Meta Content Framework (MCF) is a specification of a content format for structuring metadata about web sites and other data. == History == MCF was developed by Ramanathan V. Guha at Apple Computer's Advanced Technology Group between 1995 and 1997. Rooted in knowledge-representation systems such as CycL, KRL, and KIF, it sought to describe objects, their attributes, and the relationships between them. One application of MCF was HotSauce, also developed by Guha while at Apple. It generated a 3D visualization of a web site's table of contents, based on MCF descriptions. By late 1996, a few hundred sites were creating MCF files and Apple HotSauce allowed users to browse these MCF representations in 3D. When the research project was discontinued, Guha left Apple for Netscape, where, in collaboration with Tim Bray, he adapted MCF to use XML and created the first version of the Resource Description Framework (RDF). == MCF format == An MCF file consists of one or more blocks, each corresponding to an entity. A block looks like this:The identifier is a unique identifier for that entity (more on the scope of the identifier below) and is used to refer to that entity. The following lines each specify a property and one or more values, separated by commas. Each value can be a reference to another entity (via its identifier), a string (enclosed by double quotes) or a number. For example:NOTE: The identifier must not include a comma (,) and must not be enclosed within double quotes. A common parsing failure is due to odd number of unescaped double quotes in text. For instance, "foo bar" baz" needs to be "foo bar\" baz". Commas within double quotes are not considered as value separators. Every entity has at least one property: typeOf.
Actifsource
Actifsource is a domain-specific modeling workbench. It is realized as plug-in for the software development environment Eclipse. Actifsource supports the creation of multiple domain models which can be linked together. It comes with a UML-like graphical editor to create domain-specific languages and a general graphical editor to edit structures in the created languages. It supports code generation using user-defined generic code templates which are directly linked to the domain models. Code generation is integrated into Eclipse's incremental build process. == Interoperability == Actifsource can use models from other modelling tools by importing and exporting the ecore format which is defined by the Eclipse Modeling Framework. == Licensing policy == There are two versions of actifsource available: The free community edition which can be used freely for non-commercial projects and the enterprise edition which contains additional features. The enterprise edition comes with customer support and maintenance for a limited period of time. This package allows the customers to upgrade to new versions and maintenance releases during their support period.
Bottom-up and top-down approaches
Bottom-up and top-down are strategies of composition and decomposition in fields as diverse as information processing and ordering knowledge, software, humanistic and scientific theories (see systemics), time management, and organization. In practice they can be seen as a style of thinking, teaching, or leadership. A top-down approach (also known as stepwise design and stepwise refinement and in some cases used as a synonym of decomposition) is essentially the breaking down of a system to gain insight into its compositional subsystems in a reverse engineering fashion. In a top-down approach an overview of the system is formulated, specifying, but not detailing, any first-level subsystems. Each subsystem is then refined in yet greater detail, sometimes in many additional subsystem levels, until the entire specification is reduced to base elements. A top-down model is often specified with the assistance of black boxes, which makes it easier to manipulate. However, black boxes may fail to clarify elementary mechanisms or be detailed enough to realistically validate the model. A top-down approach starts with the big picture, then breaks down into smaller segments. A bottom-up approach is the piecing together of systems to give rise to more complex systems, thus making the original systems subsystems of the emergent system. Bottom-up processing is a type of information processing based on incoming data from the environment to form a perception. From a cognitive psychology perspective, information enters the eyes in one direction (sensory input, or the "bottom"), and is then turned into an image by the brain that can be interpreted and recognized as a perception (output that is "built up" from processing to final cognition). In a bottom-up approach the individual base elements of the system are first specified in great detail. These elements are then linked together to form larger subsystems, which then in turn are linked, sometimes in many levels, until a complete top-level system is formed. This strategy often resembles a "seed" model, by which the beginnings are small but eventually grow in complexity and completeness. But "organic strategies" may result in a tangle of elements and subsystems, developed in isolation and subject to local optimization as opposed to meeting a global purpose. == Computer science == === Software development === In the software development process, the top-down and bottom-up approaches play a key role. Top-down approaches emphasize planning and a complete understanding of the system. It is inherent that no coding can begin until a sufficient level of detail has been reached in the design of at least some part of the system. Top-down approaches are implemented by attaching the stubs in place of the module. But these delay testing of the ultimate functional units of a system until significant design is complete. Bottom-up emphasizes coding and early testing, which can begin as soon as the first module has been specified. But this approach runs the risk that modules may be coded without having a clear idea of how they