In computer science, a trie (, ), also known as a digital tree or prefix tree, is a specialized search tree data structure used to store and retrieve strings from a dictionary or set. Unlike a binary search tree, nodes in a trie do not store their associated key. Instead, each node's position within the trie determines its associated key, with the connections between nodes defined by individual characters rather than the entire key. Tries are particularly effective for tasks such as autocomplete, spell checking, and IP routing, offering advantages over hash tables due to their prefix-based organization and lack of hash collisions. Every child node shares a common prefix with its parent node, and the root node represents the empty string. While basic trie implementations can be memory-intensive, various optimization techniques such as compression and bitwise representations have been developed to improve their efficiency. A notable optimization is the radix tree, which provides more efficient prefix-based storage. While tries store character strings, they can be adapted to work with any ordered sequence of elements, such as permutations of digits or shapes. A notable variant is the bitwise trie, which uses individual bits from fixed-length binary data (such as integers or memory addresses) as keys. == History, etymology, and pronunciation == The idea of a trie for representing a set of strings was first abstractly described by Axel Thue in 1912. Tries were first described in a computer context by René de la Briandais in 1959. The idea was independently described in 1960 by Edward Fredkin, who coined the term trie, pronouncing it (as "tree"), after the middle syllable of retrieval. However, other authors pronounce it (as "try"), in an attempt to distinguish it verbally from "tree". == Overview == Tries are a form of string-indexed look-up data structure, which is used to store a dictionary list of words that can be searched on in a manner that allows for efficient generation of completion lists. A prefix trie is an ordered tree data structure used in the representation of a set of strings over a finite alphabet set, which allows efficient storage of words with common prefixes. Tries can be efficacious on string-searching algorithms such as predictive text, approximate string matching, and spell checking in comparison to binary search trees. A trie can be seen as a tree-shaped deterministic finite automaton. == Operations == Tries support various operations: insertion, deletion, and lookup of a string key. Tries are composed of nodes that contain links, which either point to other suffix child nodes or null. As for every tree, each node except the root is pointed to by only one other node, called its parent. Each node contains as many links as the number of characters in the applicable alphabet (although tries tend to have a substantial number of null links). In some cases, the alphabet used is simply that of the character encoding—resulting in, for example, a size of 128 in the case of ASCII. The null links within the children of a node emphasize the following characteristics: Characters and string keys are implicitly stored in the trie, and include a character sentinel value indicating string termination. Each node contains one possible link to a prefix of strong keys of the set. A basic structure type of nodes in the trie is as follows: Node {\displaystyle {\text{Node}}} may contain an optional Value {\displaystyle {\text{Value}}} , which is associated with the key that corresponds to the node. === Searching === Searching for a value in a trie is guided by the characters in the search string key, as each node in the trie contains a corresponding link to each possible character in the given string. Thus, following the string within the trie yields the associated value for the given string key. A null link during the search indicates the inexistence of the key. The following pseudocode implements the search procedure for a given string key in a rooted trie x. In the above pseudocode, x and key correspond to the pointer of the trie's root node and the string key, respectively. The search operation takes O ( m ) {\displaystyle O(m)} time, where m {\displaystyle m} is the size of the string parameter key. In a balanced binary search tree, on the other hand, it takes O ( m log n ) {\displaystyle O(m\log n)} time, in the worst case, since key needs to be compared with O ( log n ) {\displaystyle O(\log n)} other keys and each comparison