Digital redlining is the practice of creating and perpetuating inequities between already marginalized groups specifically through the use of digital technologies, digital content, and the internet. The concept of digital redlining is an extension of the practice of redlining in housing discrimination, a historical legal practice in the United States and Canada dating back to the 1930s where red lines were drawn on maps to indicate poor and primarily black neighborhoods that were deemed unsuitable for loans or further development, which created great economic disparities between neighborhoods. The term was popularized by Dr. Chris Gilliard, a privacy scholar, who defines digital redlining as "the creation and maintenance of tech practices, policies, pedagogies, and investment decisions that enforce class boundaries and discriminate against specific groups". Though digital redlining is related to the digital divide and techniques such as weblining and personalization, it is distinct from these concepts as part of larger complex systemic issues. It can refer to practices that create inequities of access to technology services in geographical areas, such as when internet service providers decide to not service specific geographic areas because they are perceived to be not as profitable and thus reduce access to crucial services and civic participation. It can also be used to refer to inequities caused by the policies and practices of digital technologies. For instance, with these methods inequities are accomplished through divisions that are created via algorithms which are hidden from the technology user; the use of big data and analytics allow for a much more nuanced form of discrimination that can target specific vulnerable populations. These algorithmic means are enabled through the use of unregulated data technologies that apply a score to individuals that statistically categorize personality traits or tendencies which are similar to a credit score but are proprietary to the technology companies and not under outside oversight. == Digital redlining and geography == While the roots of redlining lie in excluding populations based on geography, digital redlining occurs in both geographical and non-geographical contexts. An example of both contexts can be found in the charges brought against Facebook on March 28 of 2019, by the United States Department of Housing and Urban Development (HUD). HUD charged Facebook with violating the Fair Housing Act of 1968 by "encouraging, enabling, and causing housing discrimination through the company's advertising platform." HUD stated that Facebook allowed advertisers to “exclude people who live in a specified area from seeing an ad by drawing a red line around that area.” The discrimination called out by HUD included those that were racist, homophobic, ableist, and classist. Besides this example of geographically based digital redlining, HUD also charged that Facebook used profile information and designations to exclude classes of people. The charges stated: "Facebook enabled advertisers to exclude people whom Facebook classified as parents; non-American-born; non-Christian; interested in accessibility; interested in Hispanic culture; or a wide variety of other interests that closely align with the Fair Housing Act’s protected classes" Several media outlets pointed out HUDs own history of housing discrimination through redlining, the establishment of the Fair Housing Act to combat redlining, and how the digital platform was recreating this discriminatory practice. === Digital redlining within a geographical context === Although digital redlining refers to a complex and varied set of practices, it has been most commonly applied to practices with a geographical dimension. Common examples include when an internet service providers decide to not service specific geographic areas because those areas are seen to be not as profitable, resulting in discrimination against low-income communities, with resulting impacts on access to crucial services and civic participation. AT&T has faced specific scrutiny for this form of digital redlining, it has been reported that AT&T has been classist in its offerings of broadband internet service in areas that are more impoverished. Geographically based digital redlining can also apply to digital content or the distribution of goods sold online. Geographically based games such as Pokémon Go have been shown to offer more virtual stops and rewards in geographic areas that are less ethnically and racially diverse. In 2016, Amazon was rebuked for not offering their Prime same-day delivery service to many communities that were largely African American and had incomes that were beneath the national average. Even services such as email can be impacted, with many email administrators creating filters for flagging particular email messages as spam based on the geographical origin of the message. === Digital redlining based on personal identity === Although often aligned with discrimination that falls into a geographically based context digital redlining also refers to when vulnerable populations are targeted for or excluded from specific content or access to the internet in a way that harms them based on some aspect of their identity. Trade schools and community colleges, which typically have a more working class student body, have been found to block public internet content from their students where elite research institutions do not. The use of big data and analytics allow for a much more nuanced form of discrimination that can target specific vulnerable populations. For example, Facebook has been criticized for providing tools that allow advertisers to target ads by ethnic affinity and gender, effectively blocking minorities from seeing specific ads for housing and employment. In October 2019, a major class action lawsuit was filed against Facebook alleging gender and age discrimination in financial advertising. A broad array of consumers can be particularly vulnerable to digital redlining when it is used outside of a geographical context. Besides targeting vulnerable populations based on traditional and legally recognized classifications such as race, gender, age, etc., it has been shown that personal data mined and then resold by brokers can be used to target those who have been identified as suffering from Alzheimer's or dementia, or simply identified as impulse buyers or gullible. == Term distinctions == === Distinctions between weblining and digital redlining === Earlier distinctions have been made between weblining—the process of charging customers different prices based on profile information --- and internet or digital redlining, with digital redlining being focused not on pricing but access. As early as 2002 the Gale Encyclopedia of E-Commerce puts forth the distinction more in use today: weblining is the pervasive and generally accepted (or at least tolerated) practice of personalizing access to products and services in ways invisible to the user; digital redlining is when such personalized, data-driven schemes perpetuate traditional advantages of privileged demographics. As weblining has become more ubiquitous, the term has fallen out of use in favor of the more general term personalization. === Distinctions between the digital divide and digital redlining === Scholars have often drawn connections between the digital divide and digital redlining. In practice, the digital divide is seen as one of a number of impacts of digital redlining, and digital redlining is one of a number of ways in which the divide is maintained or extended. == Criticisms == A 2001 report looked to find if the reason for a gap in access to broadband internet by low-income and minority populations was due to a lack of availability or due to other factors. The report found that there was "little evidence of digital redlining based on income or black or Hispanic concentrations" but that there was mixed evidence of redlining based on areas in which Native American or Asian populations were larger.
