever form they please. Often these businesses have special relationships with key data providers, that when combined with other data and processed as a whole create valuable differentiation and competitive barriers. Bloomberg is an example of a powerful

  Read more →
• 

  Brooklyn Bridge (software)

  The Brooklyn Bridge from White Crane Systems was a data transfer enabler. Although it came with some hardware, it was the software which was the basis of the product. It also could transform the data's format. == Overview == The New York Times described its category as being among "communications packages used to transfer files." In an era of 300 baud, Brooklyn Bridge operated at "115,200 baud" so that a transfer which "at 300 baud took 4 minutes and 36 seconds" only needed 5 seconds. Unlike some communications packages, this one retains the original version-date, so as not to alarm people when they seem to have what looks like an update, when it's not. == Description == Once the software is installed, users comfortable with typing the word "COPY" can do so as readily as they sneakernet. An earlier review described it as "less cumbersome than conventional communications software" The use of neither specialized hardware nor specialized software is ideal in an era when this can be done using online or other "outside" services.

  Read more →
• 

  Data proliferation

  Data proliferation refers to the prodigious amount of data, structured and unstructured, that businesses and governments continue to generate at an unprecedented rate and the usability problems that result from attempting to store and manage that data. While originally pertaining to problems associated with paper documentation, data proliferation has become a major problem in primary and secondary data storage on computers. While digital storage has become cheaper, the associated costs, from raw power to maintenance and from metadata to search engines, have not kept up with the proliferation of data. Although the power required to maintain a unit of data has fallen, the cost of facilities which house the digital storage has tended to rise. Data proliferation has been documented as a problem for the U.S. military since August 1971, in particular regarding the excessive documentation submitted during the acquisition of major weapon systems. Efforts to mitigate data proliferation and the problems associated with it are ongoing. == Problems caused == The problem of data proliferation is affecting all areas of commerce as a result of the availability of relatively inexpensive data storage devices. This has made it very easy to dump data into secondary storage immediately after its window of usability has passed. This masks problem that could gravely affect the profitability of businesses and the efficient functioning of health services, police and security forces, local and national governments, and many other types of organizations. Data proliferation is problematic for several reasons: Difficulty when trying to find and retrieve information. At Xerox, on average it takes employees more than one hour per week to find hard-copy documents, costing $2,152 a year to manage and store them. For businesses with more than 10 employees, this increases to almost two hours per week at $5,760 per year. In large networks of primary and secondary data storage, problems finding electronic data are analogous to problems finding hard copy data. Data loss and legal liability when data is disorganized, not properly replicated, or cannot be found promptly. In April 2005, the Ameritrade Holding Corporation told 200,000 current and past customers that a tape containing confidential information had been lost or destroyed in transit. In May of the same year, Time Warner Incorporated reported that 40 tapes containing personal data on 600,000 current and former employees had been lost en route to a storage facility. In March 2005, a Florida judge hearing a $2.7 billion lawsuit against Morgan Stanley issued an "adverse inference order" against the company for "willful and gross abuse of its discovery obligations." The judge cited Morgan Stanley for repeatedly finding misplaced tapes of e-mail messages long after the company had claimed that it had turned over all such tapes to the court. Increased manpower requirements to manage increasingly chaotic data storage resources. Slower networks and application performance due to excess traffic as users search and search again for the material they need. High cost in terms of the energy resources required to operate storage hardware. A 100 terabyte system will cost up to $35,040 a year to run—not counting cooling costs. == Proposed solutions == Applications that better utilize modern technology Reductions in duplicate data (especially as caused by data movement) Improvement of metadata structures Improvement of file and storage transfer structures User education and discipline The implementation of Information Lifecycle Management solutions to eliminate low-value information as early as possible before putting the rest into actively managed long-term storage in which it can be quickly and cheaply accessed.