takes O ( m ) {\displaystyle O(m)} time, in the worst case. The trie occupies less space, in comparison with a binary search tree, in the case of a large number of short strings, since nodes share common initial string subsequences and store the keys implicitly. === Insertion === Insertion into a trie is guided by using the character sets as indexes to the children array until the last character of the string key is reached. Each node in the trie corresponds to one call of the radix sorting routine, as the trie structure reflects the execution pattern of the top-down radix sort. If null links are encountered before reaching the last character of the string key, new nodes are created. The input value is assigned to the value of the last node traversed, which is the node that corresponds to the key. === Deletion === Deletion of a key–value pair from a trie involves finding the node corresponding to the key, setting its value to null, and recursively removing nodes that have no children. The procedure begins by examining key; an empty string indicates arrival at the node corresponding to the (original) key, in which case its value is set to null. If the node, then, has null value and no children, it is removed from the trie by returning null; otherwise, the node is kept by returning the node itself. == Replacing other data structures == === Replacement for hash tables === A trie can be used to replace a hash table, over which it has the following advantages: Searching for a node with an associated key of size m {\displaystyle m} has the complexity of O ( m ) {\displaystyle O(m)} , whereas an imperfect hash function may have numerous colliding keys, and the worst-case lookup speed of such a table would be O ( N ) {\displaystyle O(N)} , where N {\displaystyle N} denotes the total number of nodes within the table. Tries do not need a hash function for the operation, unlike a hash table; there are also no collisions of different keys in a trie. Within a trie, keys can be efficiently sorted lexicographically. However, tries are less efficient than a hash table when the data is directly accessed on a secondary storage device such as a hard disk drive that has higher random access time than the main memory. == Implementation strategies == Tries can be represented in several ways, corresponding to different trade-offs between memory use and speed of the operations. Using a vector of pointers for representing a trie consumes enormous space; however, memory space can be reduced at the expense of running time if a singly linked list is used for each node vector, as most entries of the vector contains nil {\displaystyle {\text{nil}}} . Techniques such as alphabet reduction may reduce the large space requirements by reinterpreting the original string as a longer string over a smaller alphabet. For example, a string of n bytes can alternatively be regarded as a string of 2n four-bit units. This can reduce memory usage by a factor of eight; but lookups need to visit twice as many nodes in the worst case. Another technique includes storing a vector of 256 ASCII pointers as a bitmap of 256 bits representing ASCII alphabet, which reduces the size of individual nodes dramatically. === Bitwise tries === Bitwise tries are used to address the enormous space requirement for the trie nodes in a naive simple pointer vector implementations. Each character in the string key set is represented via individual bits, which are used to traverse the trie over a string key. The implementations for these types of trie use vectorized CPU instructions to find the first set bit in a fixed-length key input (e.g. GCC's __builtin_clz() intrinsic function). Accordingly, the set bit is used to index the first item, or child node, in the 32- or 64-entry based bitwise tree. Search then proceeds by testing each subsequent bit in the key. This procedure is also cache-local and highly parallelizable due to register independency, and thus performant on out-of-order execution CPUs. === Compressed tries === Radix tree, also known as a compressed trie, is a space-optimized variant of a trie in which any node with only one child gets merged with its parent; elimination of branches of the nodes with a single child results in better metrics in both space and time. This works best when the trie remains static and set of keys stored are very sparse within their representation space. One more approach for static tries is to "pack" the trie by storing disjoint
Shape context