Natural language processing
Natural language processing (NLP) is the processing of natural language information by a computer. NLP is a subfield of computer science and is closely associated with artificial intelligence. NLP is also related to information retrieval, knowledge representation, computational linguistics, and linguistics more broadly. Major processing tasks in an NLP system include: speech recognition, text classification, natural language understanding, and natural language generation. == History == Natural language processing has its roots in the 1950s. Already in 1950, Alan Turing published an article titled "Computing Machinery and Intelligence," which proposed what is now called the Turing test as a criterion of intelligence, though at the time that was not articulated as a problem separate from artificial intelligence. The proposed test includes a task that involves the automated interpretation and generation of natural language. === Symbolic NLP (1950s – early 1990s) === The premise of symbolic NLP is often illustrated using John Searle's Chinese room thought experiment: Given a collection of rules (e.g., a Chinese phrasebook, with questions and matching answers), the computer emulates natural language understanding (or other NLP tasks) by applying those rules to the data it confronts. 1950s: The Georgetown experiment in 1954 involved fully automatic translation of more than sixty Russian sentences into English. The authors claimed that within three or five years, machine translation would be a solved problem. However, real progress was much slower, and after the ALPAC report in 1966, which found that ten years of research had failed to fulfill the expectations, funding for machine translation was dramatically reduced. Little further research in machine translation was conducted in America (though some research continued elsewhere, such as Japan and Europe) until the late 1980s when the first statistical machine translation systems were developed. 1960s: Some notably successful natural language processing systems developed in the 1960s were SHRDLU, a natural language system working in restricted "blocks worlds" with restricted vocabularies, and ELIZA, a simulation of Rogerian psychotherapy, written by Joseph Weizenbaum between 1964 and 1966. Despite using minimal information about human thought or emotion, ELIZA was able to produce interactions that appeared human-like. When the "patient" exceeded the very small knowledge base, ELIZA might provide a generic response, for example, responding to "My head hurts" with "Why do you say your head hurts?". Ross Quillian's successful work on natural language was demonstrated with a vocabulary of only twenty words, because that was all that would fit in a computer memory at the time. 1970s: During the 1970s, many programmers began to write "conceptual ontologies", which structured real-world information into computer-understandable data. Examples are MARGIE (Schank, 1975), SAM (Cullingford, 1978), PAM (Wilensky, 1978), TaleSpin (Meehan, 1976), QUALM (Lehnert, 1977), Politics (Carbonell, 1979), and Plot Units (Lehnert 1981). During this time, the first chatterbots were written (e.g., PARRY). 1980s: The 1980s and early 1990s mark the heyday of symbolic methods in NLP. Focus areas of the time included research on rule-based parsing (e.g., the development of HPSG as a computational operationalization of generative grammar), morphology (e.g., two-level morphology), semantics (e.g., Lesk algorithm), reference (e.g., within Centering Theory) and other areas of natural language understanding (e.g., in the Rhetorical Structure Theory). Other lines of research were continued, e.g., the development of chatterbots with Racter and Jabberwacky. An important development (that eventually led to the statistical turn in the 1990s) was the rising importance of quantitative evaluation in this period. === Statistical NLP (1990s–present) === Up until the 1980s, most natural language processing systems were based on complex sets of hand-written rules. Starting in the late 1980s, however, there was a revolution in natural language processing with the introduction of machine learning algorithms for language processing. This shift was influenced by increasing computational power (see Moore's law) and a decline in the dominance of Chomskyan linguistic theories (e.g. transformational grammar), whose theoretical underpinnings discouraged the sort of corpus linguistics that underlies the machine-learning approach to language processing. 1990s: Many of the notable early successes in statistical methods in NLP occurred in the field of machine translation, due especially to work at IBM Research, such as IBM alignment models. These systems were able to take advantage of existing multilingual textual corpora that had been produced by the Parliament of Canada and the European Union as a result of laws calling for the translation of all governmental proceedings into all official languages of the corresponding systems of government. However, many systems relied on corpora that were specifically developed for the tasks they were designed to perform. This reliance has been a major limitation to their broader effectiveness and continues to affect similar systems. Consequently, significant research has focused on methods for learning effectively from limited amounts of data. 2000s: With the growth of the web, increasing amounts of raw (unannotated) language data have become available since the mid-1990s. Research has thus increasingly focused on unsupervised and semi-supervised learning algorithms. Such algorithms can learn from data that has not been hand-annotated with the desired answers or using a combination of annotated and non-annotated data. Generally, this task is much more difficult than supervised learning, and typically produces less accurate results for a given amount of input data. However, large quantities of non-annotated data are available (including, among other things, the entire content of the World Wide Web), which can often make up for the worse efficiency if the algorithm used has a low enough time complexity to be practical. 