CHAOS (Chess Heuristics and Other Stuff) is a chess playing program that was developed by programmers working at the RCA Systems Programming division in the late 1960s. It played competitively in computer chess competitions in the 1970s and 1980s. It differed from other programs of that era in its look-ahead philosophy, choosing to use chess knowledge to evaluate fewer positions and continuations as opposed to simple evaluations that relied on deep look-ahead to avoid bad moves. == Introduction == CHAOS was originally developed by Ira Ruben, Fred Swartz, Victor Berman, Joe Winograd and William Toikka while working at RCA in Cinnaminson, NJ. Its name is an acronym for 'Chess Heuristics and Other Stuff.' Program development moved to the Computing Center of the University of Michigan when Swartz changed jobs, and Mike Alexander joined the development group. Swartz, Alexander and Berman were continuously group members from that point onward in CHAOS' evolution, as others of the original authors left and new members contributed episodically. Chess Senior Master Jack O'Keefe contributed to CHAOS' development from about 1980 onwards. CHAOS was written in Fortran, except for low-level board representation manipulations written in assembly language or C. Due to this portability, it ran on RCA, Univac and IBM-compatible mainframes in its lifetime. CHAOS heralds from the mainframe computing era when only machines of that capacity were able to play at a high level. Consequently, development and testing could only take place at off-peak times for production use of the machine. In a competition, CHAOS had to run on a dedicated mainframe with a telephone link to the match venue. In its later years, CHAOS ran on computers on the machine assembly floor of Amdahl Corporation on MTS. == Background == === Chess and artificial intelligence === Mathematicians Claude Shannon and Alan Turing, working separately, were the first to view playing chess as a challenge to machines. Working for AT&T / Bell Labs with its access to telephone switching equipment, Shannon built a relay-based machine that learned how to work its way through a two-dimensional, 5x5 cell maze in 1949. Shannon viewed this as an analogue of the way that organisms learn things about their natural environment. There is a random element to searching it, a memory element to benefit from the search outcome, and a reward element that reinforces learning when the global outcome is favorable to the organism. Soon afterward, Shannon wrote a mathematical analysis of the game of chess, published in 1950. Like with the maze, he broke down game play into the necessary elements for reinforcement learning. Associated with each board configuration a move will be made from, there is a numerical score. To decide what move to make, a player wants to maximize their own position's score after the move and to minimize their opponent's score (a minimax view). Since there are about 32 possible moves at each of the early stages of the game, and about 40 moves and responses in each game, then there are about 32 80 {\displaystyle 32^{80}} or about 10 120 {\displaystyle 10^{120}} possible games - an impossibly large set to evaluate completely. Therefore, there must be a way to limit the number of moves to look ahead for to find the best one. Reducing the game to these few key elements provided a way to think about human intelligence in general. Shannon became part of a wider group using computing machines to mimic aspects of human intelligence that grew into the general idea of artificial intelligence. (Other members of this group were John McCarthy, Herbert Simon, Allen Newell, Alan Kotok, Alex Bernstein and Richard Greenblatt.) The paradigm that evolved was that there was a quantification of the position on the board into a score, an evaluation method to find favorable outcomes (minimax, later alpha-beta pruning), and a strategy to manage the combinatorial explosion of the look-ahead possibilities. By the early 1960s, there were computer programs that played chess at a rudimentary level. They used very simple evaluation functions for each position and tried to search as far forward as was practical given the time constraints and available compute power. Naturally, programmers optimized their code to use the available computing resources. This led to a major philosophical divide among chess programs: those that tried to evaluate as many positions as possible, and those that tried to evaluate the most promising move sequences as deeply as possible. CHAOS was firmly in the camp believing only the most promising moves should be evaluated in depth. Said Swartz, "The 'brute force people' ... look at every (possible move) no matter what garbage it is. Most moves are just terrible, terrible moves, and most computing time is being spent on pure garbage." The program spent more time evaluating each board position in the expectation that it would find the most promising lines of play to explore in depth. In 1983, the then-fastest chess program (Belle) evaluated 110,000 positions per second, and typical programs 1000–50,000 per second, whereas CHAOS evaluated about 50-100 per second. === Machine learning and strategies to manage search === From about 1949 onward, Arthur Samuel began work for IBM on machine learning, culminating in a checkers-playing program in 1952 and publications on the topic. Concurrently, Christopher Strachey created Checkers, a program to play the board game of checkers in 1951, but it had no capacity to learn from its play. Checkers was chosen by both authors because it was simpler than chess yet contained the basic characteristics