Shape context is a feature descriptor used in object recognition. Serge Belongie and Jitendra Malik proposed the term in their paper "Matching with Shape Contexts" in 2000. == Theory == The shape context is intended to be a way of describing shapes that allows for measuring shape similarity and the recovering of point correspondences. The basic idea is to pick n points on the contours of a shape. For each point pi on the shape, consider the n − 1 vectors obtained by connecting pi to all other points. The set of all these vectors is a rich description of the shape localized at that point but is far too detailed. The key idea is that the distribution over relative positions is a robust, compact, and highly discriminative descriptor. So, for the point pi, the coarse histogram of the relative coordinates of the remaining n − 1 points, h i ( k ) = # { q ≠ p i : ( q − p i ) ∈ bin ( k ) } {\displaystyle h_{i}(k)=\#\{q\neq p_{i}:(q-p_{i})\in {\mbox{bin}}(k)\}} is defined to be the shape context of p i {\displaystyle p_{i}} . The bins are normally taken to be uniform in log-polar space. The fact that the shape context is a rich and discriminative descriptor can be seen in the figure below, in which the shape contexts of two different versions of the letter "A" are shown. (a) and (b) are the sampled edge points of the two shapes. (c) is the diagram of the log-polar bins used to compute the shape context. (d) is the shape context for the point marked with a circle in (a), (e) is that for the point marked as a diamond in (b), and (f) is that for the triangle. As can be seen, since (d) and (e) are the shape contexts for two closely related points, they are quite similar, while the shape context in (f) is very different. For a feature descriptor to be useful, it needs to have certain invariances. In particular it needs to be invariant to translation, scaling, small perturbations, and, depending on the application, rotation. Translational invariance comes naturally to shape context. Scale invariance is obtained by normalizing all radial distances by the mean distance α {\displaystyle \alpha } between all the point pairs in the shape although the median distance can also be used. Shape contexts are empirically demonstrated to be robust to deformations, noise, and outliers using synthetic point set matching experiments. One can provide complete rotational invariance in shape contexts. One way is to measure angles at each point relative to the direction of the tangent at that point (since the points are chosen on edges). This results in a completely rotationally invariant descriptor. But of course this is not always desired since some local features lose their discriminative power if not measured relative to the same frame. Many applications in fact forbid rotational invariance e.g. distinguishing a "6" from a "9". == Use in shape matching == A complete system that uses shape contexts for shape matching consists of the following steps (which will be covered in more detail in the Details of Implementation section): Randomly select a set of points that lie on the edges of a known shape and another set of points on an unknown shape. Compute the shape context of each point found in step 1. Match each point from the known shape to a point on an unknown shape. To minimize the cost of matching, first choose a transformation (e.g. affine, thin plate spline, etc.) that warps the edges of the known shape to the unknown (essentially aligning the two shapes). Then select the point on the unknown shape that most closely corresponds to each warped point on the known shape. Calculate the "shape distance" between each pair of points on the two shapes. Use a weighted sum of the shape context distance, the image appearance distance, and the bending energy (a measure of how much transformation is required to bring the two shapes into alignment). To identify the unknown shape, use a nearest-neighbor classifier to compare its shape distance to shape distances of known objects. == Details of implementation == === Step 1: Finding a list of points on shape edges === The approach assumes that the shape of an object is essentially captured by a finite subset of the points on the internal or external contours on the object. These can be simply obtained using the Canny edge detector and picking a random set of points from the edges. Note that these points need not and in general do not correspond to key-points such as maxima of curvature or inflection points. It is preferable to sample