2003: word n-gram model, at the time the best statistical algorithm, is outperformed by a multi-layer perceptron (with a single hidden layer and context length of several words, trained on up to 14 million words, by Bengio et al.) 2010: Tomáš Mikolov (then a PhD student at Brno University of Technology) with co-authors applied a simple recurrent neural network with a single hidden layer to language modeling, and in the following years he went on to develop Word2vec. In the 2010s, representation learning and deep neural network-style (featuring many hidden layers) machine learning methods became widespread in natural language processing. This shift gained momentum due to results showing that such techniques can achieve state-of-the-art results in many natural language tasks, e.g., in language modeling and parsing. This is increasingly important in medicine and healthcare, where NLP helps analyze notes and text in electronic health records that would otherwise be inaccessible for study when seeking to improve care or protect patient privacy. == Approaches: Symbolic, statistical, neural networks == Symbolic approach, i.e., the hand-coding of a set of rules for manipulating symbols, coupled with a dictionary lookup, was historically the first approach used both by AI in general and by NLP in particular: such as by writing grammars or devising heuristic rules for stemming. Machine learning approaches, which include both statistical and neural networks, on the other hand, have many advantages over the symbolic approach: both statistical and neural network methods tend to focus more on the most common cases extracted from a corpus of texts, whereas the rule-based approach needs to provide rules for both rare and common cases equally. language models, produced by either statistical or neural networks methods, are more robust to both unfamiliar (e.g. containing words or structures that have not been seen before) and erroneous input (e.g. with misspelled words or words accidentally omitted) in comparison to the rule-based systems, which are also more costly to produce. the larger such a (probabilistic) language model is, the more accurate it becomes, in contrast to rule-based systems that can gain accuracy only by increasing the amount and complexity of the rules leading to intractability problems. Rule-based systems are commonly used: when the amount of training data is insufficient to successfully apply machine learning methods, e.g., for the machine translation of low-resource languages such as provided by the Apertium system, for preprocessing in NLP pipelines, e.g., tokenization, or for post-processing and transforming the output of NLP pipelines, e.g., for knowledge extraction from syntactic parses. === Statistical approach === In the late 1980s and mid-1990s, the statistical approach ended a peri
Divide-and-conquer algorithm
In computer science, divide and conquer is an algorithm design paradigm. A divide-and-conquer algorithm recursively breaks down a problem into two or more sub-problems of the same or related type, until these become simple enough to be solved directly. The solutions to the sub-problems are then combined to give a solution to the original problem. The divide-and-conquer technique is the basis of efficient algorithms for many problems, such as sorting (e.g., quicksort, merge sort), multiplying large numbers (e.g., the Karatsuba algorithm), finding the closest pair of points, syntactic analysis (e.g., top-down parsers), SAT solving, and computing the discrete Fourier transform (FFT). Designing efficient divide-and-conquer algorithms can be difficult. As in mathematical induction, it is often necessary to generalize the problem to make it amenable to a recursive solution. The correctness of a divide-and-conquer algorithm is usually proved by mathematical induction, and its computational cost is often determined by solving recurrence relations. == Divide and conquer == The divide-and-conquer paradigm is often used to find an optimal solution of a problem. Its basic idea is to decompose a given problem into two or more similar, but simpler, subproblems, to solve them in turn, and to compose their solutions to solve the given problem. Problems of sufficient simplicity are solved directly. For example, to sort a given list of n natural numbers, split it into two lists of about n/2 numbers each, sort each of them in turn, and interleave both results appropriately to obtain the sorted version of the given list (see the picture). This approach is known as the merge sort algorithm. The name "divide and conquer" is sometimes applied to algorithms that reduce each problem to only one sub-problem, such as the binary search algorithm for finding a record in a sorted list (or its analogue in numerical computing, the bisection algorithm for root finding). These algorithms can be implemented more efficiently than general divide-and-conquer algorithms; in particular, if they use tail recursion, they can be converted into simple loops. Under this broad definition, however, every algorithm that uses recursion or loops could be regarded as a "divide-and-conquer algorithm". Therefore, some authors consider that the name "divide and conquer" should be used only when each problem may generate two or more subproblems. The name decrease and conquer has been proposed instead for the single-subproblem class. An important application of divide and conquer is in optimization, where if the search space is reduced ("pruned") by a constant factor at each step, the overall algorithm has the same asymptotic complexity as the pruning step, with the constant depending on the pruning factor (by summing the geometric series); this is known as prune and search. == Early historical examples == Early examples of these algorithms are primarily decrease and conquer – the original problem is successively broken down into single subproblems, and indeed can be solved iteratively. Binary search, a decrease-and-conquer algorithm where the subproblems are of roughly half the original size, has a long history. While a clear description