of an intellectual activity, and, in Samuel's view, was a test-bed in which heuristic procedures and learning processes could be evaluated quickly. Checker playing programs introduced the notion of the game tree and evaluating play to various depths to choose the best move. The complexity of chess, however, promoted it to the status of an analogue for human intelligence, and it attracted computer scientists' attention, who referred to it as research into artificial intelligence (AI). Like checkers, it required a numerical assessment of each arrangement of chess pieces on a board. It also required looking ahead to future moves to decide how to play the present position. Due to the enormous number of possible moves, there had to be a way to confine the look-ahead search to the most promising lines of play. From these factors, the notion of minimax score evaluation developed and, later, alpha-beta tree pruning to abandon looking at positions worse than any that have already been examined. === Chess search strategies === The AI community viewed artificial intelligence as comprising two parts: a way to symbolically quantify the knowledge in hand (a chess board position), and a set of heuristics to limit look-ahead to the consequences of a move. The early chess playing programs attempted to look forward as far as possible, perhaps to 3 moves ahead by each player, and to choose the best outcome. This led to the horizon effect, whereby a key move 4 or more moves ahead would be unexamined and therefore missed. Consequently, the programs were quite weak and heuristics to manage the search became important in their development. CHAOS used a selective search strategy with iterative widening. As chess programs evolved, they incorporated books of opening lines of play from historic sources. Nowadays, book moves are catalogued in machine-readable form, but originally programmers had to type them in. CHAOS had an extensive book for its time of around 10,000 moves that O'Keefe helped to develop. A problem with play from an opening book is the behavior of the program when the play leaves the book: the positional advantage may be so subtle that the evaluation scheme may be unable to understand it, leading to very wide and shallow searches to establish a line of play. The horizon effect again plagues move selection after leaving the book. CHAOS mitigated these problems by only using book lines that it could understand, and by relying on cached analyses of continuations out of the book made while the opponent's clock was running. == Game Play History == CHAOS played in twelve ACM computer chess tournaments and four World Computer Chess Championships (WCCC). Its debut was the ACM computer chess tournament in 1973, taking 2nd place. In 1974, it again won 2nd place in the WCCC, defeating the tournament favorite Chess 4.0 but losing to Kaissa. CHAOS was close to winning the 1980 WCCC, but lost to Belle in a playoff. The 1985 ACM computer chess tournament was CHAOS' last competition. One of CHAOS' notable victories was over Chess 4.0 at the 1974 WCCC tournament. Chess 4.0 was unbeaten by any other program up until then. Playing as white, CHAOS made a knight sacrifice (16 Nd4-e6!!) that traded material for open lines of attack and eventually won the game. CHAOS’ authors thought the move was due to a
Active learning (machine learning)
Active learning is a special case of machine learning in which a learning algorithm can interactively query a human user (or some other information source) to label new data points with the desired outputs. The human user must possess expertise in the problem domain, including the ability to consult authoritative sources when necessary. In statistics literature, it is sometimes also called optimal experimental design. The information source is also called teacher or oracle. There are situations in which unlabeled data is abundant but manual labeling is expensive. In such a scenario, learning algorithms can actively query the teacher for labels. Since the learner chooses the examples, the number of examples to learn a concept can often be much lower than the number required in normal supervised learning. However, there is a risk that the algorithm is overwhelmed by uninformative examples. Recent developments are dedicated to multi-label active learning, hybrid active learning and active learning in a single-pass (on-line) context, combining concepts from the field of machine learning (e.g. conflict and ignorance) with adaptive, incremental learning policies in the field of online machine learning. Using active learning allows for faster development of a machine learning algorithm, when comparative updates would require a quantum or super computer. Large-scale active learning projects may benefit from crowdsourcing frameworks such as Amazon Mechanical Turk that include many humans in the active learning loop. == Definitions == Let T be the total set of all data under consideration. For example, in a protein engineering problem, T would include all proteins that are known to have a certain interesting activity and all additional proteins that one might want to test for that activity. During each iteration, i, T is broken up into three subsets T K , i {\displaystyle \mathbf {T} _{K,i}} : Data points where the label is known. T U , i {\displaystyle \mathbf {T} _{U,i}} : Data points where the label is unknown. T C , i {\displaystyle \mathbf {T} _{C,i}} : A subset of TU,i that is chosen to be labeled. Most of the current research in active learning involves the best method to choose the data points for TC,i. == Scenarios == Pool-based sampling: In this