the shape with roughly uniform spacing, though it is not critical. === Step 2: Computing the shape context === This step is described in detail in the Theory section. === Step 3: Computing the cost matrix === Consider two points p and q that have normalized K-bin histograms (i.e. shape contexts) g(k) and h(k). As shape contexts are distributions represented as histograms, it is natural to use the χ2 test statistic as the "shape context cost" of matching the two points: C S = 1 2 ∑ k = 1 K [ g ( k ) − h ( k ) ] 2 g ( k ) + h ( k ) {\displaystyle C_{S}={\frac {1}{2}}\sum _{k=1}^{K}{\frac {[g(k)-h(k)]^{2}}{g(k)+h(k)}}} The values of this range from 0 to 1. In addition to the shape context cost, an extra cost based on the appearance can be added. For instance, it could be a measure of tangent angle dissimilarity (particularly useful in digit recognition): C A = 1 2 ‖ ( cos ( θ 1 ) sin ( θ 1 ) ) − ( cos ( θ 2 ) sin ( θ 2 ) ) ‖ {\displaystyle C_{A}={\frac {1}{2}}{\begin{Vmatrix}{\dbinom {\cos(\theta _{1})}{\sin(\theta _{1})}}-{\dbinom {\cos(\theta _{2})}{\sin(\theta _{2})}}\end{Vmatrix}}} This is half the length of the chord in unit circle between the unit vectors with angles θ 1 {\displaystyle \theta _{1}} and θ 2 {\displaystyle \theta _{2}} . Its values also range from 0 to 1. Now the total cost of matching the two points could be a weighted-sum of the two costs: C = ( 1 − β ) C S + β C A {\displaystyle C=(1-\beta )C_{S}+\beta C_{A}\!\,} Now for each point pi on the first shape and a point qj on the second shape, calculate the cost as described and call it Ci,j. This is the cost matrix. === Step 4: Finding the matching that minimizes total cost === Now, a one-to-one matching π ( i ) {\displaystyle \pi (i)} that matches each point pi on shape 1 and qj on shape 2 that minimizes the total cost of matching, H ( π ) = ∑ i C ( p i , q π ( i ) ) {\displaystyle H(\pi )=\sum _{i}C\left(p_{i},q_{\pi (i)}\right)} is needed. This can be done in O ( N 3 ) {\displaystyle O(N^{3})} time using the Hungarian method, although there are more efficient algorithms. To have robust handling of outliers, one can add "dummy" nodes that have a constant but reasonably large cost of matching to the cost matrix. This would cause the matching algorithm to match outliers to a "dummy" if there is no real match. === Step 5: Modeling transformation === Given the set of correspondences between a finite set of points on the two shapes, a transformation T : R 2 → R 2 {\displaystyle T:\mathbb {R} ^{2}\to \mathbb {R} ^{2}} can be estimated to map any point from one shape to the other. There are several choices for this transformation, described below. ==== Affine ==== The affine model is a standard choice: T ( p ) = A p + o {\displaystyle T(p)=Ap+o\!} . The least squares solution for the matrix A {\displaystyle A} and the translational offset vector o is obtained by: o = 1 n ∑ i = 1 n ( p i − q π ( i ) ) , A = ( Q + P ) t {\displaystyle o={\frac {1}{n}}\sum _{i=1}^{n}\left(p_{i}-q_{\pi (i)}\right),A=(Q^{+}P)^{t}} Where P = ( 1 p 11 p 12 ⋮ ⋮ ⋮ 1 p n 1 p n 2 ) {\displaystyle P={\begin{pmatrix}1&p_{11}&p_{12}\\\vdots &\vdots &\vdots \\1&p_{n1}&p_{n2}\end{pmatrix}}} with a similar expression for Q {\displaystyle Q\!} . Q + {\displaystyle Q^{+}\!} is the pseudoinverse of Q {\displaystyle Q\!} . ==== Thin plate spline ==== The thin plate spline (TPS) model is the most widely used model for transformations when working with shape contexts. A 2D transformation can be separated into two TPS function to model a coordinate transform: T ( x , y ) = ( f x ( x , y ) , f y ( x , y ) ) {\displaystyle T(x,y)=\left(f_{x}(x,y),f_{y}(x,y)\right)} where each of the ƒx and ƒy have the form: f ( x , y ) = a 1 + a x x + a y y + ∑ i = 1 n ω i U ( ‖ ( x i , y i ) − ( x , y ) ‖ ) , {\displaystyle f(x,y)=a_{1}+a_{x}x+a_{y}y+\sum _{i=1}^{n}\omega _{i}U\left({\begin{Vmatrix}(x_{i},y_{i})-(x,y)\end{Vmatrix}}\right),} and the kernel function U ( r ) {\displaystyle U(r)\!} is defined by U ( r ) = r 2 log r 2 {\displaystyle U(r)=r^{2}\log r^{2}\!} . The exact details of how to solve for the parameters can be found elsewhere but it essentially involves solving a linear system of equations. The bending energy (a measure of how much transformation is needed to align the points) will also be easily obtained. ==== Regularized TPS ==== The TPS formulation above has exact matching requirement for the pairs of points on the two shapes. For noisy data, it is best to
Is an AI Website Builder Worth It in 2026?