of the algorithm on computers appeared in 1946 in an article by John Mauchly, the idea of using a sorted list of items to facilitate searching dates back at least as far as Babylonia in 200 BC. Another ancient decrease-and-conquer algorithm is the Euclidean algorithm to compute the greatest common divisor of two numbers by reducing the numbers to smaller and smaller equivalent subproblems, which dates to several centuries BC. An early example of a divide-and-conquer algorithm with multiple subproblems is Gauss's 1805 description of what is now called the Cooley–Tukey fast Fourier transform (FFT) algorithm, although he did not analyze its operation count quantitatively, and FFTs did not become widespread until they were rediscovered over a century later. An early two-subproblem D&C algorithm that was specifically developed for computers and properly analyzed is the merge sort algorithm, invented by John von Neumann in 1945. Another notable example is the algorithm invented by Anatolii A. Karatsuba in 1960 that could multiply two n-digit numbers in O ( n log 2 3 ) {\displaystyle O(n^{\log _{2}3})} operations (in Big O notation). This algorithm disproved Andrey Kolmogorov's 1956 conjecture that Ω ( n 2 ) {\displaystyle \Omega (n^{2})} operations would be required for that task. As another example of a divide-and-conquer algorithm that did not originally involve computers, Donald Knuth gives the method a post office typically uses to route mail: letters are sorted into separate bags for different geographical areas, each of these bags is itself sorted into batches for smaller sub-regions, and so on until they are delivered. This is related to a radix sort, described for punch-card sorting machines as early as 1929. == Advantages == === Solving difficult problems === Divide and conquer is a powerful tool for solving conceptually difficult problems: all it requires is a way of breaking the problem into sub-problems, of solving the trivial cases, and of combining sub-problems to the original problem. Similarly, decrease and conquer only requires reducing the problem to a single smaller problem, such as the classic Tower of Hanoi puzzle, which reduces moving a tower of height n {\displaystyle n} to move a tower of height n − 1 {\displaystyle n-1} . === Algorithm efficiency === The divide-and-conquer paradigm often helps in the discovery of efficient algorithms. It was the key, for example, to Karatsuba's fast multiplication method, the quicksort and mergesort algorithms, the Strassen algorithm for matrix multiplication, and fast Fourier transforms. In all these examples, the D&C approach led to an improvement in the asymptotic cost of the solution. For example, if (a) the base cases have constant-bounded size, the work of splitting the problem and combining the partial solutions is proportional to the problem's size n {\displaystyle n} , and (b) there is a bounded number p {\displaystyle p} of sub-problems of size ~ n p {\displaystyle {\frac {n}{p}}} at each stage, then the cost of the divide-and-conquer algorithm will be O ( n log p n ) {\displaystyle O(n\log _{p}n)} . For other types of divide-and-conquer approaches, running times can also be generalized. For example, when a) the work of splitting the problem and combining the partial solutions take c n {\displaystyle cn} time, where n {\displaystyle n} is the input size and c {\displaystyle c} is some constant; b) when n < 2 {\displaystyle n<2} , the algorithm takes time upper-bounded by c {\displaystyle c} , and c) there are q {\displaystyle q} subproblems where each subproblem has size ~ n 2 {\displaystyle {\frac {n}{2}}} . Then, the running times are as follows: if the number of subproblems q > 2 {\displaystyle q>2} , then the divide-and-conquer algorithm's running time is bounded by O ( n log 2 q ) {\displaystyle O(n^{\log _{2}q})} . if the number of subproblems is exactly one, then the divide-and-conquer algorithm's running time is bounded by O ( n ) {\displaystyle O(n)} . If, instead, the work of splitting the problem and combining the partial solutions take c n 2 {\displaystyle cn^{2}} time, and there are 2 subproblems where each has size n 2 {\displaystyle {\frac {n}{2}}} , then the running time of the divide-and-conquer algorithm is bounded by O ( n 2 ) {\displaystyle O(n^{2})} . === Parallelism === Divide-and-conquer algorithms are naturally adapted for execution in multi-processor machines, especially shared-memory systems where the communication of data between processors does not need to be planned in advance because distinct sub-problems can be executed on different processors. === Memory access === Divide-and-conquer algorithms naturally tend to make efficient use of memory caches. The reason is that once a sub-problem is small enough, it and all its sub-problems can, in principle, be solved within the cache, without accessing the slower main memory. An algorithm designed to exploit the cache in this way is called cache-oblivious, because it does not contain the cache size as an explicit parameter. Moreover, D&C algorithms can be designed for important algorithms (e.g., sorting, FFTs, and matrix multiplication) to be optimal cache-oblivious algorithms–they use the cache in a probably optimal way, in an asymptotic sense, regardless of the cache size. In contrast, the traditional approach to exploiting the cache is blocking, as in loop nest optimization, where the problem is explicitly divided into chunks of the appropriate size—this can also use the cache optimally, but only when the algorithm is tuned for the specific cache sizes of a particular machine. The same advantage exists with regards to other hierarchical storage systems, such as NUMA or virtual memory, as well as for multip
Data quality