link to other parts of the system, and that such linking may not be as easy as first thought. Re-usability of code is one of the main benefits of a bottom-up approach. Top-down design was promoted in the 1970s by IBM researchers Harlan Mills and Niklaus Wirth. Mills developed structured programming concepts for practical use and tested them in a 1969 project to automate the New York Times morgue index. The engineering and management success of this project led to the spread of the top-down approach through IBM and the rest of the computer industry. Among other achievements, Niklaus Wirth, the developer of Pascal programming language, wrote the influential paper Program Development by Stepwise Refinement. Since Niklaus Wirth went on to develop languages such as Modula and Oberon (where one could define a module before knowing about the entire program specification), one can infer that top-down programming was not strictly what he promoted. Top-down methods were favored in software engineering until the late 1980s, and object-oriented programming assisted in demonstrating the idea that both aspects of top-down and bottom-up programming could be used. Modern software design approaches usually combine top-down and bottom-up approaches. Although an understanding of the complete system is usually considered necessary for good design—leading theoretically to a top-down approach—most software projects attempt to make use of existing code to some degree. Pre-existing modules give designs a bottom-up flavor. === Programming === Top-down is a programming style, the mainstay of traditional procedural languages, in which design begins by specifying complex pieces and then dividing them into successively smaller pieces. The technique for writing a program using top-down methods is to write a main procedure that names all the major functions it will need. Later, the programming team looks at the requirements of each of those functions and the process is repeated. These compartmentalized subroutines eventually will perform actions so simple they can be easily and concisely coded. When all the various subroutines have been coded the program is ready for testing. By defining how the application comes together at a high level, lower-level work can be self-contained. In a bottom-up approach the individual base elements of the system are first specified in great detail. These elements are then linked together to form larger subsystems, which in turn are linked, sometimes at many levels, until a complete top-level system is formed. This strategy often resembles a "seed" model, by which the beginnings are small, but eventually grow in complexity and completeness. Object-oriented programming (OOP) is a paradigm that uses "objects" to design applications and computer programs. In mechanical engineering with software programs such as Pro/ENGINEER, Solidworks, and Autodesk Inventor users can design products as pieces not part of the whole and later add those pieces together to form assemblies like building with Lego. Engineers call this "piece part design". === Parsing === Parsing is the process of analyzing an input sequence (such as that read from a file or a keyboard) in order to determine its grammatical structure. This method is used in the analysis of both natural languages and computer languages, as in a compiler. Bottom-up parsing is parsing strategy that recognizes the text's lowest-level small details first, before its mid-level structures, and leaves the highest-level overall structure to last. In top-down parsing, on the other hand, one first looks at the highest level of the parse tree and works down the parse tree by using the rewriting rules of a formal grammar. == Natural sciences == === Nanotechnology === Top-down and bottom-up are two approaches for the manufacture of products. These terms were first applied to the field of nanotechnology by the Foresight Institute in 1989 to distinguish between molecular manufacturing (to mass-produce large atomically precise objects) and conventional manufacturing (which can mass-produce large objects that are not atomically precise). Bottom-up approaches seek to have smaller (usually molecular) components built up into more complex assemblies, while top-down approaches seek to create nanoscale devices by using larger, externally controlled ones to direct their assembly. Certain valuable nanostructures, such as Silicon nanowires, can be fabricated using either approach, with processing methods selected on the basis of targeted applications. A top-down approach often uses the traditional workshop or microfabrication methods where externally controlled tools are used to cut, mill, and shape materials into the desired shape and order. Micropatterning techniques, such as photolithography and inkjet printing belong to this category. Vapor treatment can be regarded as a new top-down secondary approaches to engineer nanostructures. Bottom-up approaches, in contrast, use the chemical properties of single molecules to cause single-molecule components to (a) self-organize or self-assemble into some useful conformation, or (b) rely on positional assembly. These approaches use the concepts of molecular self-assembly and/or molecular recognition. See also Supramolecular chemistry. Such bottom-up approaches should, broadly speaking, be able to produce devices in parallel and much cheaper than top-down methods but could potentially be overwhelmed as the size and complexity of the desired assembly increases. === Neuroscience and psychology === These terms are also employed in cognitive sciences including neuroscience, cognitive neuroscience and cognitive psychology to discuss the flow of information in processing. Typically, sensory input is considered bottom-up, and higher cognitive processes, which have more information from other sources, are considered top-down. A bottom-up proc
Brian Deer Classification System