Comparing the best AI website builder? An AI website builder is software that uses machine learning to help you get more done — it lowers the barrier so anyone can produce professional output. Privacy matters too: check whether your data trains the model and whether a no-log or enterprise tier is available. Whether you are a beginner or a pro, the right AI website builder slots into your workflow and pays for itself fast. We tested the leading options and ranked them by quality, value, and ease of use.
MemoQ
memoQ is a computer-assisted translation software suite which runs on Microsoft Windows operating systems. It is developed by the Hungarian software company memoQ Fordítástechnológiai Zrt. (memoQ Translation Technologies), formerly Kilgray, a provider of translation management software established in 2004 and cited as one of the fastest-growing companies in the translation technology sector in 2012, and 2013. memoQ provides translation memory, terminology, machine translation integration and reference information management in desktop, client/server and web application environments. == History == memoQ, a translation environment tool first released in 2006, was the first product created by memoQ Translation Technologies, a company founded in Hungary by the three language technologists Balázs Kis, István Lengyel and Gábor Ugray. In the years since the software was first presented, it has grown in popularity and is now among the most frequent TEnT applications used for translation (it was rated as the third most used CAT tool in a Proz.com study in 2013 and as the second most widely used tool in a June 2010 survey of 458 working translators), after SDL Trados, Wordfast, Déjà Vu, OmegaT and others. Today it is available in desktop versions for translators (Translator Pro edition), and project managers (Project Manager edition), as well as site-installed and hosted server applications offering integration with the desktop versions and a web browser interface. There are currently several active online forums in which users provide each other with independent advice and support on the software's functions, as well as many online tutorials created by professional trainers and active users. Before its commercial debut, a version of memoQ (2.0) was distributed as postcardware.
Joseph Keshet
Joseph (Yossi) Keshet (Hebrew: יוסף (יוסי) קשת; born: 28 February 1973) is an Israeli professor in the Electrical and Computer Engineering Faculty of the Technion, where he is the director of the Speech, Language, and Deep Learning Lab. His research focuses on human speech processing and machine learning. == Early life and education == Keshet was born in Tel-Aviv. He graduated from the Amal School and began his academic studies at the Department of Electrical Engineering-Systems at Tel-Aviv University in 1991 and received his B.Sc. (Cum Laude) in 1994. Keshet served in the IDF Unit 8200 from 1995 to 2002 as the head of the speech processing research section in the R&D Center. During his service, he received a national award from the Administration for the Development of Weapons and Technological Infrastructure (Maf’at). Keshet was award his M.Sc. from the same department after he completed his Israel Defense Force service in 2002. His Dissertation was titled: Stop consonant spotting in continuous speech and was supervised by Dan Chazan from IBM Research Labs, Haifa. He continued his Ph.D. studies at the Hebrew University of Jerusalem until 2008. Prof. Yoram Singer supervised his thesis on Large Margin Algorithms for Discriminative Continuous Speech. == Career == Keshet was a Research Associate (postdoc) at IDIAP Research Institute, Martigny, Switzerland in 2007, and joined the TTI-Chicago and Department of Computer Science, University of Chicago, Chicago, IL in 2009 as Research Assistant Professor. In 2013, he returned to Israel and joined the Computer Science department at Bar-Ilan University as a senior lecturer and head of the Speech, Language, and Deep Learning Lab. In 2020, Keshet became a Founding Venture Partner at the Disruptive AI Venture Capital. In the same year, he also joined Amazon in