Data quality refers to the state of qualitative or quantitative pieces of information. There are many definitions of data quality, but data is generally considered high quality if it is "fit for [its] intended uses in operations, decision making and planning". Data is deemed of high quality if it correctly represents the real-world construct to which it refers. Apart from these definitions, as the number of data sources increases, the question of internal data consistency becomes significant, regardless of fitness for use for any particular external purpose. People's views on data quality can often be in disagreement, even when discussing the same set of data used for the same purpose. When this is the case, businesses may adopt recognised international standards for data quality (See #International Standards for Data Quality below). Data governance can also be used to form agreed upon definitions and standards, including international standards, for data quality. In such cases, data cleansing, including standardization, may be required in order to ensure data quality. == Definitions == Defining data quality is difficult due to the many contexts data are used in, as well as the varying perspectives among end users, producers, and custodians of data. From a consumer perspective, data quality is: "data that are fit for use by data consumers" data "meeting or exceeding consumer expectations" data that "satisfies the requirements of its intended use" From a business perspective, data quality is: data that are "'fit for use' in their intended operational, decision-making and other roles" or that exhibits "'conformance to standards' that have been set, so that fitness for use is achieved" data that "are fit for their intended uses in operations, decision making and planning" "the capability of data to satisfy the stated business, system, and technical requirements of an enterprise" From a standards-based perspective, data quality is: the "degree to which a set of inherent characteristics (quality dimensions) of an object (data) fulfills requirements" "the usefulness, accuracy, and correctness of data for its application" Arguably, in all these cases, "data quality" is a comparison of the actual state of a particular set of data to a desired state, with the desired state being typically referred to as "fit for use," "to specification," "meeting consumer expectations," "free of defect," or "meeting requirements." These expectations, specifications, and requirements are usually defined by one or more individuals or groups, standards organizations, laws and regulations, business policies, or software development policies. == Dimensions of data quality == Drilling down further, those expectations, specifications, and requirements are stated in terms of characteristics or dimensions of the data, such as: accessibility or availability accuracy or correctness comparability completeness or comprehensiveness consistency, coherence, or clarity credibility, reliability, or reputation flexibility plausibility relevance, pertinence, or usefulness timeliness or latency uniqueness validity or reasonableness A systematic scoping review of the literature suggests that data quality dimensions and methods with real world data are not consistent in the literature, and as a result quality assessments are challenging due to the complex and heterogeneous nature of these data. == International standards for data quality == ISO 8000 is an international standard for data quality. Managed by the International Organization for Standardization, the ISO 8000 standards address and describe general aspects of data quality including principles, vocabulary and measurement data governance data quality management data quality assessment quality of master data, including exchange of characteristic data and identifiers quality of industrial data == History == Before the rise of the inexpensive computer data storage, massive mainframe computers were used to maintain name and address data for delivery services. This was so that mail could be properly routed to its destination. The mainframes used business rules to correct common misspellings and typographical errors in name and address data, as well as to track customers who had moved, died, gone to prison, married, divorced, or experienced other life-changing events. Government agencies began to make postal data available to a few service companies to cross-reference customer data with the National Change of Address registry (NCOA). This technology saved large companies millions of dollars in comparison to manual correction of customer data. Large companies saved on postage, as bills and direct marketing materials made their way to the intended customer more accurately. Initially sold as a service, data quality moved inside the walls of corporations, as low-cost and powerful server technology became available. Companies with an emphasis on marketing often focused their quality efforts on name and address information, but data quality is recognized as an important property of all types of data. Principles of data quality can be applied to supply chain data, transactional data, and nearly every other category of data found. For example, making supply chain data conform to a certain standard has value to an organization by: 1) avoiding overstocking of similar but slightly different stock; 2) avoiding false stock-out; 3) improving the understanding of vendor purchases to negotiate volume discounts; and 4) avoiding logistics costs in stocking and shipping parts across a large organization. For companies with significant research efforts, data quality can include developing protocols for research methods, reducing measurement error, bounds checking of data, cross tabulation, modeling and outlier detection, verifying data integrity, etc. == Overview == There are a number of theoretical frameworks for understanding data quality. A systems-theoretical approach influenced by American pragmatism expands the definition of data quality to include information quality, and emphasizes the inclusiveness of the fundamental dimensions of accuracy and precision on the basis of the theory of science (Ivanov, 1972). One framework, dubbed "Zero Defect Data" (Hansen, 1991) adapts the principles of statistical process control to data quality. Another framework seeks to integrate the product perspective (conformance to specifications) and the service perspective (meeting consumers' expectations) (Kahn et al. 2002). Another