  Read more →
• 

  Residual neural network

  A residual neural network (also referred to as a residual network or ResNet) is a deep learning architecture in which the layers learn residual functions with reference to the layer inputs. It was developed in 2015 for image recognition, and won the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) of that year. As a point of terminology, "residual connection" refers to the specific architectural motif of x ↦ f ( x ) + x {\displaystyle x\mapsto f(x)+x} , where f {\displaystyle f} is an arbitrary neural network module. The motif had been used previously (see §History for details). However, the publication of ResNet made it widely popular for feedforward networks, appearing in neural networks that are seemingly unrelated to ResNet. The residual connection stabilizes the training and convergence of deep neural networks with hundreds of layers, and is a common motif in deep neural networks, such as transformer models (e.g., BERT, and GPT models such as ChatGPT), the AlphaGo Zero system, the AlphaStar system, and the AlphaFold system. == Mathematics == === Residual connection === In a multilayer neural network model, consider a (non-residual) subnetwork with a certain number of stacked layers (e.g., 2 or 3). Let H ( x ; α ) {\displaystyle H(x;\alpha )} denote the subnetwork. Suppose H ∗ {\displaystyle H^{}} is the desired optimal output of this subnetwork. Residual learning simply adds x {\displaystyle x} directly to the output, such that the optimal learned output now becomes be H ∗ − x {\displaystyle H^{}-x} , which is interpreted as a "residual" with respect to x {\displaystyle x} . The operation of "adding x {\displaystyle x} " is implemented via a "skip connection" that performs an identity mapping to connect the input of the subnetwork with its output. This connection is referred to as a "residual connection" in later work. Let F ( x ; α ) = H ( x ; a ) + x {\displaystyle F(x;\alpha )=H(x;a)+x} . The function F {\displaystyle F} is often represented by matrix multiplication interlaced with activation functions and normalization operations (e.g., batch normalization or layer normalization). As a whole, one of these subnetworks is referred to as a "residual block". A deep residual network is constructed by simply stacking these blocks. Long short-term memory (LSTM) has a memory mechanism that serves as a residual connection. In an LSTM without a forget gate, an input x t {\displaystyle x_{t}} is processed by a function F {\displaystyle F} and added to a memory cell c t {\displaystyle c_{t}} , resulting in c t + 1 = c t + F ( x t ) {\displaystyle c_{t+1}=c_{t}+F(x_{t})} . An LSTM with a forget gate essentially functions as a highway network. To stabilize the variance of the layers' inputs, it is recommended to replace the residual connections x + f ( x ) {\displaystyle x+f(x)} with x / L + f ( x ) {\displaystyle x/L+f(x)} , where L {\displaystyle L} is the total number of residual layers. === Projection connection === If the function F {\displaystyle F} is of type F : R n → R m {\displaystyle F:\mathbb {R} ^{n}\to \mathbb {R} ^{m}} where n ≠ m {\displaystyle n\neq m} , then F ( x ) + x {\displaystyle F(x)+x} is undefined. To handle this special case, a projection connection is used: y = F ( x ) + P ( x ) {\displaystyle y=F(x)+P(x)} where P {\displaystyle P} is typically a linear projection, defined by P ( x ) = M x {\displaystyle P(x)=Mx} where M {\displaystyle M} is a m × n {\displaystyle m\times n} matrix. The matrix is trained via backpropagation, as is any other parameter of the model. === Signal propagation === The introduction of identity mappings facilitates signal propagation in both forward and backward paths. ==== Forward propagation ==== If the output of the ℓ {\displaystyle \ell } -th residual block is the input to the ( ℓ + 1 ) {\displaystyle (\ell +1)} -th residual block (assuming no activation function between blocks), then the ( ℓ + 1 ) {\displaystyle (\ell +1)} -th input is: x ℓ + 1 = F ( x ℓ ) + x ℓ {\displaystyle x_{\ell +1}=F(x_{\ell })+x_{\ell }} Applying this formulation recursively, e.g.: x ℓ + 2 = F ( x ℓ + 1 ) + x ℓ + 1 = F ( x ℓ + 1 ) + F ( x ℓ ) + x ℓ {\displaystyle {\begin{aligned}x_{\ell +2}&=F(x_{\ell +1})+x_{\ell +1}\\&=F(x_{\ell +1})+F(x_{\ell })+x_{\ell }\end{aligned}}} yields the general relationship: x L = x ℓ + ∑ i = ℓ L − 1 F ( x i ) {\displaystyle x_{L}=x_{\ell }+\sum _{i=\ell }^{L-1}F(x_{i})} where L {\textstyle L} is the index of a residual block