The Brian Deer Classification System (BDC) is a library classification system used to organize materials in libraries with specialized Indigenous collections. The system was created in the mid-1970s by Canadian librarian A. Brian Deer, a Kahnawake Mohawk. It has been adapted for use in a British Columbia version, and also by a small number of First Nations libraries in Canada. == History and usage == Deer designed his classification system while working in the library of the National Indian Brotherhood (now the Assembly of First Nations) from 1974 to 1976. Instead of using a standard library classification scheme, such as that of the Library of Congress, he created a new system to organize the library's historic indigenous research materials and papers. He later worked at the library of the Union of British Columbia Indian Chiefs, where he developed a system for its holdings. He returned to Kahnawake, working at its Cultural Centre at Kahnawake and the Kahnawake Branch branch of the Mohawk Nation Office. His system was flexible, and he created new forms for their collections. The new systems Deer created were designed specifically for the materials in each collection according to the concerns of local Indigenous people at the time (for example, categories included land claims, treaty rights, resource management, and Elders' stories). Between 1978 and 1980, the system was adapted for use in British Columbia by Gene Joseph and Keltie McCall while they were working at the Union of British Columbia Indian Chiefs, becoming known as BDC-BC. Joseph later adapted it further for use in the Xwi7xwa Library at University of British Columbia, Vancouver. Though the Brian Deer Classification was not created as a universal classification solution for Indigenous resources, the system has provided a foundation for specialized libraries to create their own localized classification schemes. Variations of the Brian Deer Classification System are used in a small number of Canadian libraries. One prominent library using BDC is the X̱wi7x̱wa Library at the University of British Columbia, which uses a British Columbia-focused version of BDC along with First Nations House of Learning subject headings. The Union of British Columbia Indian Chiefs Resource Centre issued a revised BDC-BC in 2014, with the goal of providing users with a more flexible and culturally appropriate approach to organizing their resources. The Aanischaaukamikw Cree Cultural Institute in Oujé-Bougoumou, Quebec, implemented a local adaptation of BDC when they opened in 2012. In 2020 the Carrier Sekani Tribal Council in Prince George, British Columbia, shifted from organizing its library with the Dewey Decimal Classification to using a version of the BDC. They added new subject heading categories for topics of local interest such as the crisis of Missing and murdered Indigenous women. Simon Fraser University Library began developing the Indigenous Curriculum Resource Centre (ICRC) in 2020, with the physical space opening in 2023. The ICRC is Call to Action 21 of SFU's Aboriginal Reconciliation Council's final report, Walk This Path With Us. Through its collection, the ICRC supports those interested in learning about how and why decolonizing pedagogy and teaching practices are important. The physical items in the collection are catalogued using a modified Brian Deer Classification system. In 2022 Kwantlen Polytechnic University’s χʷəχʷéy̓əm Indigenous Collection released a revised BDC-BC System. This BDC contains works exclusively with Indigenous authored materials and expands the cuttering systems of previous BDC, with the result that much of the collection reflects a spatial relationality. The implementation of this BDC was possible due to the tireless work at Xwi7xwa Library, Union of British Columbia Indian Chiefs Resource Centre, and Simon Fraser University Library's Indigenous Curriculum Resource Centre. == Structure == The high-level organizational structure of BDC reflects a First Nations worldview, with an emphasis on relationships between and among people, animals, and the land. Subcategories demonstrate the relationships among First Nations by grouping them geographically as opposed to alphabetically; the latter is a practice frequently used for specific topics in the Library of Congress Classification. The top-level hierarchy of the X̱wi7x̱wa Library adaptation of BDC-BC demonstrates the emphasis on access to subjects prioritized by a First Nation collection: Reference Materials Local History History International Education Economic Development Housing and Community Development Criminal Justice System Constitution (Canada) and First Nations Self Government Rights and Title Natural Resources Community Resources Health World View Fine Arts Languages Literature The system is not designed to provide a comprehensive description of all topics of interest to North American Indigenous peoples; in addition, its use is limited in scope, being intended for small and specialized libraries. While English is used in the classification scheme as a common language among First Nations peoples and non-Indigenous library users, Indigenous spellings and terminology that local library users would expect to find are used to provide access. Short and easily remembered call numbers are used to facilitate use by both library workers and patrons, with the recognition that Indigenous libraries often have a small staff and limited resources to devote to cataloging. Beyond its simplicity, one potential drawback of the system is its shortage of clear guidelines for application, which provides flexibility but can also result in inconsistencies within and between library catalogs. Because few libraries use the BDC and there are limited examples for use as case studies, implementing the system and keeping it up-to-date can prove a challenge for libraries with limited resources. However, X̱wi7x̱wa Library head librarian Ann Doyle describes the system as "an important part of the body of Indigenous scholarship" that should be retained as a reflection of Indigenous worldviews, as well as for ease of access for Indigenous library users.