framework is based in semiotics to evaluate the quality of the form, meaning and use of the data (Price and Shanks, 2004). One highly theoretical approach analyzes the ontological nature of information systems to define data quality rigorously (Wand and Wang, 1996). A considerable amount of data quality research involves investigating and describing various categories of desirable attributes (or dimensions) of data. Nearly 200 such terms have been identified and there is little agreement in their nature (are these concepts, goals or criteria?), their definitions or measures (Wang et al., 1993). Software engineers may recognize this as a similar problem to "ilities". MIT has an Information Quality (MITIQ) Program, led by Professor Richard Wang, which produces a large number of publications and hosts a significant international conference in this field (International Conference on Information Quality, ICIQ). This program grew out of the work done by Hansen on the "Zero Defect Data" framework (Hansen, 1991). In practice, data quality is a concern for professionals involved with a wide range of information systems, ranging from data warehousing and business intelligence to customer relationship management and supply chain management. One industry study estimated the total cost to the U.S. economy of data quality problems at over U.S. $600 billion per annum (Eckerson, 2002). Incorrect data – which includes invalid and outdated information – can originate from different data sources – through data entry, or data migration and conversion projects. In 2002, the USPS and PricewaterhouseCoopers released a report stating that 23.6 percent of all U.S. mail sent is incorrectly addressed. One reason contact data becomes stale very quickly in the average database – more than 45 million Americans change their address every year. In fact, the problem is such a concern that companies are beginning to set up a data governance team whose sole role in the corporation is to be responsible for data quality. In some organizations, this data governance function has been established as part of a larger Regulatory Compliance function - a recognition of the importance of Data/Information Quality to organizations. Problems with data quality don't only arise from incorrect data; inconsistent data is a problem as well. Eliminating data shadow systems and centralizing data in a warehouse is one of the initiatives a company can take to ensure data consistency. En
Paper data storage
Paper data storage refers to the use of paper as a data storage device. This includes writing, illustrating, and the use of data that can be interpreted by a machine or is the result of the functioning of a machine. A defining feature of paper data storage is the ability of humans to produce it with only simple tools and interpret it visually. Though now mostly obsolete, paper was once an important form of computer data storage as both paper tape and punch cards were a common staple of working with computers before the 1980s. == History == Before paper was used for storing data, it had been used in several applications for storing instructions to specify a machine's operation. The earliest use of paper to store instructions for a machine was the work of Basile Bouchon who, in 1725, used punched paper rolls to control textile looms. This technology was later developed into the wildly successful Jacquard loom. The 19th century saw several other uses of paper for controlling machines. In 1846, telegrams could be prerecorded on punched tape and rapidly transmitted using Alexander Bain's automatic telegraph. Several inventors took the concept of a mechanical organ and used paper to represent the music. In the late 1880s Herman Hollerith invented the recording of data on a medium that could then be read by a machine. Prior uses of machine readable media, above, had been for control (automatons, piano rolls, looms, ...), not data. "After some initial trials with paper tape, he settled on punched cards..." Hollerith's method was used in the 1890 census. Hollerith's company eventually became the core of IBM. Other technologies were also developed that allowed machines to work with marks on paper instead of punched holes. This technology was widely used for tabulating votes and grading standardized tests. Banks used magnetic ink on checks, supporting MICR scanning. In an early electronic computing device, the Atanasoff–Berry Computer, electric sparks were used to singe small holes in paper cards to represent binary data. The altered dielectric constant of the paper at the location of the holes could then be used to read the binary data back into the machine by means of electric sparks of lower voltage than the sparks used to create the holes. This form of paper data storage was never made reliable and was not used in any subsequent machine. == Modern techniques == === 1D barcodes === Barcodes make it possible for any object that was to be sold or transported to have some computer readable information securely attached to it. Universal Product Code barcodes, first used in 1974, are ubiquitous today. Some people recommend a width of at least 3 pixels for each minimum-width gap and each minimum-width bar for 1D barcodes. The density is about 50 bits per linear inch (about 2 bit/mm). === 2D barcodes === 2D barcodes allow to store much more data on paper, up to 2.9 kbyte per barcode. It is recommended to have a width of at least 4 pixels—e.g., a 4 × 4 pixel = 16 pixel module. == Limits == The limits of data storage depend on the technology to write and read such data. The theoretical limits assume a scanner that can perfectly reproduce the printed image at its printing resolution, and a program which can accurately interpret such an image. For example, an 8 in × 10 in (200 mm × 250 mm) 600 dpi black-and-white image contains 3.43 MiB of data, as does a 300 dpi CMYK printed image. A 2,400 ppi True color (24-bit) image contains about 1.29 GiB of information; printing an image maintaining this data would require a printing resolution of about 120,000 dpi in black and white, or 60,000 dpi with CMYK dots.