and ℓ {\textstyle \ell } is the index of some earlier block. This formulation suggests that there is always a signal that is directly sent from a shallower block ℓ {\textstyle \ell } to a deeper block L {\textstyle L} . ==== Backward propagation ==== The residual learning formulation provides the added benefit of mitigating the vanishing gradient problem to some extent. However, it is crucial to acknowledge that the vanishing gradient issue is not the root cause of the degradation problem, which is tackled through the use of normalization. To observe the effect of residual blocks on backpropagation, consider the partial derivative of a loss function E {\displaystyle {\mathcal {E}}} with respect to some residual block input x ℓ {\displaystyle x_{\ell }} . Using the equation above from forward propagation for a later residual block L > ℓ {\displaystyle L>\ell } : ∂ E ∂ x ℓ = ∂ E ∂ x L ∂ x L ∂ x ℓ = ∂ E ∂ x L ( 1 + ∂ ∂ x ℓ ∑ i = ℓ L − 1 F ( x i ) ) = ∂ E ∂ x L + ∂ E ∂ x L ∂ ∂ x ℓ ∑ i = ℓ L − 1 F ( x i ) {\displaystyle {\begin{aligned}{\frac {\partial {\mathcal {E}}}{\partial x_{\ell }}}&={\frac {\partial {\mathcal {E}}}{\partial x_{L}}}{\frac {\partial x_{L}}{\partial x_{\ell }}}\\&={\frac {\partial {\mathcal {E}}}{\partial x_{L}}}\left(1+{\frac {\partial }{\partial x_{\ell }}}\sum _{i=\ell }^{L-1}F(x_{i})\right)\\&={\frac {\partial {\mathcal {E}}}{\partial x_{L}}}+{\frac {\partial {\mathcal {E}}}{\partial x_{L}}}{\frac {\partial }{\partial x_{\ell }}}\sum _{i=\ell }^{L-1}F(x_{i})\end{aligned}}} This formulation suggests that the gradient computation of a shallower layer, ∂ E ∂ x ℓ {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{\ell }}}} , always has a later term ∂ E ∂ x L {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{L}}}} that is directly added. Even if the gradients of the F ( x i ) {\displaystyle F(x_{i})} terms are small, the total gradient ∂ E ∂ x ℓ {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{\ell }}}} resists vanishing due to the added term ∂ E ∂ x L {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{L}}}} . == Variants of residual blocks == === Basic block === A basic block is the simplest building block studied in the original ResNet. This block consists of two sequential 3x3 convolutional layers and a residual connection. The input and output dimensions of both layers are equal. === Bottleneck block === A bottleneck block consists of three sequential convolutional layers and a residual connection. The first layer in this block is a 1×1 convolution for dimension reduction (e.g., to 1/2 of the input dimension); the second layer performs a 3×3 convolution; the last layer is another 1×1 convolution for dimension restoration. The models of ResNet-50, ResNet-101, and ResNet-152 are all based on bottleneck blocks. === Pre-activation block === The pre-activation residual block applies activation functions before applying the residual function F {\displaystyle F} . Formally, the computation of a pre-activation residual block can be written as: x ℓ + 1 = F ( ϕ ( x ℓ ) ) + x ℓ {\displaystyle x_{\ell +1}=F(\phi (x_{\ell }))+x_{\ell }} where ϕ {\displaystyle \phi } can be any activation (e.g. ReLU) or normalization (e.g. LayerNorm) operation. This design reduces the number of non-identity mappings between residual blocks, and allows an identity mapping directly from the input to the output. This design was used to train models with 200 to over 1000 layers, and was found to consistently outperform variants where the residual path is not an identity function. The pre-activation ResNet with 200 layers took 3 weeks to train for ImageNet on 8 GPUs in 2016. Since GPT-2, transformer blocks have been mostly implemented as pre-activation blocks. This is often referred to as "pre-normalization" in the literature of transformer models. == Applications == Originally, ResNet was designed for computer vision. All transformer architectures include residual connections. Indeed, very deep transformers cannot be trained without them. The original ResNet paper made no claim on being inspired by biological systems. However, later research has related ResNet to biologically-plausible algorithms. A study published in Science in 2023 disclosed the complete connectome of an insect brain (specifically that of a fruit fly larva). This study discovered "multilayer shortcuts" that resemble the skip connections in artificial neural networks, including ResNets. == History == === Previous work === Residual connections were noticed in neu

