The neocognitron is a hierarchical, multilayered artificial neural network proposed by Kunihiko Fukushima in 1979. It has been used for Japanese handwritten character recognition and other pattern recognition tasks, and served as the inspiration for convolutional neural networks. Previously in 1969, he published a similar architecture, but with hand-designed kernels inspired by convolutions in mammalian vision. In 1975 he improved it to the Cognitron, and in 1979 he improved it to the neocognitron, which learns all convolutional kernels by unsupervised learning (in his terminology, "self-organized by 'learning without a teacher'"). The neocognitron was inspired by the model proposed by Hubel & Wiesel in 1959. They found two types of cells in the visual primary cortex called simple cell and complex cell, and also proposed a cascading model of these two types of cells for use in pattern recognition tasks. The neocognitron is a natural extension of these cascading models. The neocognitron consists of multiple types of cells, the most important of which are called S-cells and C-cells. The local features are extracted by S-cells, and these features' deformation, such as local shifts, are tolerated by C-cells. Local features in the input are integrated gradually and classified in the higher layers. The idea of local feature integration is found in several other models, such as the Convolutional Neural Network model, the SIFT method, and the HoG method. There are various kinds of neocognitron. For example, some types of neocognitron can detect multiple patterns in the same input by using backward signals to achieve selective attention.
Valantic
Valantic GmbH (stylised as valantic) is an IT service and consulting company headquartered in Munich, Germany. == History == Valantic GmbH was founded in 2012 under the name Dabero Service Group. Until it was renamed Valantic GmbH in 2017, the company merged with IT service providers and consulting firms. These included, among others, Realtime AG, a company for SAP systems. The companies involved in these mergers were also renamed in 2017 and have since used the Valantic brand name. Realtime AG, for example, became Valantic ERP Services AG. During the COVID-19 pandemic and the resulting economic pressures, demand increased for IT service providers, particularly those offering customised software, IT consulting, SAP services, customer experience, cybersecurity, IoT, and digital work environments. In the following years, Valantic expanded by integrating additional companies. In 2021, Valantic expanded into other European countries through the integration of the Dutch company ISM eCompany and the Portuguese consulting firm Abaco. In 2022, the consulting firm C-Clear/Atom Ideas from Belgium joined Valantic. In February 2019, DPE Deutsche Private Equity Management III GmbH (DPE) took over the majority shareholding in Valantic. The founder, Holger von Daniels, and the further management retained a 25% stake. By 2025, DPE had invested €500 million in Valantic. In the following years, Valantic expanded its international locations. In 2023, Valantic incorporated the Danish company Inspari into the group, thereby entering the Scandinavian market. Inspari is a company for Microsoft technologies such as Azure and Power Platform. In the same year, Valantic joined forces with the Aiopsgroup, an international provider of online shopping applications for private and business customers of large companies. The company is based in Bulgaria with additional locations across Eastern Europe and other places. Additionally, the SAP applications division was expanded through the merger with the Spanish company Saptools. As a result, the companies became one of the largest European end-to-end consulting and implementation house for SAP services. By the end of 2023, Valantic had locations in 18 countries. In November 2024, Valantic announced its merger with the Danish digital consultancy Venzo. Through the integration of the company, founded in 2007 and oriented towards Microsoft technologies and digital transformation projects in the areas of automation, artificial intelligence, security, infrastructure and change management, Valantic further expanded its presence in Denmark and the Nordic countries. In July 2025, Valantic announced its merger with Utiligence GmbH, a Mannheim-based consulting firm for SAP technologies. Utiligence works primarily for the energy industry and supports companies in the integration of SAP S/4HANA and the digitalisation of business processes. == Company structure == Valantic is a partnership-based organisation, with partners acting as decision-makers in matters relating to corporate strategy, employee development and acquisitions. Valantic pursues a holacratic approach, promoting an open and self-organised way of working instead of hierarchical structures. By merging with other companies, Valantic is expanding its range of services and tapping into international markets and market shares. The new companies use Valantic's core systems and support processes, but usually retain their original structure. In the 2024 financial year, the company generated revenue of €544 million and employed 3,874 on average. Valantic has over 40 locations internationally. == Services == Valantic GmbH is a consulting firm, software provider and implementation partner. The company offers services in the areas of digital strategy and analytics (business intelligence and data science), customer experience management, SAP services, smart industries (Industry 4.0, supply chain management, and production planning and control processes), and financial services automation. The automation of financial services is aimed at financial service providers and banks. Valantic has been offering services in the field of generative artificial intelligence (GenAI) since 