Tel-Aviv as an Amazon Scholar. In 2022, Keshet joined the Faculty of Electrical and Computer Engineering at the Technion. == Research == Keshet's research work focuses on both machine learning and computational study of human speech and language. His work on speech and language concentrates on speech processing, speech recognition, acoustic phonetics, and pathological speech. In machine learning, Keshet is focused on deep learning and structured tasks. According to Google Scholar (September 2020), Keshet is one of the 15 most cited researchers in the field of spoken language processing. The algorithms that were developed in the Speech, Language, and Deep Learning Lab can analyze different pathological conditions in the throat and vocal cords based on the subject's voice. Other algorithms showed that the voice can be used to estimate physical and emotional state of the speaker. Another research led by Keshet suggested that it is possible to fool structured AI systems (like Google Voice). == Membership in professional societies == Keshet is the founder and chair of the Machine Learning for Speech and Language Processing Special Interest Group (SIGML) of the International Speech Communication Association (ISCA), from 2011. He is a senior member of the IEEE Signal Processing Society since 2018 and a member of ISCA since 2002. == Publications == Prof. Keshet has authored more than 70 scientific publications and edited one book. === Book === Joseph Keshet and Samy Bengio, Eds., Automatic Speech and Speaker Recognition: Large Margin and Kernel Methods, John Wiley & Sons, March 2009. === Selected articles === Jacob T. Cohen, Alma Cohen, Limor Benyamini, Yossi Adi, Joseph Keshet, Predicting glottal closure insufficiency using fundamental frequency contour analysis, Head & Neck, Journal of the Sciences and Specialities of the Head and Neck, Volume 41, Issue 7, pp. 2324–2331, July 2019. Yehoshua Dissen, Jacob Goldberger, and Joseph Keshet, Formant Estimation and Tracking: A Deep Learning Approach, Journal of the Acoustical Society of America, 145 (2), February 2019. Joseph Keshet, Automatic speech recognition: A primer for speech-language pathology researchers, International Journal of Speech-Language Pathology, Vol. 20 No. 6, pp. 599–609, 2018. Yossi Adi, Carsten Baum, Moustapha Cisse, Benny Pinkas, Joseph Keshet, Turning Your Weakness Into a Strength: Watermarking Deep Neural Networks by Backdooring, Usenix, 2018. Tzeviya Fuchs, Joseph Keshet, Spoken Term Detection Automatically Adjusted for a Given Threshold, IEEE Journal of Selected Topics in Signal Processing, Dec 2017, Volume 11, Issue 8, pp. 1–8. Moustapha Cisse, Yossi Adi, Natalia Neverova, Joseph Keshet, Houdini: Fooling Deep Structured Visual and Speech Recognition Models with Adversarial Examples, Neural Information and Processing Systems (NIPS), 2017. Joseph Keshet, Subhransu Maji, Tamir Hazan, and Tommi Jaakkola, Perturbation Models and PAC-Bayesian Generalization Bounds, in Perturbations, Optimization, and Statistics, Tamir Hazan, George Papandreou, and Daniel Tarlow, Eds., The MIT Press, 2016. Matthew Goldrick, Joseph Keshet, Erin Gustafson, Jordana Heller, and Jeremy Needle, Automatic Analysis of Slips of the Tongue: Insights into the Cognitive Architecture of Speech Production, Cognition, 149, 31–39, 2016. Joseph Keshet, Optimizing the Measure of Performance in Structured Prediction, in Advanced Structured Prediction, Sebastian Nowozin, Peter V. Gehler, Jeremy January, and Christoph H. Lampert, Eds., The MIT Press, 2014. Morgan Sonderegger and Joseph Keshet, Automatic Measurement of Voice Onset Time using Discriminative Structured Prediction, Journal of the Acoustical Society of America, Vol. 132, Issue 6, pp. 3965−3979, 2012. David McAllester, Tamir Hazan and Joseph Keshet, Direct Loss Minimization for Structured Prediction, The 24th Annual Conference on Neural Information Processing Systems (NIPS), 2010. Joseph Keshet, David Grangier and Samy Bengio, Discriminative Keyword Spotting, Speech Communication, Volume 51, Issue 4, pp. 317–329, April 2009. == Personal life == Keshet is married to Lital. They have three children.