  Read more →
• 

  Knapsack problem

  The knapsack problem is the following problem in combinatorial optimization: Given a set of items, each with a weight and a value, determine which items to include in the collection so that the total weight is less than or equal to a given limit and the total value is as large as possible. It derives its name from the problem faced by someone who is constrained by a fixed-size knapsack and must fill it with the most valuable items. The problem often arises in resource allocation where the decision-makers have to choose from a set of non-divisible projects or tasks under a fixed budget or time constraint, respectively. The knapsack problem has been studied for more than a century, with early works dating back to 1897. The subset sum problem is a special case of the decision and 0-1 problems where for each kind of item, the weight equals the value: w i = v i {\displaystyle w_{i}=v_{i}} . In the field of cryptography, the term knapsack problem is often used to refer specifically to the subset sum problem. The subset sum problem is one of Karp's 21 NP-complete problems. == Applications == Knapsack problems appear in real-world decision-making processes in a wide variety of fields, such as finding the least wasteful way to cut raw materials, selection of investments and portfolios, selection of assets for asset-backed securitization, and generating keys for the Merkle–Hellman and other knapsack cryptosystems. One early application of knapsack algorithms was in the construction and scoring of tests in which the test-takers have a choice as to which questions they answer. For small examples, it is a fairly simple process to provide the test-takers with such a choice. For example, if an exam contains 12 questions each worth 10 points, the test-taker need only answer 10 questions to achieve a maximum possible score of 100 points. However, on tests with a heterogeneous distribution of point values, it is more difficult to provide choices. Feuerman and Weiss proposed a system in which students are given a heterogeneous test with a total of 125 possible points. The students are asked to answer all of the questions to the best of their abilities. Of the possible subsets of problems whose total point values add up to 100, a knapsack algorithm would determine which subset gives each student the highest possible score. A 1999 study of the Stony Brook University Algorithm Repository showed that, out of 75 algorithmic problems related to the field of combinatorial algorithms and algorithm engineering, the knapsack problem was the 19th most popular and the third most needed after suffix trees and the bin packing problem. == Definition == The most common problem being solved is the 0-1 knapsack problem, which restricts the number x i {\displaystyle x_{i}} of copies of each kind of item to zero or one. Given a set of n {\displaystyle n} items numbered from 1 up to n {\displaystyle n} , each with a weight w i {\displaystyle w_{i}} and a value v i {\displaystyle v_{i}} , along with a maximum weight capacity W {\displaystyle W} , maximize ∑ i = 1 n v i x i {\displaystyle \sum _{i=1}^{n}v_{i}x_{i}} subject to ∑ i = 1 n w i x i ≤ W {\displaystyle \sum _{i=1}^{n}w_{i}x_{i}\leq W} and x i ∈ { 0 , 1 } {\displaystyle x_{i}\in \{0,1\}} . Here x i {\displaystyle x_{i}} represents the number of instances of item i {\displaystyle i} to include in the knapsack. Informally, the problem is to maximize the sum of the values of the items in the knapsack so that the sum of the weights is less than or equal to the knapsack's capacity. The bounded knapsack problem (BKP) removes the restriction that there is only one of each item, but restricts the number x i {\displaystyle x_{i}} of copies of each kind of item to a maximum non-negative integer value c {\displaystyle c} : maximize ∑ i = 1 n v i x i {\displaystyle \sum _{i=1}^{n}v_{i}x_{i}} subject to ∑ i = 1 n w i x i ≤ W {\displaystyle \sum _{i=1}^{n}w_{i}x_{i}\leq W} and x i ∈ { 0 , 1 , 2 , … , c } . {\displaystyle x_{i}\in \{0,1,2,\dots ,c\}.} The unbounded knapsack problem (UKP) places no upper bound on the number of copies of each kind of item and can be formulated as above except that the only restriction on x i {\displaystyle x_{i}} is that it is a non-negative integer. maximize ∑ i = 1 n v i x i {\displaystyle \sum _{i=1}^{n}v_{i}x_{i}} subject to ∑ i = 1 n w i x i ≤ W {\displaystyle \sum _{i=1}^{n}w_{i}x_{i}\leq W} and x i ∈ N . {\displaystyle x_{i}\in \mathbb {N} .