2023. Part of these services involves enabling companies to use GenAI securely and in compliance with regulations in order to make internal work processes more efficient. Its customers include large corporations, several medium-sized companies and DAX-listed companies. == Research == Since 2018, Valantic has published an annual study on the development of the SAP landscape in German-speaking countries. The study examines topics such as the migration to SAP S/4HANA, cloud strategies, technological trends and the use of artificial intelligence in business processes. The 2025 survey of 201 SAP professionals from the DACH region showed, for example, an increase in ongoing and completed S/4HANA migration projects, as well as a further shift towards private-cloud systems. The use of artificial intelligence continued to grow, as did the use of the SAP Business Technology Platform and the Business Data Cloud. In 2025, Valantic, together with the Handelsblatt Research Institute, published the trend study Digital 2030 – The Rise of Applied AI. The study was based on a survey of around 700 executives from companies in Germany, Austria, and Switzerland on the economic effects of current digitalisation trends. According to the study, most respondents consider artificial intelligence, cybersecurity, and cloud computing to hold the greatest strategic importance for business success by 2030. Around 70% of the participating companies stated that they are already achieving measurable business benefits through the use of AI applications, for example in quality control, document management, logistics, or customer service.
JustWatch
JustWatch is a website that provides information on the availability of films and TV shows on various streaming platforms such as Netflix, HBO Max, Disney+, Hulu, Peacock, Fandango at Home, Apple TV, and Amazon Prime Video, among others. It is also available as a mobile application and smart TV application. JustWatch provides a search engine that allows users to discover which digital platforms host a particular movie or TV series. As of November 2023, JustWatch is available to users in 139 countries. == Features == JustWatch functions as a search engine by aggregating information about the online availability of films and TV series from video-on-demand streaming services. It aggregates information from more than 100 video content libraries, as well providing information about video resolution quality, pricing, and purchase or rental options. The website includes various filters for searching, including genre, price, release date, rating, and popularity. Users are also able to create lists of shows and movies and to share these lists with other users. == History == JustWatch GmbH is an international database company that is privately held and headquartered in Berlin, Germany. The company specializes in the online availability of movies and TV series. In addition to its user-facing website, the company also has an advertising-focused arm, JustWatch Media, that works with corporate clients, using data about what people watch that it gleans from user behavior to help entertainment companies tailor their marketing strategies. Its clients include Universal Pictures, Paramount Pictures, and Sony Pictures, among others. Development of the website began in 2014, and it was launched in the U.S. and Germany in February 2015. In 2018, the company received funding to improve databases within the European Union. In December 2019, the company acquired a rival streaming aggregation service, GoWatchIt, from Plexus Entertainment. JustWatch also used the acquisition to open its first New York office. In 2019, JustWatch had over 30 million users across 38 countries. By 2020, the company's streaming aggregation service was available in over 45 countries. By November 2023, it was available in 139 countries, and had over 40 million monthly users. === Founding === JustWatch was co-founded in 2013 by David Croyé, Cristoph Hoyer, Kevin Hiller, Dominik Raute, Ingke Weimert, and Michael Wilken. In a company blog post from February 2017, Croyé described the group of co-founders as all having previously "worked in leading roles at successful international tech-startups in Berlin." Croyé, who currently holds the title of CEO at JustWatch GmbH, had previously worked as the chief marketing officer at kaufDA, a European location-based mobile coupon and promotion service, and the background of other co-founders included time at the adtech company Trademob and the streaming site MyVideo. Startup capital for the website initially came from the founders themselves. Croyé in particular was able to reinvest funds he had obtained from the sale of kaufDA to Axel Springer, a European media company, in March 2011. Since 2015, the company has had at least one additional round of seed funding, with investors including venture capital groups CG Partners and STS Ventures.
NNDB
The Notable Names Database (NNDB) is an online database of biographical details of over 40,000 people. Soylent Communications, a sole proprietorship that also hosted the later defunct Rotten.com, describes NNDB as an "intelligence aggregator" of noteworthy persons, highlighting their interpersonal connections. The Rotten.com domain was registered in 1996 by former Apple and Netscape software engineer Thomas E. Dell, who was also known by his internet alias, "Soylent". == Entries == Each entry has an executive summary followed by a brief narrative about their life. It also lists date and cause of death if deceased. Businesspeople and government officials are listed with chronologies of their posts, positions, and board memberships. As of 2022, the site is no longer updated. == NNDB Mapper == The NNDB Mapper, a visual tool for exploring connections between people, was made available in May 2008. It required Adobe Flash 7.