Meta-learning (computer science)
Meta-learning is a subfield of machine learning where automatic learning algorithms are applied to metadata about machine learning experiments. As of 2017, the term had not found a standard interpretation, however the main goal is to use such metadata to understand how automatic learning can become flexible in solving learning problems, hence to improve the performance of existing learning algorithms or to learn (induce) the learning algorithm itself, hence the alternative term learning to learn. Flexibility is important because each learning algorithm is based on a set of assumptions about the data, its inductive bias. This means that it will only learn well if the bias matches the learning problem. A learning algorithm may perform very well in one domain, but not on the next. This poses strong restrictions on the use of machine learning or data mining techniques, since the relationship between the learning problem (often some kind of database) and the effectiveness of different learning algorithms is not yet understood. By using different kinds of metadata, like properties of the learning problem, algorithm properties (like performance measures), or patterns previously derived from the data, it is possible to learn, select, alter or combine different learning algorithms to effectively solve a given learning problem. Critiques of meta-learning approaches bear a strong resemblance to the critique of metaheuristic, a possibly related problem. A good analogy to meta-learning, and the inspiration for Jürgen Schmidhuber's early work (1987) and Yoshua Bengio et al.'s work (1991), considers that genetic evolution learns the learning procedure encoded in genes and executed in each individual's brain. In an open-ended hierarchical meta-learning system using genetic programming, better evolutionary methods can be learned by meta evolution, which itself can be improved by meta meta evolution, etc. == Definition == A proposed definition for a meta-learning system combines three requirements: The system must include a learning subsystem. Experience is gained by exploiting meta knowledge extracted in a previous learning episode on a single dataset, or from different domains. Learning bias must be chosen dynamically. Bias refers to the assumptions that influence the choice of explanatory hypotheses and not the notion of bias represented in the bias-variance dilemma. Meta-learning is concerned with two aspects of learning bias. Declarative bias specifies the representation of the space of hypotheses, and affects the size of the search space (e.g., represent hypotheses using linear functions only). Procedural bias imposes constraints on the ordering of the inductive hypotheses (e.g., preferring smaller hypotheses). == Common approaches == There are three common approaches: using (cyclic) networks with external or internal memory (model-based) learning effective distance metrics (metrics-based) explicitly optimizing model parameters for fast learning (optimization-based). === Model-Based === Model-based meta-learning models updates its parameters rapidly with a few training steps, which can be achieved by its internal architecture or controlled by another meta-learner model. ==== Memory-Augmented Neural Networks ==== A Memory-Augmented Neural Network, or MANN for short, is claimed to be able to encode new information quickly and thus to adapt to new tasks after only a few examples. ==== Meta Networks ==== Meta Networks (MetaNet) learns a meta-level knowledge across tasks and shifts its inductive biases via fast parameterization for rapid generalization. === Metric-Based === The core idea in metric-based meta-learning is similar to nearest neighbors algorithms, which weight is generated by a kernel function. It aims to learn a metric or distance function over objects. The notion of a good metric is problem-dependent. It should represent the relationship between inputs in the task space and facilitate problem solving. ==== Convolutional Siamese Neural Network ==== Siamese neural network is composed of two twin networks whose output is jointly trained. There is a function above to learn the relationship between input data sample pairs. The two networks are the same, sharing the same weight and network parameters. ==== Matching Networks ==== Matching Networks learn a network that maps a small labelled support set and an unlabelled example to its label, obviating the need for fine-tuning to adapt to new class types. ==== Relation Network ==== The Relation Network (RN), is trained end-to-end from scratch. During meta-learning, it learns to learn a deep distance metric to compare a small number of images within episodes, each of which is designed to simulate the few-shot setting. ==== Prototypical Networks ==== Prototypical Networks learn a metric space in which classification can be performed by computing distances to prototype representations of each