} One example of the unbounded knapsack problem is given using the figure shown at the beginning of this article and the text "if any number of each book is available" in the caption of that figure. == Computational complexity == The knapsack problem is interesting from the perspective of computer science for many reasons: The decision problem form of the knapsack problem (Can a value of at least V be achieved without exceeding the weight W?) is NP-complete, thus there is no known algorithm that is both correct and fast (polynomial-time) in all cases. There is no known polynomial algorithm which can tell, given a solution, whether it is optimal (which would mean that there is no solution with a larger V). This problem is co-NP-complete. There is a pseudo-polynomial time algorithm using dynamic programming. There is a fully polynomial-time approximation scheme, which uses the pseudo-polynomial time algorithm as a subroutine, described below. Many cases that arise in practice, and "random instances" from some distributions, can nonetheless be solved exactly. There is a link between the "decision" and "optimization" problems in that if there exists a polynomial algorithm that solves the "decision" problem, then one can find the maximum value for the optimization problem in polynomial time by applying this algorithm iteratively while increasing the value of k. On the other hand, if an algorithm finds the optimal value of the optimization problem in polynomial time, then the decision problem can be solved in polynomial time by comparing the value of the solution output by this algorithm with the value of k. Thus, both versions of the problem are of similar difficulty. One theme in research literature is to identify what the "hard" instances of the knapsack problem look like, or viewed another way, to identify what properties of instances in practice might make them more amenable than their worst-case NP-complete behaviour suggests. The goal in finding these "hard" instances is for their use in public-key cryptography systems, such as the Merkle–Hellman knapsack cryptosystem. More generally, better understanding of the structure of the space of instances of an optimization problem helps to advance the study of the particular problem and can improve algorithm selection. Furthermore, notable is the fact that the hardness of the knapsack problem depends on the form of the input. If the weights and profits are given as integers, it is weakly NP-complete, while it is strongly NP-complete if the weights and profits are given as rational numbers. However, in the case of rational weights and profits it still admits a fully polynomial-time approximation scheme. === Unit-cost models === The NP-hardness of the Knapsack problem relates to computational models in which the size of integers matters (such as the Turing machine). In contrast, decision trees count each decision as a single step. Dobkin and Lipton show an 1 2 n 2 {\displaystyle {1 \over 2}n^{2}} lower bound on linear decision trees for the knapsack problem, that is, trees where decision nodes test the sign of affine functions. This was generalized to algebraic decision trees by Steele and Yao. If the elements in the problem are real numbers or rationals, the decision-tree lower bound extends to the real random-access machine model with an instruction set that includes addition, subtraction and multiplication of real numbers, as well as comparison and either division or remaindering ("floor"). This model covers more algorithms than the algebraic decision-tree model, as it encompasses algorithms that use indexing into tables. However, in this model all program steps are counted, not just decisions. An upper bound for a decision-tree model was given by Meyer auf der Heide who showed that for every n there exists an O(n4)-deep linear decision tree that solves the subset-sum problem with n items. Note that this does not imply any upper bound for an algorithm that should solve the problem for any given n. == Solving == Several algorithms are available to solve knapsack problems, based on the dynamic programming approach, the branch and bound approach or hybridizations of both approaches. === Dynamic programming in-advance algorithm === The unbounded knapsack problem (UKP) places no restriction on the number of copies of each kind of item. Besides, here we assume that x i > 0 {\displaystyle x_{i}>0} m [ w ′ ] = max ( ∑ i = 1 n v i x i ) {\displaystyle m[w']=\max \left(\sum _{i=1}^{n}v_{i}x_{i}\right)} subject to ∑