Hierarchical RBF
In computer graphics, hierarchical RBF is an interpolation method based on radial basis functions (RBFs). Hierarchical RBF interpolation has applications in treatment of results from a 3D scanner, terrain reconstruction, and the construction of shape models in 3D computer graphics (such as the Stanford bunny, a popular 3D model). This problem is informally named as "large scattered data point set interpolation." == Method == The steps of the interpolation method (in three dimensions) are as follows: Let the scattered points be presented as set P = { c i = ( x i , y i , z i ) | i = 1 N ⊂ R 3 } {\displaystyle \mathbf {P} =\{\mathbf {c} _{i}=(\mathbf {x} _{i},\mathbf {y} _{i},\mathbf {z} _{i})\vert _{i=1}^{N}\subset \mathbb {R} ^{3}\}} Let there exist a set of values of some function in scattered points H = { h i | i = 1 N ⊂ R } {\displaystyle \mathbf {H} =\{\mathbf {h} _{i}\vert _{i=1}^{N}\subset \mathbb {R} \}} Find a function f ( x ) {\displaystyle \mathbf {f} (\mathbf {x} )} that will meet the condition f ( x ) = 1 {\displaystyle \mathbf {f} (\mathbf {x} )=1} for points lying on the shape and f ( x ) ≠ 1 {\displaystyle \mathbf {f} (\mathbf {x} )\neq 1} for points not lying on the shape As J. C. Carr et al. showed, this function takes the form f ( x ) = ∑ i = 1 N λ i φ ( x , c i ) {\displaystyle \mathbf {f} (\mathbf {x} )=\sum _{i=1}^{N}\lambda _{i}\varphi (\mathbf {x} ,\mathbf {c} _{i})} where φ {\displaystyle \varphi } is a radial basis function and λ {\displaystyle \lambda } are the coefficients that are the solution of the following linear system of equations: [ φ ( c 1 , c 1 ) φ ( c 1 , c 2 ) . . . φ ( c 1 , c N ) φ ( c 2 , c 1 ) φ ( c 2 , c 2 ) . . . φ ( c 2 , c N ) . . . . . . . . . . . . φ ( c N , c 1 ) φ ( c N , c 2 ) . . . φ ( c N , c N ) ] ∗ [ λ 1 λ 2 . . . λ N ] = [ h 1 h 2 . . . h N ] {\displaystyle {\begin{bmatrix}\varphi (c_{1},c_{1})&\varphi (c_{1},c_{2})&...&\varphi (c_{1},c_{N})\\\varphi (c_{2},c_{1})&\varphi (c_{2},c_{2})&...&\varphi (c_{2},c_{N})\\...&...&...&...\\\varphi (c_{N},c_{1})&\varphi (c_{N},c_{2})&...&\varphi (c_{N},c_{N})\end{bmatrix}}{\begin{bmatrix}\lambda _{1}\\\lambda _{2}\\...\\\lambda _{N}\end{bmatrix}}={\begin{bmatrix}h_{1}\\h_{2}\\...\\h_{N}\end{bmatrix}}} For determination of surface, it is necessary to estimate the value of function f ( x ) {\displaystyle \mathbf {f} (\mathbf {x} )} in specific points x. A lack of such method is a considerable complication on the order of O ( n 2 ) {\displaystyle \mathbf {O} (\mathbf {n} ^{2})} to calculate RBF, solve system, and determine surface. == Other methods == Reduce interpolation centers ( O ( n 2 ) {\displaystyle \mathbf {O} (\mathbf {n} ^{2})} to calculate RBF and solve system, O ( m n ) {\displaystyle \mathbf {O} (\mathbf {m} \mathbf {n} )} to determine surface) Compactly support RBF ( O ( n log n ) {\displaystyle \mathbf {O} (\mathbf {n} \log {\mathbf {n} })} to calculate RBF, O ( n 1.2..1.5 ) {\displaystyle \mathbf {O} (\mathbf {n} ^{1.2..1.5})} to solve system, O ( m log n ) {\displaystyle \mathbf {O} (\mathbf {m} \log {\mathbf {n} })} to determine surface) FMM ( O ( n 2 ) {\displaystyle \mathbf {O} (\mathbf {n} ^{2})} to calculate RBF, O ( n log n ) {\displaystyle \mathbf {O} (\mathbf {n} \log {\mathbf {n} })} to solve system, O ( m + n log n ) {\displaystyle \mathbf {O} (\mathbf {m} +\mathbf {n} \log {\mathbf {n} })} to determine surface) == Hierarchical algorithm == A hierarchical algorithm allows for an acceleration of calculations due to decomposition of intricate problems on the great number of simple (see picture). In this case, hierarchical division of space contains points on elementary parts, and the system of small dimension solves for each. The calculation of surface in this case is taken to the hierarchical (on the basis of tree-structure) calculation of interpolant. A method for a 2D case is offered by Pouderoux J. et al. For a 3D case, a method is used in the tasks of 3D graphics by W. Qiang et al. and modified by Babkov V.