approach, which is the most well known scenario, the learning algorithm attempts to evaluate the entire dataset before selecting data points (instances) for labeling. It is often initially trained on a fully labeled subset of the data using a machine-learning method such as logistic regression or SVM that yields class-membership probabilities for individual data instances. The candidate instances are those for which the prediction is most ambiguous. Instances are drawn from the entire data pool and assigned a confidence score, a measurement of how well the learner "understands" the data. The system then selects the instances for which it is the least confident and queries the teacher for the labels. The theoretical drawback of pool-based sampling is that it is memory-intensive and is therefore limited in its capacity to handle enormous datasets, but in practice, the rate-limiting factor is that the teacher is typically a (fatiguable) human expert who must be paid for their effort, rather than computer memory. Stream-based selective sampling: Here, each consecutive unlabeled instance is examined one at a time with the machine evaluating the informativeness of each item against its query parameters. The learner decides for itself whether to assign a label or query the teacher for each datapoint. As contrasted with Pool-based sampling, the obvious drawback of stream-based methods is that the learning algorithm does not have sufficient information, early in the process, to make a sound assign-label-vs ask-teacher decision, and it does not capitalize as efficiently on the presence of already labeled data. Therefore, the teacher is likely to spend more effort in supplying labels than with the pool-based approach. Membership query synthesis: This is where the learner generates synthetic data from an underlying natural distribution. For example, if the dataset are pictures of humans and animals, the learner could send a clipped image of a leg to the teacher and query if this appendage belongs to an animal or human. This is particularly useful if the dataset is small. The challenge here, as with all synthetic-data-generation efforts, is in ensuring that the synthetic data is consistent in terms of meeting the constraints on real data. As the number of variables/features in the input data increase, and strong dependencies between variables exist, it becomes increasingly difficult to generate synthetic data with sufficient fidelity. For example, to create a synthetic data set for human laboratory-test values, the sum of the various white blood cell (WBC) components in a white blood cell differential must equal 100, since the component numbers are really percentages. Similarly, the enzymes alanine transaminase (ALT) and aspartate transaminase (AST) measure liver function (though AST is also produced by other tissues, e.g., lung, pancreas) A synthetic data point with AST at the lower limit of normal range (8–33 units/L) with an ALT several times above normal range (4–35 units/L) in a simulated chronically ill patient would be physiologically impossible. == Query strategies == Algorithms for determining which data points should be labeled can be organized into a number of different categories, based upon their purpose: Balance exploration and exploitation: the choice of examples to label is seen as a dilemma between the exploration and the exploitation over the data space representation. This strategy manages this compromise by modelling the active learning problem as a contextual bandit problem. For example, Bouneffouf et al. propose a sequential algorithm named Active Thompson Sampling (ATS), which, in each round, assigns a sampling distribution on the pool, samples one point from this distribution, and queries the oracle for this sample point label. Expected model change: label those points that would most change the current model. Expected error reduction: label those points that would most reduce the model's generalization error. Exponentiated Gradient Exploration for Active Learning: In this paper, the author proposes a sequential algorithm named exponentiated gradient (EG)-active that can improve any active learning algorithm by an optimal random exploration. Uncertainty sampling: label those points for which the current model is least certain as to what the correct output should be. Query by committee: a variety of models are trained on the current labeled data, and vote on the output for unlabeled data; label those points for which the "committee" disagrees the most Querying from diverse subspaces or partitions: When the underlying model is a forest of trees, the leaf nodes might represent (overlapping) partitions of the original feature space. This offers the possibility of selecting instances from non-overlapping or minimally overlapping partitions for labeling. Variance reduction: label those points that would minimize output variance, which is one of the components of error. Conformal prediction: predicts that a new data point will have a label similar to old data points in some specified way and degree of the similarity within the old examples is used to estimate the confidence in the prediction. Mismatch-first farthest-traversal: The primary selection criterion is the prediction mismatch between the current model and nearest-neighbour prediction. It targets on wrongly predicted data points. The second selection criterion is the distance to previously selected data, the farthest first. It aims at optimizing the diversity of selected data. User-centered labeling strategies: Learning is accomplished by applying dimensionality reduction to graphs and figures like