  Read more →
• 

  Signals intelligence

  Signals intelligence (SIGINT) is the act and field of intelligence-gathering by interception of signals, whether communications between people (communications intelligence—abbreviated to COMINT) or from electronic signals not directly used in communication (electronic intelligence—abbreviated to ELINT). As classified and sensitive information is usually encrypted, signals intelligence may necessarily involve cryptanalysis (to decipher the messages). Traffic analysis—the study of who is signaling to whom and in what quantity—is also used to integrate information, and it may complement cryptanalysis. == History == === Origins === Electronic interceptions appeared as early as 1900, during the Boer War of 1899–1902. The British Royal Navy had installed wireless sets produced by Marconi on board their ships in the late 1890s, and the British Army used some limited wireless signalling. The Boers captured some wireless sets and used them to make vital transmissions. Since the British were the only people transmitting at the time, the British did not need special interpretation of the signals that they were. The birth of signals intelligence in a modern sense dates from the Russo-Japanese War of 1904–1905. As the Russian fleet prepared for conflict with Japan in 1904, the British ship HMS Diana stationed in the Suez Canal intercepted Russian naval wireless signals being sent out for the mobilization of the fleet, for the first time in history. === Development in World War I === Over the course of the First World War, a new method of signals intelligence reached maturity. Russia's failure to properly protect its communications fatally compromised the Russian Army's advance early in World War I and led to their disastrous defeat by the Germans under Ludendorff and Hindenburg at the Battle of Tannenberg. In 1918, French intercept personnel captured a message written in the new ADFGVX cipher, which was cryptanalyzed by Georges Painvin. This gave the Allies advance warning of the German 1918 Spring Offensive. The British in particular, built up great expertise in the newly emerging field of signals intelligence and codebreaking (synonymous with cryptanalysis). On the declaration of war, Britain cut all German undersea cables. This forced the Germans to communicate exclusively via either (A) a telegraph line that connected through the British network and thus could be tapped; or (B) through radio which the British could then intercept. Rear Admiral Henry Oliver appointed Sir Alfred Ewing to establish an interception and decryption service at the Admiralty; Room 40. An interception service known as 'Y' service, together with the post office and Marconi stations, grew rapidly to the point where the British could intercept almost all official German messages. The German fleet was in the habit each day of wirelessing the exact position of each ship and giving regular position reports when at sea. It was possible to build up a precise picture of the normal operation of the High Seas Fleet, to infer from the routes they chose where defensive minefields had been placed and where it was safe for ships to operate. Whenever a change to the normal pattern was seen, it immediately signalled that some operation was about to take place, and a warning could be given. Detailed information about submarine movements was also available. The use of radio-receiving equipment to pinpoint the location of any single transmitter was also developed during the war. Captain H.J. Round, working for Marconi, began carrying out experiments with direction-finding radio equipment for the army in France in 1915. By May 1915, the Admiralty was able to track German submarines crossing the North Sea. Some of these stations also acted as 'Y' stations to collect German messages, but a new section was created within Room 40 to plot the positions of ships from the directional reports. Room 40 played an important role in several naval engagements during the war, notably in detecting major German sorties into the North Sea. The battle of Dogger Bank was won in no small part due to the intercepts that allowed the Navy to position its ships in the right place. It played a vital role in subsequent naval clashes, including at the Battle of Jutland as the British fleet was sent out to intercept them. The direction-finding capability allowed for the tracking and location of German ships, submarines, and Zeppelins. The system was so successful that by the end of the