Rendezvous hashing
Rendezvous or highest random weight (HRW) hashing is an algorithm that allows clients to achieve distributed agreement on a set of k {\displaystyle k} options out of a possible set of n {\displaystyle n} options. A typical application is when clients need to agree on which sites (or proxies) objects are assigned to. Consistent hashing addresses the special case k = 1 {\displaystyle k=1} using a different method. Rendezvous hashing is both much simpler and more general than consistent hashing (see below). == History == Rendezvous hashing was invented by David Thaler and Chinya Ravishankar at the University of Michigan in 1996. Consistent hashing appeared a year later in the literature. Given its simplicity and generality, rendezvous hashing is now being preferred to consistent hashing in real-world applications. Rendezvous hashing was used very early on in many applications including mobile caching, router design, secure key establishment, and sharding and distributed databases. Other examples of real-world systems that use Rendezvous Hashing include the GitHub load balancer, the Apache Ignite distributed database, the Tahoe-LAFS file store, the CoBlitz large-file distribution service, Apache Druid, IBM's Cloud Object Store, the Arvados Data Management System, Apache Kafka, and the Twitter EventBus pub/sub platform. One of the first applications of rendezvous hashing was to enable multicast clients on the Internet (in contexts such as the MBONE) to identify multicast rendezvous points in a distributed fashion. It was used in 1998 by Microsoft's Cache Array Routing Protocol (CARP) for distributed cache coordination and routing. Some Protocol Independent Multicast routing protocols use rendezvous hashing to pick a rendezvous point. == Problem definition and approach == === Algorithm === Rendezvous hashing solves a general version of the distributed hash table problem: We are given a set of n {\displaystyle n} sites (servers or proxies, say). How can any set of clients, given an object O {\displaystyle O} , agree on a k-subset of sites to assign to O {\displaystyle O} ? The standard version of the problem uses k = 1. Each client is to make its selection independently, but all clients must end up picking the same subset of sites. This is non-trivial if we add a minimal disruption constraint, and require that when a site fails or is removed, only objects mapping to that site need be reassigned to other sites. The basic idea is to give each site S j {\displaystyle S_{j}} a score (a weight) for each object O i {\displaystyle O_{i}} , and assign the object to the highest scoring site. All clients first agree on a hash function h ( ⋅ ) {\displaystyle h(\cdot )} . For object O i {\displaystyle O_{i}} , the site S j {\displaystyle S_{j}} is defined to have weight w i , j = h ( O i , S j ) {\displaystyle w_{i,j}=h(O_{i},S_{j})} . Each client independently computes these weights w i , 1 , w i , 2 … w i , n {\displaystyle w_{i,1},w_{i,2}\dots w_{i,n}} and picks the k sites that yield the k largest hash values. The clients have thereby achieved distributed k {\displaystyle k} -agreement. If a site S {\displaystyle S} is added or removed, only the objects mapping to S {\displaystyle S} are remapped to different sites, satisfying the minimal disruption constraint above. The HRW assignment can be computed independently by any client, since it depends only on the identifiers for the set of sites S 1 , S 2 … S n {\displaystyle S_{1},S_{2}\dots S_{n}} and the object being assigned. HRW easily accommodates different capacities among sites. If site S k {\displaystyle S_{k}} has twice the capacity of the other sites, we simply represent S k {\displaystyle S_{k}} twice in the list, say, as S k , 1 , S k , 2 {\displaystyle S_{k,1},S_{k,2}} . Clearly, twice as many objects will now map to S k {\displaystyle S_{k}} as to the other sites. === Properties === Consider the simple version of the problem, with k = 1, where all clients are to agree on a single site for an object O. Approaching the problem naively, it might appear sufficient to treat the n sites as buckets in a hash table and hash the object name O into this table. Unfortunately, if any of the sites fails or is unreachable, the hash table size changes, forcing all objects to be remapped. This massive disruption makes such direct hashing unworkable. Under rendezvous hashing, however, clients handle site failures by picking the site that yields the next largest weight. Remapping is required only for objects currently mapped to the failed site, and disruption is minimal. Rendezvous hashing has the following properties: Low overhead: The hash function used is efficient, so overhead at the clients is very low. Load balancing: Since the hash function is randomizing, each of the n sites is equally likely to receive the object O. Loads are uniform across the sites. Site capacity: Sites with different capacities can be represented in the site list with multiplicity in proportion to capacity. A site with twice the capacity of the other sites will be represented twice in the list, while every other site is represented once. High hit rate: Since all clients agree on placing an object O into the same site SO, each fetch or placement of O into SO