scatter plots. Then the user is asked to label the compiled data (categorical, numerical, relevance scores, relation between two instances). A wide variety of algorithms have been studied that fall into these categories. While the traditional AL strategies can achieve remarkable performance, it is often challenging to predict in advance which strategy is the most suitable in a particular situation. In recent years, meta-learning algorithms have been gaining in popularity. Some of them have been proposed to tackle the problem of learning AL strategies instead of relying on manually designed strategies. A benchmark which compares 'meta-learning approaches to active learning' to 'traditional heuristic-based Active Learning' may give intuitions if 'Learning active learning' is at the crossroads == Minimum marginal hyperplane == Some active learning algorithms are built upon support-vector machines (SVMs) and exploit the structure of the SVM to determine which data points to label. Such methods usually calculate the margin, W, of each u
Attention inequality
Attention inequality is the inequality of distribution of attention across users on social networks, people in general, and for scientific papers. Yun Family Foundation introduced "Attention Inequality Coefficient" as a measure of inequality in attention and arguments it by the close interconnection with wealth inequality. == Relationship to economic inequality == Attention inequality is related to economic inequality since attention is an economically scarce good. The same measures and concepts as in classical economy can be applied for attention economy. The relationship develops also beyond the conceptual level—considering the AIDA process, attention is the prerequisite for real monetary income on the Internet. On data of 2018, a significant relationship between likes and comments on Facebook to donations is proven for non-profit organizations. == Attention economy == The attention economy refers to the practice of maximizing the attention users give to a product for advertising-related reasons. Attention economy remains one of the most common forms of advertising, and has been steadily increasing thanks to new technologies such as television, internet and social media. It is one of the most widely-used approaches to economy for its effectiveness for maximising the noticeability of a certain product. == Attention inequality in social media == In social media, attention inequality refers to the unequal distribution of users' attention on social media platforms. This means that instead of an equal distribution of attention, fewer sources receive a disproportionate share of attention, leaving many unnoticed. This phenomenon is possibly the result of social media algorithms, which are commonly designed to drive maximum engagement. This phenomenon is a large factor in the polarization and creation of echo-chambers. Social media algorithms tend to note content that is already performing well and display it to more users, while content that is equally engaging or well-made is not recommended to users. Posts that trigger strong emotions usually out-perform more "uncontroversial" content. When many users interact with the post, it signals the algorithm that the specific post drives engagement. The algorithm then tends to recommend that type of content to an exponential number of people, potentially outperforming "un-emotional" content. These factors, when combined, tend to create an unequal social media environment. == Attention inequality in science == According to a recent 2025 study about research inequality among scientists published in Information Processing and Management, scientific discourse is restricted to a small group of connected scientists, and is frequently not an accurate representation of the whole scientific community. Using citation-network analysis in the fields of nanoscience and chemical physics, the study claims that a group of connected scientists has a significant notability in the scientific community. The calculated connection strength between these scientists is estimated to be about 4.5, the study also says that these authors cite each other four times more often than would be predicted in a random network, whereas ordinary scientists that exist outside of this group only reach an estimated connection strength of 0.9. The study findings suggest that that scientific attention is not distributed by merit, but rather by the connectedness of the scientists involved in the research. == Extent == As data of 2008 shows, 50% of the attention is concentrated on approximately 0.2% of all hostnames, and 80% on 5% of hostnames. The Gini coefficient of attention distribution lay in 2008 at over 0.921 for such commercial domains names as ac.jp and at 0.985 for .org-domains. The Gini coefficient was measured on Twitter in 2016 for the number of followers as 0.9412, for the number of mentions as 0.9133, and for the number of retweets as 0.9034. For comparison, the world's income Gini coefficient was 0.68 in 2005 and 0.904 in 2018. More than 96% of all followers, 93% of the retweets, and 93% of all mentions are owned by 20% of Twitter. == Causes == At least for scientific papers, today's consensus states that inequality is unexplainable by variations of quality and individual talent. The Matthew effect plays a significant role in the emergence of attention inequality—those who already enjoy large amounts of attention get even more attention, and those who do not lose even more. Ranking algorithms based on relevance to the user have been found to alleviate the inequality of the number of posts across topics.