class. Compared to recent approaches for few-shot learning, they reflect a simpler inductive bias that is beneficial in this limited-data regime, and achieve satisfied results. === Optimization-Based === What optimization-based meta-learning algorithms intend for is to adjust the optimization algorithm so that the model can be good at learning with a few examples. ==== LSTM Meta-Learner ==== LSTM-based meta-learner is to learn the exact optimization algorithm used to train another learner neural network classifier in the few-shot regime. The parametrization allows it to learn appropriate parameter updates specifically for the scenario where a set amount of updates will be made, while also learning a general initialization of the learner (classifier) network that allows for quick convergence of training. ==== Temporal Discreteness ==== Model-Agnostic Meta-Learning (MAML) is a fairly general optimization algorithm, compatible with any model that learns through gradient descent. ==== Reptile ==== Reptile is a remarkably simple meta-learning optimization algorithm, given that both of its components rely on meta-optimization through gradient descent and both are model-agnostic. == Examples == Some approaches which have been viewed as instances of meta-learning: Recurrent neural networks (RNNs) are universal computers. In 1993, Jürgen Schmidhuber showed how "self-referential" RNNs can in principle learn by backpropagation to run their own weight change algorithm, which may be quite different from backpropagation. In 2001, Sepp Hochreiter & A.S. Younger & P.R. Conwell built a successful supervised meta-learner based on Long short-term memory RNNs. It learned through backpropagation a learning algorithm for quadratic functions that is much faster than backpropagation. Researchers at Deepmind (Marcin Andrychowicz et al.) extended this approach to optimization in 2017. In the 1990s, Meta Reinforcement Learning or Meta RL was achieved in Schmidhuber's research group through self-modifying policies written in a universal programming language that contains special instructions for changing the policy itself. There is a single lifelong trial. The goal of the RL agent is to maximize reward. It learns to accelerate reward intake by continually improving its own learning algorithm which is part of the "self-referential" policy. An extreme type of Meta Reinforcement Learning is embodied by the Gödel machine, a theoretical construct which can inspect and modify any part of its own software which also contains a general theorem prover. It can achieve recursive self-improvement in a provably optimal way. Model-Agnostic Meta-Learning (MAML) was introduced in 2017 by Chelsea Finn et al. Given a sequence of tasks, the parameters of a given model are trained such that few iterations of gradient descent with few training data from a new task will lead to good generalization performance on that task. MAML "trains the model to be easy to fine-tune." MAML was successfully applied to few-shot image classification benchmarks and to policy-gradient-based reinforcement learning. Variational Bayes-Adaptive Deep RL (VariBAD) was introduced in 2019. While MAML is optimization-based, VariBAD is a model-based method for meta reinforcement learning, and leverages a variational autoencoder to capture the task information in an internal memory, thus conditioning its decision making on the task. When addressing a set of tasks, most meta learning approaches optimize the average score across all tasks. Hence, certain tasks may be sacrificed in favor of the average score, which is often unacceptable in real-world applications. By contrast, Robust Meta Reinforcement Learning (RoML) focuses on improving low-score tasks, increasing robustness to the selection of task. RoML works as a meta-algorithm, as it can be applied on top of other meta learning algorithms (such as MAML and VariBAD) to increase their robustness. It is applicable to both supervised meta learning and meta reinforcement learning. Discovering meta-knowledge works by inducing knowledge
Lin-Shan Lee
Lin-Shan Lee (Chinese: 李琳山; born 23 September 1952) is a Taiwanese computer scientist. == Education and career == Lee earned a bachelor's degree in electrical engineering from National Taiwan University in 1974, and pursued a doctorate in the same subject at Stanford University, graduating in 1977. He subsequently returned to Taiwan and joined the NTU faculty in 1982. Lee is a 1993 fellow of the Institute of Electrical and Electronics Engineers, recognized "[f]or contributions to computer voice input/output techniques for Mandarin Chinese and to engineering education." The International Speech Communication Association elevated him to fellow status in 2010 "[f]or his contributions to Chinese spoken language processing and speech information retrieval, and his service to the speech language community." In 2016, Lee was elected a member of Academia Sinica.