yields the maximum utility in terms of hit rate. The object O will always be found unless it is evicted by some replacement algorithm at SO. Minimal disruption: When a site fails, only the objects mapped to that site need to be remapped. Disruption is at the minimal possible level. Distributed k-agreement: Clients can reach distributed agreement on k sites simply by selecting the top k sites in the ordering. == O(log n) running time via skeleton-based hierarchical rendezvous hashing == The standard version of Rendezvous Hashing described above works quite well for moderate n, but when n {\displaystyle n} is extremely large, the hierarchical use of Rendezvous Hashing achieves O ( log n ) {\displaystyle O(\log n)} running time. This approach creates a virtual hierarchical structure (called a "skeleton"), and achieves O ( log n ) {\displaystyle O(\log n)} running time by applying HRW at each level while descending the hierarchy. The idea is to first choose some constant m {\displaystyle m} and organize the n {\displaystyle n} sites into c = ⌈ n / m ⌉ {\displaystyle c=\lceil n/m\rceil } clusters C 1 = { S 1 , S 2 … S m } , C 2 = { S m + 1 , S m + 2 … S 2 m } … {\displaystyle C_{1}=\left\{S_{1},S_{2}\dots S_{m}\right\},C_{2}=\left\{S_{m+1},S_{m+2}\dots S_{2m}\right\}\dots } Next, build a virtual hierarchy by choosing a constant f {\displaystyle f} and imagining these c {\displaystyle c} clusters placed at the leaves of a tree T {\displaystyle T} of virtual nodes, each with fanout f {\displaystyle f} . In the accompanying diagram, the cluster size is m = 4 {\displaystyle m=4} , and the skeleton fanout is f = 3 {\displaystyle f=3} . Assuming 108 sites (real nodes) for convenience, we get a three-tier virtual hierarchy. Since f = 3 {\displaystyle f=3} , each virtual node has a natural numbering in octal. Thus, the 27 virtual nodes at the lowest tier would be numbered 000 , 001 , 002 , . . . , 221 , 222 {\displaystyle 000,001,002,...,221,222} in octal (we can, of course, vary the fanout at each level - in that case, each node will be identified with the corresponding mixed-radix number). The easiest way to understand the virtual hierarchy is by starting at the top, and descending the virtual hierarchy. We successively apply Rendezvous Hashing to the set of virtual nodes at each level of the hierarchy, and descend the branch defined by the winning virtual node. We can in fact start at any level in the virtual hierarchy. Starting lower in the hierarchy requires more hashes, but may improve load distribution in the case of failures. For example, instead of applying HRW to all 108 real nodes in the diagram, we can first apply HRW to the 27 lowest-tier virtual nodes, selecting one. We then apply HRW to the four real nodes in its cluster, and choose the winning site. We only need 27 + 4 = 31 {\displaystyle 27+4=31} hashes, rather than 108. If we apply this method starting one level higher in the hierarchy, we would need 9 + 3 + 4 = 16 {\displaystyle 9+3+4=16} hashes to get to the winning site. The figure shows how, if we proceed starting from the root of the skeleton, we may successively choose the virtual nodes ( 2 ) 3 {\displaystyle (2)_{3}} , ( 20 ) 3 {\displaystyle (20)_{3}} , and ( 200 ) 3 {\displaystyle (200)_{3}} , and finally end up with site 74. The virtual hierarchy need not be stored, but can be created on demand, since the virtual nodes names are simply prefixes of base- f {\displaystyle f} (or mixed-radix) representations. We can easily create appropriately sorted strings from the digits, as required. In the example, we would be working with the strings 0 , 1 , 2 {\displaystyle 0,1,2} (at tier 1), 20 , 21 , 22 {\displaystyle 20,21,22} (at tier 2), and 200 , 201 , 202
Confusion matrix
In machine learning, a confusion matrix, also known as error matrix, is a specific table layout that allows visualization of the performance of an algorithm, typically a supervised learning one. In unsupervised learning it is usually called a matching matrix. The term is used specifically in the problem of statistical classification. Each row of the matrix represents the instances in an actual class while each column represents the instances in a predicted class, or vice versa – both variants are found in the literature. The diagonal of the matrix therefore represents all instances that are correctly predicted. The name stems from the fact that it makes it easy to identify whether the system is confusing two classes (i.e., commonly mislabeling one class as another). The confusion matrix has its origins in human perceptual studies of auditory stimuli. It was adapted for machine learning studies and used by Frank Rosenblatt, among other early researchers, to compare human and machine classifications of visual (and later auditory) stimuli. It is a special kind of contingency table, with two dimensions ("actual" and "predicted"), and identical sets of "classes" in both dimensions (each combination of dimension and class is a variable in the contingency table). == Example == Given a sample of 12 individuals, 8 that have been diagnosed with cancer and 4 that are cancer-free, where individuals with cancer belong to class 1 (positive) and non-cancer individuals belong to class 0 (negative), we can display that data