Server-sent events
Server-Sent Events (SSE) is a server push technology enabling a client to receive automatic updates from a server via an HTTP connection, and describes how servers can initiate data transmission towards clients once an initial client connection has been established. They are commonly used to send message updates or continuous data streams to a browser client and designed to enhance native, cross-browser streaming through a JavaScript API called EventSource, through which a client requests a particular URL in order to receive an event stream. The EventSource API is standardized as part of HTML Living Standard by the WHATWG. The media type for SSE is text/event-stream. All modern browsers support server-sent events: Firefox 6+, Google Chrome 6+, Opera 11.5+, Safari 5+, Microsoft Edge 79+, Brave. Since SSE does not use either persistent connections nor chunked transfer encoding, HTTP/1.1 is not a technical requirement. == History == The SSE mechanism was first specified by Ian Hickson as part of the "WHATWG Web Applications 1.0" proposal starting in 2004. In September 2006, the Opera web browser implemented the experimental technology in a feature called "Server-Sent Events". The W3C published Server-Sent Events as a Recommendation on February 3, 2015, after years of development through Working Drafts and Candidate Recommendations. == Example == == Technology == When sending high-frequency data , the server must manage backpressure to prevent saturating clients. This is mitigated in the following ways: Client-side buffering: Browsers have limited buffer space for incoming server-sent events Adaptive rate limiting: Servers can adjust event frequency and monitor connection health Event batching: Combining multiple events into larger and less frequent transmissions
Content reference identifier
A content reference identifier or CRID is a concept from the standardization work done by the TV-Anytime forum. It is or closely matches the concept of the Uniform Resource Locator, or URL, as used on the World-Wide Web: A unit of content, in a broadcast stream, can be referred to by its globally unique CRID in the same way that a webpage can be referred to by its globally unique URL on the web. The concept of CRID permits referencing contents unambiguously, regardless of their location, i.e., without knowing specific broadcast information (time, date and channel) or how to obtain them through a network, for instance, by means of a streaming service or by downloading a file from an Internet server. The receiver must be capable of resolving these unambiguous references, i.e. of translating them into specific data that will allow it to obtain the location of that content in order to acquire it. This makes it possible for recording processes to take place without knowing that information, and even without knowing beforehand the duration of the content to be recorded: a complete series by a simple click, a program that has not been scheduled yet, a set of programs grouped by a specific criterion... This framework allows for the separation between the reference to a given content (the CRID) and the necessary information to acquire it, which is called a “locator”. Each CRID may lead to one or more locators which will represent different copies of the same content. They may be identical copies broadcast in different channels or dates, or cost different prices. They may also be distinct copies with different technical parameters such as format or quality. It may also be the case that the resolution process of a CRID provides another CRID as a result (for example, its reference in a different network, where it has an alternative identifier assigned by a different operator) or a set of CRIDs (for instance, if the original CRID represents a TV series, in which case the resolution process would result in the list of CRIDs representing each episode). From the above it can be concluded that provided that a given content can belong to many groups (each possibly defined by distinctive qualities), it is possible that many CRIDs carry the same content. That is, several CRIDs may be resolved into the same locator. A CRID is not exactly a universal, unique and exclusive identifier for a given content. It is closely related to the authority that creates it, to the resolution service provider, and to the content provider in such a way that the same content may have different CRIDs depending on the field in which they are used (for example, a different one for each television operator that has the rights to broadcast the content). == Format == A CRID is specified much like URLs. In fact, a CRID is a so-called URI. Typically, the content creator, the broadcaster or a third party will use their DNS-names in a combination with a product-specific name to create globally unique CRIDs. That is, the syntax of a CRID is: crid://authority/data The authority field represents the entity that created the CRID and its format is that of a DNS name. The data field represents a string of characters that will unambiguously identify the content within the authority scope (it is a string of characters assigned by the authority itself). As an example, let's assume that BBC wanted to make a CRID for (all the programs of) the Olympics in China. It may have looked something like this crid://bbc.co.uk/olympics/2008/ This would be a group CRID, that is, a CRID representing a group of contents. Then, to refer to a specific event – such as the women's shot-put final – they could have used the following inside their metadata. crid://bbc.co.uk/olympics/2008/final/shotput/women Currently, four types of CRIDs are playing a major role in some unidirectional television networks: programme CRID, series CRID, group CRID, and recommendation CRID. One of the most important applications of CRIDs is the so-called series link recording function (SL) of modern digital video recorders (DVR, PVR). In turn, a locator is a string of characters that contains all the necessary information for a receiver to find and acquire a given content, whether it is received through a transport stream, located in local storage, downloaded as a file from an Internet server, or through a streaming service. For example, a DVB locator will include all the necessary parameters to identify a specific content within a transport stream: network, transport stream, service, table and/or event identifiers. The locators' format, as established in TV-Anytime, is quite generic and simple, and corresponds to: [transport-mechanism]:[specific-data] The first part of the locator's format (the transport mechanism) must be a string of characters that is unique for each mechanism (transport stream, local file, HTTP Internet access...). The second part must be unambiguous only within the scope of a given transport mechanism and will be standardized by the organism in charge of the regulation of the mechanism itself. For instance, a DVB locator to identify a content within the transport stream of networks that follow this standard would be: dvb://112.4a2.5ec;2d22~20121212T220000Z—PT01H30M which would indicate a content (identified by the string “2d22”) that airs on a channel available on a DVB network identified by the address “112.4a2.5ec” (network “112”, transport stream “4a2” and service “5ec”), on 12 December 2012 at 10 p.m. and with a duration of 90 minutes. == The location resolution process == The location resolution process is the procedure by which, starting from the CRID of a given content, one or several locators of that content are obtained. Resolving a CRID can be a direct process, which leads immediately to one or many locators, or it may also happen that in the first place one or many intermediate CRIDs are returned, which must undergo the same procedure to finally obtain one or several locators. This procedure involves some information elements, among which we find two structures named resolving authority record (RAR) and ContentReferencingTable, respectively. Consulting them repeatedly will take the receiver from a CRID to one or many locators that will allow it to acquire the content. The RAR table The RAR table is one or many data structures that provide the receiver, for each authority that submits CRIDs, information on the corresponding resolution service provider. Among other things, it informs about which mechanism is used to provide information to resolve the CRIDs from each authority. That is, one or many RAR records must exist for each authority that indicate the receiver where it has to go to resolve the CRIDs of that particular authority. For example, in the record of the figure (expressed by means of a XML structure, according to the XML Schema defined in the TV-Anytime) there is an authority called “tve.es”, whose resolution service provider is the entity “rtve.es”, available on the URL "http://tva.rtve.es/locres/tve", which means there is resolution information in that URL. These RAR records will have reached the receiver in an indefinite form, unimportant for the TV-Anytime specification, which will depend on the specific transport mechanism of the network to which the receiver is connected. Each family of standards that regulates distribution networks (DVB, ATSC, ISDB, IPTV...) will have previously defined such procedure, which will be used by devices certified according to those standards. The ContentReferencingTable table The second structure involved in the location resolution process is a proper resolution table which, given a content's CRID, returns one or several locators that enable the receiver to access an instance of that content, or one or many CRIDs that allow it to move forward in the resolution process. The figure shows an example of this second structure, an XML document according to the specifications of the XML Schema defined in TV-Anytime. In it, several sections are included (
Simple interactive object extraction
Simple interactive object extraction (SIOX) is an algorithm for extracting foreground objects from color images and videos with very little user interaction. It has been implemented as "foreground selection" tool in the GIMP (since version 2.3.3), as part of the tracer tool in Inkscape (since 0.44pre3), and as function in ImageJ and Fiji (plug-in). Experimental implementations were also reported for Blender and Krita. Although the algorithm was originally designed for videos, virtually all implementations use SIOX primarily for still image segmentation. In fact, it is often said to be the current de facto standard for this task in the open-source world. Initially, a free hand selection tool is used to specify the region of interest. It must contain all foreground objects to extract and as few background as possible. The pixels outside the region of interest form the sure background while the inner region define a superset of the foreground, i.e. the unknown region. A so-called foreground brush is then used to mark representative foreground regions. The algorithm outputs a selection mask. The selection can be refined by either adding further foreground markings or by adding background markings using the background brush. Technically, the algorithm performs the following steps: Create a set of representative colors for sure foreground and sure background, the so-called color signatures. Assign all image points to foreground or background by a weighted nearest neighbor search in the color signatures. Apply some standard image processing operations like erode, dilate, and blur to remove artifacts. Find the connected foreground components that are either large enough or marked by the user. For video segmentation the sure background and sure foreground regions are learned from motion statistics. SIOX also features tools that allow sub-pixel accurate refinement of edges and high texture areas, the so-called "detail refinement brushes". As with all segmentation algorithms, there are always pictures where the algorithm does not yield perfect results. The most critical drawback of SIOX is the color dependence. Although many photos are well-separable by color, the algorithm cannot deal with camouflage. If the foreground and background share many identical shades of similar colors, the algorithm might give a result with parts missing or incorrectly classified foreground. SIOX performs about equally well on different benchmarks compared to graph-based segmentation methods, such as Grabcut. SIOX is, however, more noise robust and can therefore also be used for the segmentation of videos. Graph-based segmentation methods search for a minimum cut and therefore tend to not perform optimally with complex structures. The algorithm has initially been developed at the department of computer science at Freie Universitaet Berlin. The main developer, Gerald Friedland, is now faculty at the EECS department of the University of California at Berkeley and also a Principal Data Scientist at Lawrence Livermore National Lab. He continues to support the development through mentoring, e.g. in the Google Summer of Code.