PenTile matrix family
PenTile matrix is a family of patented subpixel matrix schemes used in electronic device displays. PenTile is a trademark of Samsung. PenTile matrices are used in AMOLED and LCD displays. These subpixel layouts are specifically designed to operate with proprietary algorithms for subpixel rendering embedded in the display driver, allowing plug and play compatibility with conventional RGB (Red-Green-Blue) stripe panels. == Overview == "PenTile Matrix" (a neologism from penta-, meaning "five" in Greek and tile) describes the geometric layout of the prototypical subpixel arrangement developed in the early 1990s. The layout consists of a quincunx comprising two red subpixels, two green subpixels, and one central blue subpixel in each unit cell. It was inspired by biomimicry of the human retina, which has nearly equal numbers of L and M type cone cells, but significantly fewer S cones. As the S cones are primarily responsible for perceiving blue colors, which do not appreciably affect the perception of luminance, reducing the number of blue subpixels with respect to the red and green subpixels in a display does not reduce the image quality. However, the layout may cause color leakage image distortion, which can be reduced by filters. In some cases the layout causes reduced moiré and blockiness compared to conventional RGB layouts. The PenTile layout is specifically designed to work with and be dependent upon subpixel rendering that uses only one and a quarter subpixel per pixel, on average, to render an image. That is, that any given input pixel is mapped to either a red-centered logical pixel, or a green-centered logical pixel. === History === PenTile was invented by Candice H. Brown Elliott, for which she was awarded the Society for Information Display's Otto Schade Prize in 2014. The technology was licensed by the company Clairvoyante from 2000 until 2008, during which time several prototype PenTile displays were developed by a number of Asian liquid crystal display (LCD) manufacturers. In March 2008, Samsung Electronics acquired Clairvoyante's PenTile IP assets. Samsung then funded a new company, Nouvoyance, Inc. to continue development of the PenTile technology. == PenTile RGBG == PenTile RGBG layout used in AMOLED and plasma displays uses green pixels interleaved with alternating red and blue pixels. The human eye is most sensitive to green, especially for high resolution luminance information. The green subpixels are mapped to input pixels on a one-to-one basis. The red and blue subpixels are subsampled, reconstructing the chroma signal at a lower resolution. The luminance signal is processed using adaptive subpixel rendering filters to optimize reconstruction of high spatial frequencies from the input image, wherein the green subpixels provide the majority of the reconstruction. The red and blue subpixels are capable of reconstructing the horizontal and vertical spatial frequencies, but not the highest of the diagonal. Diagonal high spatial frequency information in the red and blue channels of the input image are transferred to the green subpixels for image reconstruction. Thus the RG-BG scheme creates a color display with one third fewer subpixels than a traditional RGB-RGB scheme but with the same measured luminance display resolution. This is similar to the Bayer filter commonly used in digital cameras. === Devices === As of 2021, "almost all" OLED screens in portable consumer devices use some form of Pentile subpixel layout. == PenTile RGBW == PenTile RGBW technology, used in LCD, adds an extra subpixel to the traditional red, green and blue subpixels that is a clear area without color filtering material and with the only purpose of letting backlight come through, hence W for white. This makes it possible to produce a brighter image compared to an RGB-matrix while using the same amount of power, or produce an equally bright image while using less power. The PenTile RGBW layout uses each red, green, blue and white subpixel to present high-resolution luminance information to the human eyes' red-sensing and green-sensing cone cells, while using the combined effect of all the color subpixels to present lower-resolution chroma (color) information to all three cone cell types. Combined, this optimizes the match of display technology to the biological mechanisms of human vision. The layout uses one third fewer subpixels for the same resolution as the RGB stripe (RGB-RGB) layout, in spite of having four color primaries instead of the conventional three, using subpixel rendering combined with metamer rendering. Metamer rendering optimizes the energy distribution between the white subpixel and the combined red, green, and blue subpixels: W <> RGB, to improve image sharpness. The display driver chip has an RGB to RGBW color vector space converter and gamut mapping algorithm, followed by metamer and subpixel rendering algorithms. In order to maintain saturated color quality, to avoid simultaneous contrast error between saturated colors and peak white brightness, while simultaneously reducing backlight power requirements, the display backlight brightness is under control of the PenTile driver engine. When the image is mostly desaturated colors, those near white or grey, the backlight brightness is significantly reduced, often to less than 50% peak, while the LCD levels are increased to compensate. When the image has very bright saturated colors, the backlight brightness is maintained at higher levels. The PenTile RGBW also has an optional high-brightness mode that doubles the brightness of the desaturated color image areas, such as black-and-white text, for improved outdoor viewability. === Devices === Motorola MC65 Motorola ES55 Motorola ES400 Motorola Atrix 4G Samsung Galaxy Note 10.1 2014 version Lenovo Yoga 2 Pro Lenovo Yoga 3 Pro HP ENVY TouchSmart 14-k022tx Sleekbook MSI GS60 Ghost Pro 4K Lenovo IdeaPad Y50 4K Asus ZenBook UX303LN 4K Asus ZenBook Pro UX501JW LG UH7500/6500/6100 LG