as follows: Assume that we have a classifier that distinguishes between individuals with and without cancer in some way, we can take the 12 individuals and run them through the classifier. The classifier then makes 9 accurate predictions and misses 3: 2 individuals with cancer wrongly predicted as being cancer-free (sample 1 and 2), and 1 person without cancer that is wrongly predicted to have cancer (sample 9). Notice, that if we compare the actual classification set to the predicted classification set, there are 4 different outcomes that could result in any particular column: The actual classification is positive and the predicted classification is positive (1,1). This is called a true positive result because the positive sample was correctly identified by the classifier. The actual classification is positive and the predicted classification is negative (1,0). This is called a false negative result because the positive sample is incorrectly identified by the classifier as being negative. The actual classification is negative and the predicted classification is positive (0,1). This is called a false positive result because the negative sample is incorrectly identified by the classifier as being positive. The actual classification is negative and the predicted classification is negative (0,0). This is called a true negative result because the negative sample gets correctly identified by the classifier. We can then perform the comparison between actual and predicted classifications and add this information to the table, making correct results appear in green so they are more easily identifiable. The template for any binary confusion matrix uses the four kinds of results discussed above (true positives, false negatives, false positives, and true negatives) along with the positive and negative classifications. The four outcomes can be formulated in a 2×2 confusion matrix, as follows: The color convention of the three data tables above were picked to match this confusion matrix, in order to easily differentiate the data. Now, we can simply total up each type of result, substitute into the template, and create a confusion matrix that will concisely summarize the results of testing the classifier: In this confusion matrix, of the 8 samples with cancer, the system judged that 2 were cancer-free, and of the 4 samples without cancer, it predicted that 1 did have cancer. All correct predictions are located in the diagonal of the table (highlighted in green), so it is easy to visually inspect the table for prediction errors, as values outside the diagonal will represent them. By summing up the 2 rows of the confusion matrix, one can also deduce the total number of positive (P) and negative (N) samples in the original dataset, i.e. P = T P + F N {\displaystyle P=TP+FN} and N = F P + T N {\displaystyle N=FP+TN} . == Table of confusion == In predictive analytics, a table of confusion (sometimes also called a confusion matrix) is a table with two rows and two columns that reports the number of true positives, false negatives, false positives, and true negatives. This allows more detailed analysis than simply observing the proportion of correct classifications (accuracy). Accuracy will yield misleading results if the data set is unbalanced; that is, when the numbers of observations in different classes vary greatly. For example, if there were 95 cancer samples and only 5 non-cancer samples in the data, a particular classifier might classify all the observations as having cancer. The overall accuracy would be 95%, but in more detail the classifier would have a 100% recognition rate (sensitivity) for the cancer class but a 0% recognition rate for the non-cancer class. F1 score is even more unreliable in such cases, and here would yield over 97.4%, whereas informedness removes such bias and yields 0 as the probability of an informed decision for any form of guessing (here always guessing cancer). According to Davide Chicco and Giuseppe Jurman, the most informative metric to evaluate a confusion matrix is the Matthews correlation coefficient (MCC). Other metrics can be included in a confusion matrix, each of them having their significance and use. Some researchers have argued that the confusion matrix, and the metrics derived from it, do not truly reflect a model's knowledge. In particular, the confusion matrix cannot show whether correct predictions were reached through sound reasoning or merely by chance (a problem known in philosophy as epistemic luck). It also does not capture situations where the facts used to make a prediction later change or turn out to be wrong (defeasibility). This means that while the confusion matrix is a useful tool for measuring classification performance, it may give an incomplete picture of a model’s true reliability. == Confusion matrices with more than two categories == Confusion matrix is not limited to binary classification and can be used in multi-class classifiers as well. The confusion matrices discussed above have only two conditions: positive and negative. For example, the table below summarizes communication of a whistled language between two speakers, with zero values omitted for clarity. == Confusion matrices in multi-label and soft-label classification == Confusion matrices are not limited to single-label classification (where only one class is present) or hard-label settings (where classes are either fully present, 1, or absent, 0). They can also be extended to Multi-label classification (where multiple classes can be predicted at once) and soft-label classification (where classes can be partially present). One such extension is the Transport-based Confusion Matrix (TCM), which builds on the theory of optimal transport and the principle of maximum entropy. TCM applies to single-label, multi-label, and soft-label settings. It retains the familiar structure of the standard confusion matrix: a square matrix sized by the number of classes, with diagonal entries indicating correct predictions and off-diagonal entries indicating confusion. In the single-label case, TCM is identical to the standard confusion matrix. TCM follows the same reasoning as the standard confusion matrix: if class A is overestimated (its predicted value is greater than its label value) and class B is underestimated (its predicted value is less than its label value), A is considered confused with B, and the entry (B, A) is increased. If a class is both predicted and present, it is correctly identified, and the diagonal entry (A, A) increases. Optimal transport and maximum entropy are used to determine the extent to which these entries are updated. TCM enables clearer comparison between predictions and labels in complex classification tasks, while maintaining a consistent matrix format across settings.
Controlled vocabulary
A controlled vocabulary provides a way to organize knowledge for subsequent retrieval. Controlled vocabularies are used in subject indexing schemes, subject headings, thesauri, taxonomies and other knowledge organization systems. Controlled vocabulary schemes mandate the use of predefined, preferred terms that have been preselected by the designers of the schemes, in contrast to natural language vocabularies, which have no such restriction. == In library and information science == In library and information science, controlled vocabulary is a carefully selected list of words and phrases, which are used to tag units of information (document or work) so that they may be more easily retrieved by a search. Controlled vocabularies solve the problems of homographs, synonyms and polysemes by a bijection between concepts and preferred terms. In short, controlled vocabularies reduce unwanted ambiguity inherent in normal human languages where the same concept can be given different names and ensure consistency. For example, in the Library of Congress Subject Headings (a subject heading system that uses a controlled vocabulary), preferred terms—subject headings in this case—have to be chosen to handle choices between variant spellings of the same word (American versus British), choice among scientific and popular terms (cockroach versus Periplaneta americana), and choices between synonyms (automobile versus car), among other difficult issues. Choices of preferred terms are based on the principles of user warrant (what terms users are likely to use), literary warrant (what terms are generally used in the literature and documents), and structural warrant (terms chosen by considering the structure, scope of the controlled vocabulary). Controlled vocabularies also typically handle the problem of homographs with qualifiers. For example, the term pool has to be qualified to refer to either swimming pool or the game pool to ensure that each preferred term or heading refers to only one concept. === Types used in libraries === There are two main kinds of controlled vocabulary tools used in libraries: subject headings and thesauri. While the differences between the two are diminishing, there are still some minor differences: Historically, subject headings were designed to describe books in library catalogs by catalogers while thesauri were used by indexers to apply index terms to documents and articles. Subject headings tend to be broader in scope describing whole books, while thesauri tend to be more specialized covering very specific disciplines. Because of the card catalog system, subject headings tend to have terms that are in indirect order (though with the rise of automated systems this is being removed), while thesaurus terms are always in direct order. Subject headings tend to use more pre-coordination of terms such that the designer of the controlled vocabulary will combine various concepts together to form one preferred subject heading. (e.g., children and terrorism) while thesauri tend to use singular direct terms. Thesauri list not only equivalent terms but also narrower, broader terms and related terms among various preferred and non-preferred (but potentially synonymous) terms, while historically most subject headings did not. For example, the Library of Congress Subject Heading itself did not have much syndetic structure until 1943, and it was not until 1985 when it began to adopt the thesauri type term "Broader term" and "Narrow term". The terms are chosen and organized by trained professionals (including librarians and information scientists) who possess expertise in the subject area. Controlled vocabulary terms can accurately describe what a given document