PagedAttention
PagedAttention is an attention algorithm for efficient serving of large language models (LLMs). It was introduced in 2023 by Woosuk Kwon and colleagues in the paper Efficient Memory Management for Large Language Model Serving with PagedAttention, alongside the vLLM serving engine. The method stores the key–value cache used during autoregressive decoding in fixed-size blocks that can be mapped to non-contiguous physical memory, borrowing ideas from virtual memory, paging, and operating system design. == Background == In transformer inference, the key–value cache grows with sequence length and the number of concurrent requests. Kwon et al. argued that earlier serving systems typically reserved contiguous cache regions in advance, which caused reserved space, internal fragmentation, and external fragmentation. In their experiments, the paper reported that the effective memory utilization of previous systems could fall as low as 20.4%. == Description == PagedAttention partitions the cache of each sequence into fixed-size KV blocks. A request's cache is represented as a sequence of logical blocks, while a block table maps those logical blocks to physical GPU-memory blocks. As a result, neighboring logical blocks do not need to be contiguous in physical memory, and new blocks can be allocated on demand as generation proceeds. The design also makes it easier to share cache state across related decoding paths. In vLLM, physical blocks can be reference-counted and shared among requests or branches, with block-granularity copy-on-write used when a shared block must be modified. The original paper applied this design to parallel sampling, beam search, and prompts with shared prefixes. == Mathematical formulation == For a query token i {\displaystyle i} in causal self-attention, the standard attention output can be written as a i j = exp ( q i ⊤ k j / d ) ∑ t = 1 i exp ( q i ⊤ k t / d ) , o i = ∑ j = 1 i a i j v j {\displaystyle a_{ij}={\frac {\exp(\mathbf {q} _{i}^{\top }\mathbf {k} _{j}/{\sqrt {d}})}{\sum _{t=1}^{i}\exp(\mathbf {q} _{i}^{\top }\mathbf {k} _{t}/{\sqrt {d}})}},\;\mathbf {o} _{i}=\sum _{j=1}^{i}a_{ij}\mathbf {v} _{j}} where q i {\displaystyle \mathbf {q} _{i}} , k j {\displaystyle \mathbf {k} _{j}} , and v j {\displaystyle \mathbf {v} _{j}} are the query, key, and value vectors, and d {\displaystyle d} is the attention dimension. If the cache is partitioned into blocks of size B {\displaystyle B} , the key and value blocks may be written as K j = ( k ( j − 1 ) B + 1 , … , k j B ) , V j = ( v ( j − 1 ) B + 1 , … , v j B ) {\displaystyle \mathbf {K} _{j}=(\mathbf {k} _{(j-1)B+1},\ldots ,\mathbf {k} _{jB}),\;\mathbf {V} _{j}=(\mathbf {v} _{(j-1)B+1},\ldots ,\mathbf {v} _{jB})} PagedAttention then performs the computation blockwise: A i j = exp ( q i ⊤ K j / d ) ∑ t = 1 ⌈ i / B ⌉ exp ( q i ⊤ K t / d ) , o i = ∑ j = 1 ⌈ i / B ⌉ V j A i j ⊤ {\displaystyle \mathbf {A} _{ij}={\frac {\exp(\mathbf {q} _{i}^{\top }\mathbf {K} _{j}/{\sqrt {d}})}{\sum _{t=1}^{\lceil i/B\rceil }\exp(\mathbf {q} _{i}^{\top }\mathbf {K} _{t}/{\sqrt {d}})}},\;\mathbf {o} _{i}=\sum _{j=1}^{\lceil i/B\rceil }\mathbf {V} _{j}\mathbf {A} _{ij}^{\top }} where A i j {\displaystyle \mathbf {A} _{ij}} is the vector of attention scores for the j {\displaystyle j} -th KV block. In the formulation given by Kwon et al., this preserves the causal attention calculation while allowing the key and value blocks to reside in non-contiguous physical memory. == Performance and use == The vLLM paper reported that, on its evaluated workloads, the use of PagedAttention and the associated memory-management design improved serving throughput by 2–4× over the compared baselines, including FasterTransformer and Orca, while preserving model outputs. In experiments on OPT-13B with the Alpaca trace, the paper also reported memory savings of 6.1–9.8% for parallel sampling and 