T.38
T.38 is an ITU recommendation for allowing transmission of fax over IP networks (FoIP) in real time. == History == The T.38 fax relay standard was devised in 1998 as a way to transport faxes across IP networks between existing Group 3 (G3) fax terminals. T.4 and related fax standards were published by the ITU in 1980, before the rise of the Internet. In the late 1990s, VoIP, or voice over IP, began to gain ground as an alternative to the conventional public switched telephone network (PSTN). However, because most VoIP systems are optimized (through their use of aggressive lossy bandwidth-saving compression) for voice rather than data calls, conventional fax machines worked poorly or not at all on them due to the network impairments such as delay, jitter, packet loss, and so on. Thus, some way of transmitting fax over IP was needed. == Overview == In practical scenarios, a T.38 fax call has at least part of the call being carried over PSTN, although this is not required by the T.38 definition, and two T.38 devices can send faxes to each other. This particular type of device is called Internet-Aware Fax device, or IAF, and it is capable of initiating or completing a fax call towards the IP network. The typical scenario where T.38 is used is – T.38 fax relay – where a T.30 fax device sends a fax over PSTN to a T.38 fax gateway which converts or encapsulates the T.30 protocol into a T.38 data stream. This is then sent either to a T.38-enabled end point such as fax machine or fax server or another T.38 gateway that converts it back to a PSTN PCM or analog signal and terminates the fax on a T.30 device. The T.38 recommendation defines the use of both TCP and UDP to transport T.38 packets. Implementations tend to use UDP, due to TCP's requirement for acknowledgement packets and resulting retransmission during packet loss, which introduces delays. When using UDP, T.38 copes with packet loss by using redundant data packets. T.38 is not a call setup protocol, thus the T.38 devices need to use standard call setup protocols to negotiate the T.38 call, e.g. H.323, SIP & MGCP. == Operation == There are two primary ways that fax transactions are conveyed across packet networks. The T.37 standard specifies how a fax image is encapsulated in e-mail and transported, ultimately, to the recipient using a store-and-forward process through intermediary entities. T.38, however, defines a protocol that supports the use of the T.30 protocol in both the sender and recipient terminals. (See diagram above.) T.38 lets one transmit a fax across an IP network in real time, just as the original G3 fax standards did for the traditional (time-division multiplexed (TDM)) network, also called the public switched telephone network or PSTN. A special protocol is needed for real-time fax over IP (Internet Protocol) since existing fax terminals only supported PSTN connections, where the information flow was generally smooth and uninterrupted, as opposed to the jittery arrival of IP packets. The trick was to come up with a protocol that makes the IP network “invisible” to the endpoint fax terminals, which would mean the user of a legacy fax terminal need not know that the fax call was traversing an IP network. The network interconnections supported by T.38 are shown above. The two fax terminals on either side of the figure communicate using the T.30 fax protocol published by the ITU in 1980. Interconnection of the PSTN with the IP packet network requires a “gateway” between the PSTN and IP networks. PSTN-IP Gateways support TDM voice on the PSTN side and VoIP and FoIP on the packet side. For voice sessions, the gateway will take in voice packets on the IP side, accumulate a few packets to ensure a smooth flow of TDM data upon their release, and then meter them out over TDM where they eventually are heard by a human or stored on a computer for later playback. The gateway employs packet-management techniques to enhance the quality of the speech in the presence of network errors by taking advantage of the natural ability of a listener to not really hear the occasional missing or repeated packet. But facsimile data are transmitted by modems, which aren't as forgiving as the human ear is for speech. Missing packets will often cause a fax session to fail at worst or create one or more image lines in error at best. So the job of T.38 is to “fool” the terminal into “thinking” that it's communicating directly with another T.30 terminal. It will also correct for