is actually about, even if the terms themselves do not occur within the document's text. Well known subject heading systems include the Library of Congress system, Medical Subject Headings (MeSH) created by the United States National Library of Medicine, and Sears. Well known thesauri include the Art and Architecture Thesaurus and the ERIC Thesaurus. When selecting terms for a controlled vocabulary, the designer has to consider the specificity of the term chosen, whether to use direct entry, inter consistency and stability of the language. Lastly the amount of pre-coordination (in which case the degree of enumeration versus synthesis becomes an issue) and post-coordination in the system is another important issue. Controlled vocabulary elements (terms/phrases) employed as tags, to aid in the content identification process of documents, or other information system entities (e.g. DBMS, Web Services) qualifies as metadata. == Indexing languages == There are three main types of indexing languages. Controlled indexing language – only approved terms can be used by the indexer to describe the document Natural language indexing language – any term from the document in question can be used to describe the document Free indexing language – any term (not only from the document) can be used to describe the document When indexing a document, the indexer also has to choose the level of indexing exhaustivity, the level of detail in which the document is described. For example, using low indexing exhaustivity, minor aspects of the work will not be described with index terms. In general the higher the indexing exhaustivity, the more terms indexed for each document. In recent years free text search as a means of access to documents has become popular. This involves using natural language indexing with an indexing exhaustively set to maximum (every word in the text is indexed). These methods have been compared in some studies, such as the 2007 article, "A Comparative Evaluation of Full-text, Concept-based, and Context-sensitive Search". === Advantages === Controlled vocabularies are often claimed to improve the accuracy of free text searching, such as to reduce irrelevant items in the retrieval list. These irrelevant items (false positives) are often caused by the inherent ambiguity of natural language. Take the English word football for example. Football is the name given to a number of different team sports. Worldwide the most popular of these team sports is association football, which also happens to be called soccer in several countries. The word football is also applied to rugby football (rugby union and rugby league), American football, Australian rules football, Gaelic football, and Canadian football. A search for football therefore will retrieve documents that are about several completely different sports. Controlled vocabulary solves this problem by tagging the documents in such a way that the ambiguities are eliminated. Compared to free text searching, the use of a controlled vocabulary can dramatically increase the performance of an information retrieval system, if performance is measured by precision (the percentage of documents in the retrieval list that are actually relevant to the search topic). In some cases controlled vocabulary can enhance recall as well, because unlike natural language schemes, once the correct preferred term is searched, there is no need to search for other terms that might be synonyms of that term. === Disadvantages === A controlled vocabulary search may lead to unsatisfactory recall, in that it will fail to retrieve some documents that are actually relevant to the search question. This is particularly problematic when the search question involves terms that are sufficiently tangential to the subject area such that the indexer might have decided to tag it using a different term (but the searcher might consider the same). Essentially, this can be avoided only by an experienced user of controlled vocabulary whose understanding of the vocabulary coincides with that of the indexer. Another possibility is that the article is just not tagged by the indexer because indexing exhaustivity is low. For example, an article might mention football as a secondary focus, and the indexer might decide not to tag it with "football" because it is not important enough compared to the main focus. But it turns out that for the searcher that article is relevant and hence recall fails. A free text search would automatically pick up that article regardless. On the other hand, free text searches have high exhaustivity (every word is searched) so although it has much lower precision, it has potential for high recall as long as the searcher overcome the problem of synonyms by entering every combination. Controlled vocabularies may become outdated rapidly in fast developing fields of knowledge, unless the preferred terms are updated regularly. Even in an ideal scenario, a controlled vocabulary is often less specific than the words of the text itself. Indexers trying to choose the appropriate index terms might misinterpret the author, while this precise problem is not a factor in a free text, as it uses the author's own words. The use of controlled vocabularies can be costly compared to free