37.6–55.2% for beam search through KV-block sharing. A 2024 survey of LLM serving systems described PagedAttention as having become an industry norm in LLM serving frameworks, citing support in TGI, vLLM, and TensorRT-LLM. == Limitations and alternatives == Subsequent work has described trade-offs in the approach. The 2025 vAttention paper argued that PagedAttention requires attention kernels to be rewritten to support paging and increases software complexity, portability issues, redundancy, and execution overhead, proposing instead a memory manager that keeps the cache contiguous in virtual memory while relying on demand paging for physical allocation. === vAttention === Unlike PagedAttention, vAttention does not introduce a different attention rule; it retains the standard attention computation Attention ( q i , K , V ) = softmax ( q i K ⊤ s c a l e ) V . {\displaystyle \operatorname {Attention} (q_{i},K,V)=\operatorname {softmax} \left({\frac {q_{i}K^{\top }}{\mathrm {scale} }}\right)V.} In the notation of Prabhu et al., the key and value tensors for a request seen so far are K , V ∈ R L ′ × ( H × D ) {\displaystyle K,V\in \mathbb {R} ^{L'\times (H\times D)}} , where L ′ {\displaystyle L'} is the context length seen so far, H {\displaystyle H} is the number of KV heads on a worker, and D {\displaystyle D} is the dimension of each KV head. In systems prior to PagedAttention, the K cache (or V cache) at each layer of a worker is typically allocated as a 4D tensor of shape [ B , L , H , D ] , {\displaystyle [B,L,H,D],} where B {\displaystyle B} is batch size and L {\displaystyle L} is the maximum context length supported by the model. vAttention preserves this contiguous virtual-memory view while deferring physical-memory allocation to runtime. A serving framework maintains separate K and V tensors for each layer, so vAttention reserves 2 N {\displaystyle 2N} virtual-memory buffers on a worker, where N {\displaystyle N} is the number of layers managed by that worker. The maximum size of one virtual-memory buffer is B S = B × S , {\displaystyle BS=B\times S,} where S {\displaystyle S} is the maximum size of a single request's per-layer K cache (or V cache) on a worker. The paper defines S = L × H × D × P , {\displaystyle S=L\times H\times D\times P,} where P {\displaystyle P} is the number of bytes needed to store one element. In this formulation, vAttention keeps the KV cache contiguous in virtual memory and relies on demand paging for physical allocation, rather than modifying the attention kernel to operate over non-contiguous KV-cache blocks.
Visible (mobile app)
Visible is a health tracking mobile app for people with long COVID and myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). The company was founded by a Harry Leeming, an engineer from London living with long Covid since 2020, and Luke Martin-Fuller. In November 2022, Visible released an open beta of an app that aims to help people pace their activities to avoid post-exertional malaise. The app gathers data on exertion levels, symptom severity, and heart-rate variability. HRV is approximated using a smartphone's camera via a technique called photoplethysmography, and according to the app's developers, can indicate how much someone needs rest. The app is currently free, but is expected to be freemium in the future. Users can also opt to allow their data be used for research purposes. In July 2023, Visible and Imperial College London announced the start of the first two studies. One is on the effects of the menstrual cycle on long COVID symptoms, and the other is on the condition's epidemiology and economic impact. Visible has announced plans to couple the app with activity trackers for continuous monitoring of heart-rate and actimetry data, which the developers claim will be more effective. As of 2022, no clinical trials on Visible's effectiveness have been conducted.