network delays with so-called spoofing techniques, and missing or delayed packets with fax-aware buffer-management techniques. Spoofing refers to the logic implemented in the protocol engine of a T.38 relay that modifies the protocol commands and responses on the TDM side to keep network delays on the IP side from causing the transaction to fail. This is done, for example, by padding image lines or deliberately causing a message to be re-transmitted to render network delays transparent to the sending/receiving fax terminals. Networks that do not have packet loss or excessive delay can exhibit acceptable fax performance without T.38, provided the PCM clocks in all gateways are of very high accuracy (explained below). T.38 not only removes the effect of PCM clocks not being synchronized, but also reduces the required network bandwidth by a factor of 10, while it corrects for packet loss and delay. === Bandwidth reduction === As shown in the diagram below, a T.38 gateway is composed of two primary elements: the fax modems and the T.38 subsystem. The fax modems modulate and demodulate the PCM samples of the analog data, turning the sampled-data representation of the fax terminal's analog signal to its binary translation, and vice versa. The PSTN network samples the analog signal of a voice or modem signal (it doesn't know the difference) 8,000 times per second (SPS), and encodes them as 8-bit data bytes. This means 8000 samples-per-second times 8-bits per sample, or 64,000 bits per second (bit/s) to represent the modem (or voice) data in one direction. For both directions the modem transaction consumes 128,000 bits of network bandwidth. However, the typical modem in a fax terminal transmits the image data at 33,600 bit/s, so if the analog data are first converted to the digital content they represent, only 33,600 bits (plus network overhead of a few bytes) are needed. And since T.30 fax is a half-duplex protocol, the network is only needed for one direction at a time. Refer to RFC 3261 === PCM clock synchronization === In the diagram above, there is a sample-rate clock in the fax terminal and one in the gateway's modems that is used to trigger the sampling of the analog line 8,000 times per second. These clocks are usually quite accurate, but in some low-cost terminal adapters (a one or two-line gateway) the PCM clock can be surprisingly inaccurate. If the terminal is sending data to the gateway, and the gateway's clock is too slow, the buffers (jitter buffers) in the gateway will eventually overflow, causing the transaction to fail. Since the difference is often quite small, this problem occurs on long, detailed fax images giving the clocks more time to cause the jitter buffer in gateway to either underflow or overflow, which is just the same as missing or duplicated packets. === Packet loss === T.38 provides facilities to eliminate the effects of packet loss through data redundancy. When a packet is sent, either zero, one, two, three, or even more of the previously sent packets are repeated. (The specification does not impose a limit.) This increases the network bandwidth required (it's still much less than not using T.38) but it allows the receiving gateway to reconstruct the complete packet sequence, even with a fairly high level of packet loss. == Related standards == T.4 is the umbrella specification for fax. It specifies the standard image sizes, two forms of image-data compression (encoding), the image-data format, and references, T.30 and the various modem standards. T.6 specifies a compression scheme that reduces the time required to transmit an image by roughly 50-percent. T.30 specifies the procedures that a sending and receiving terminal use to set up a fax call, determine the image size, encoding, and transfer speed, the demarcation between pages, and the termination of the call. T.30 also references the various modem standards. V.21, V.27ter, V.29, V.17, V.34: ITU modem standards used in facsimile. The first three were ratified prior to 1980, and were specified in the original T.4 and T.30 standards. V.34 was published for fax in 1994. T.37 The ITU standard for sending a fax-image file via e-mail to the intended recipient of a fax. G.711 pass through - this is where the T.30 fax call is carried in a VoIP call encoded as audio. This is sensitive to network packet loss, jitter and clock synchronization. When using voice high-compression encoding techniques such as, but not limited to, G.729, some fax tonal signa