ThinQ G7/G7+ Oculus Quest 1 == Controversy == An ongoing controversy regarding the definition or measurement of resolution of color subpixelated flat panel displays led many people to question the resolution claims of PenTile display products. Journalists have noted that in "just about every flat-panel TV in existence, each pixel is composed of one red, one green, and one blue subpixel (RGB), all of uniform size". In traditional flat-panel screens, the resolution is defined by the number of red, green, and blue subpixels, in groups of three, in an array in each axis. As a result, each pixel or group of subpixels can render any colour on the screen, regardless of neighbouring pixels. This is not the case with PenTile screens. The Video Electronics Standards Association (VESA) method of measuring and defining resolution in color displays is to measure the contrast of line pairs, requiring a minimum of 50% Michelson contrast for displays intended for rendering text. The developers of PenTile displays use this VESA criterion for contrast of line pairs to calculate the resolutions specified. In the RGBG layout the alternate red and blue subpixels are 'shared' or sub-sampled with neighboring pixels. Due to the one third lower subpixel density on PenTile displays the pixel structure may be more visible when compared to RGB stripe displays with the same pixel density. The loss of subpixels for a given resolution specification has led some journalists to describe the use of PenTile as "shady practice" and "sort of cheating". For a given size and resolution specification, the PenTile screen can appear grainy, pixelated, speckled, with blurred text on some saturated colors and backgrounds when compared to RGB stripe color. This effect is understood to be caused by the restriction of the number of subpixels that may participate in the image reconstruction when colors are highly saturated to primaries. In the RGBW case, this is caused as the W subpixel will not be available in order to maintain the saturated color. In the RGBG case, this effect will occur when the color boundary is primarily red or blue, as the fully populated (one green per pixel) sub-pixel cannot contribute. For all other cases, text and especially full color images are effectively reconstructed. == Advantages and disadvantages == The PenTile layout reduces the number of subpixels needed to create a specified resolution. Consequently it is possible to achieve an HD resolution on a PenTile AMOLED screen at lower cost than other technologies, and most reviewers note that "300 ppi" (as per VESA - not full pixels) resolution displays (such as Samsung Galaxy S III) make the PenTile effect less obvious than lower resolution PenTile displays (Droid Razr). The second advantage is lower power consumption: the HTC One S's use of a PenTile display makes it more energy efficient and thinner than equivalent LCD screens, giving it better battery life than the HTC One X's IPS LCD. A PenTile AMOLED screen is also
Higuchi dimension
In fractal geometry, the Higuchi dimension (or Higuchi fractal dimension (HFD)) is an approximate value for the box-counting dimension of the graph of a real-valued function or time series. This value is obtained via an algorithmic approximation so one also talks about the Higuchi method. It has many applications in science and engineering and has been applied to subjects like characterizing primary waves in seismograms, clinical neurophysiology and analyzing changes in the electroencephalogram in Alzheimer's disease. == Formulation of the method == The original formulation of the method is due to T. Higuchi. Given a time series X : { 1 , … , N } → R {\displaystyle X:\{1,\dots ,N\}\to \mathbb {R} } consisting of N {\displaystyle N} data points and a parameter k m a x ≥ 2 {\displaystyle k_{\mathrm {max} }\geq 2} the Higuchi Fractal dimension (HFD) of X {\displaystyle X} is calculated in the following way: For each k ∈ { 1 , … , k m a x } {\displaystyle k\in \{1,\dots ,k_{\mathrm {max} }}\} and m ∈ { 1 , … , k } {\displaystyle m\in \{1,\dots ,k}\} define the length L m ( k ) {\displaystyle L_{m}(k)} by L m ( k ) = N − 1 ⌊ N − m k ⌋ k 2 ∑ i = 1 ⌊ N − m k ⌋ | X N ( m + i k ) − X N ( m + ( i − 1 ) k ) | . {\displaystyle L_{m}(k)={\frac {N-1}{\lfloor {\frac {N-m}{k}}\rfloor k^{2}}}\sum _{i=1}^{\lfloor {\frac {N-m}{k}}\rfloor }|X_{N}(m+ik)-X_{N}(m+(i-1)k)|.} The length L ( k ) {\displaystyle L(k)} is defined by the average value of the k {\displaystyle k} lengths L 1 ( k ) , … , L k ( k ) {\displaystyle L_{1}(k),\dots ,L_{k}(k)} , L ( k ) = 1 k ∑ m = 1 k L m ( k ) . {\displaystyle L(k)={\frac {1}{k}}\sum _{m=1}^{k}L_{m}(k).} The slope of the best-fitting linear function through the data points { ( log 1 k , log L ( k ) ) } {\displaystyle \left\{\left(\log {\frac {1}{k}},\log L(k)\right)\right\}} is defined to be the Higuchi fractal dimension of the time-series X {\displaystyle X} . == Application to functions == For a real-valued function f : [ 0 , 1 ] → R {\displaystyle f:[0,1]\to \mathbb {R} } one can partition the unit interval [ 0 , 1 ] {\displaystyle [0,1]} into N {\displaystyle N} equidistantly intervals [ t j , t j + 1 ) {\displaystyle [t_{j},t_{j+1})} and apply the Higuchi algorithm to the times series X ( j ) = f ( t j ) {\displaystyle X(j)=f(t_{j})} . This results into the Higuchi fractal dimension of the function f {\displaystyle f} . It was shown that in this case the Higuchi method yields an approximation for the box-counting dimension of the graph of f {\displaystyle f} as it follows a geometrical approach (see Liehr & Massopust 2020). == Robustness and stability == Applications to fractional Brownian functions and the Weierstrass function reveal that the Higuchi fractal dimension can be close to the box-dimension. On the other hand, the method can be unstable in the case where the data X ( 1 ) , … , X ( N ) {\displaystyle X(1),\dots ,X(N)} are periodic or if subsets of it lie on a horizontal line (see Liehr & Massopust 2020).