war, over 80 million words, comprising the totality of German wireless transmission over the course of the war, had been intercepted by the operators of the Y-stations and decrypted. However, its most astonishing success was in decrypting the Zimmermann Telegram, a telegram from the German Foreign Office sent via Washington to its ambassador Heinrich von Eckardt in Mexico. === Postwar consolidation === With the importance of interception and decryption firmly established by the wartime experience, countries established permanent agencies dedicated to this task in the interwar period. In 1919, the British Cabinet's Secret Service Committee, chaired by Lord Curzon, recommended that a peace-time codebreaking agency should be created. The Government Code and Cypher School (GC&CS) was the first peace-time codebreaking agency, with a public function "to advise as to the security of codes and cyphers used by all Government departments and to assist in their provision", but also with a secret directive to "study the methods of cypher communications used by foreign powers". GC&CS officially formed on 1 November 1919, and produced its first decrypt on 19 October. By 1940, GC&CS was working on the diplomatic codes and ciphers of 26 countries, tackling over 150 diplomatic cryptosystems. The US Cipher Bureau was established in 1919 and achieved some success at the Washington Naval Conference in 1921, through cryptanalysis by Herbert Yardley. Secretary of War Henry L. Stimson closed the US Cipher Bureau in 1929 with the words "Gentlemen do not read each other's mail." === World War II === The use of SIGINT had even greater implications during World War II. The combined effort of intercepts and cryptanalysis for the whole of the British forces in World War II came under the code name "Ultra", managed from Government Code and Cypher School at Bletchley Park. Properly used, the German Enigma and Lorenz ciphers should have been virtually unbreakable, but flaws in German cryptographic procedures, and poor discipline among the personnel carrying them out, created vulnerabilities which made Bletchley's attacks feasible. Bletchley's work was essential to defeating the U-boats in the Battle of the Atlantic, and to the British naval victories in the Battle of Cape Matapan and the Battle of North Cape. In 1941, Ultra exerted a powerful effect on the North African desert campaign against German forces under General Erwin Rommel. General Sir Claude Auchinleck wrote that were it not for Ultra, "Rommel would have certainly got through to Cairo". Ultra decrypts featured prominently in the story of Operation SALAM, László Almásy's mission across the desert behind Allied lines in 1942. Prior to the Normandy landings on D-Day in June 1944, the Allies knew the locations of all but two of Germany's fifty-eight Western Front divisions. Winston Churchill was reported to have told King George VI: "It is thanks to the secret weapon of General Menzies, put into use on all the fronts, that we won the war!" Supreme Allied Commander, Dwight D. Eisenhower, at the end of the war, described Ultra as having been "decisive" to Allied victory. Official historian of British Intelligence in World War II Sir Harry Hinsley argued that Ultra shortened the war "by not less than two years and probably by four years"; and that, in the absence of Ultra, it is uncertain how the war would have ended. At a lower level, German cryptanalysis, direction finding, and traffic analysis were vital to Rommel's early successes in the Western Desert Campaign until British forces tightened their communications discipline and Australian raiders destroyed his principal SIGINT Company. == Technical definitions == The United States Department of Defense has defined the term "signals intelligence" as: A category of intelligence comprising either individually or in combination all communications intelligence (COMINT), electronic intelligence (ELINT), and foreign instrumentation signals intelligence (FISINT), however transmitted. Intelligence derived from communications, electronic, and foreign instrumentation signals. Being a broad field, SIGINT has many sub-disciplines. The two main ones are communications intelligence (COMINT) and electronic intelligence (ELINT). == Disciplines shared across the branches == === Targeting === A collection system has to know to look for a particular signal. "System", in this context, has several nuances. Targeting is the process of developing collection requirements: "1. A

  Read more →