Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

N G van Kampen 1981 Amsterdam: North-Holland xiv + 419 pp price Dfl 180 This is a book which, at a lower price, could be expected to become an essential part of the library of every physical scientist concerned with problems involving fluctuations and stochastic processes, as well as those who just enjoy a beautifully written book. It provides an extensive graduate-level introduction which is clear, cautious, interesting and readable.

https://iopscience.iop.org/article/10.1088/0031-9112/34/4/040/pdf

Stochastic Processes in Physics and Chemistry

Spark SQL is a new module in Apache Spark that integrates relational processing with Spark's functional programming API. Built on our experience with Shark, Spark SQL lets Spark programmers leverage the benefits of relational processing (e.g. declarative queries and optimized storage), and lets SQL users call complex analytics libraries in Spark (e.g. machine learning). Compared to previous systems, Spark SQL makes two main additions. First, it offers much tighter integration between relational and procedural processing, through a declarative DataFrame API that integrates with procedural Spark code. Second, it includes a highly extensible optimizer, Catalyst, built using features of the Scala programming language, that makes it easy to add composable rules, control code generation, and define extension points. Using Catalyst, we have built a variety of features (e.g. schema inference for JSON, machine learning types, and query federation to external databases) tailored for the complex needs of modern data analysis. We see Spark SQL as an evolution of both SQL-on-Spark and of Spark itself, offering richer APIs and optimizations while keeping the benefits of the Spark programming model.

http://people.csail.mit.edu/matei/papers/2015/sigmod_spark_sql.pdf

Spark SQL: Relational Data Processing in Spark

Principal components analysis (PCA) is a classic method for the reduction of dimensionality of data in the form of n observations (or cases) of a vector with p variables. Contemporary datasets often have p comparable with or even much larger than n. Our main assertions, in such settings, are (a) that some initial reduction in dimensionality is desirable before applying any PCA-type search for principal modes, and (b) the initial reduction in dimensionality is best achieved by working in a basis in which the signals have a sparse representation. We describe a simple asymptotic model in which the estimate of the leading principal component vector via standard PCA is consistent if and only if p(n)/n → 0. We provide a simple algorithm for selecting a subset of coordinates with largest sample variances, and show that if PCA is done on the selected subset, then consistency is recovered, even if p(n) ≫ n.

/pdf/on-consistency-and-sparsity-for-principal-components-3zjjz9i18o.pdf

On Consistency and Sparsity for Principal Components Analysis in High Dimensions

The problem of inducing, learning or inferring grammars has been studied for decades, but only in recent years has grammatical inference emerged as an independent field with connections to many scientific disciplines, including bio-informatics, computational linguistics and pattern recognition. This book meets the need for a comprehensive and unified summary of the basic techniques and results, suitable for researchers working in these various areas. In Part I, the objects of use for grammatical inference are studied in detail: strings and their topology, automata and grammars, whether probabilistic or not. Part II carefully explores the main questions in the field: What does learning mean? How can we associate complexity theory with learning? In Part III the author describes a number of techniques and algorithms that allow us to learn from text, from an informant, or through interaction with the environment. These concern automata, grammars, rewriting systems, pattern languages or transducers.

Grammatical Inference: Learning Automata and Grammars

Among the various proposals answering the shortcomings of Document Type Definitions (DTDs), XML Schema is the most widely used. Although DTDs and XML Schema Definitions (XSDs) differ syntactically, they are still quite related on an abstract level. Indeed, freed from all syntactic sugar, XML Schemas can be seen as an extension of DTDs with a restricted form of specialization. In the present paper, we inspect a number of DTDs and XSDs harvested from the web and try to answer the following questions: (1) which of the extra features/expressiveness of XML Schema not allowed by DTDs are effectively used in practice; and, (2) how sophisticated are the structural properties (i.e. the nature of regular expressions) of the two formalisms. It turns out that at present real-world XSDs only sparingly use the new features introduced by XML Schema: on a structural level the vast majority of them can already be defined by DTDs. Further, we introduce a class of simple regular expressions and obtain that a surprisingly high fraction of the content models belong to this class. The latter result sheds light on the justification of simplifying assumptions that sometimes have to be made in XML research.

DTDs versus XML schema: a practical study

The common abstraction of XML Schema by unranked regular tree languages is not entirely accurate. To shed some light on the actual expressive power of XML Schema, intuitive semantical characterizations of the Element Declarations Consistent (EDC) rule are provided. In particular, it is obtained that schemas satisfying EDC can only reason about regular properties of ancestors of nodes. Hence, with respect to expressive power, XML Schema is closer to DTDs than to tree automata. These theoretical results are complemented with an investigation of the XML Schema Definitions (XSDs) occurring in practice, revealing that the extra expressiveness of XSDs over DTDs is only used to a very limited extent. As this might be due to the complexity of the XML Schema specification and the difficulty of understanding the effect of constraints on typing and validation of schemas, a simpler formalism equivalent to XSDs is proposed. It is based on contextual patterns rather than on recursive types and it might serve as a light-weight front end for XML Schema. Next, the effect of EDC on the way XML documents can be typed is discussed. It is argued that a cleaner, more robust, larger but equally feasible class is obtained by replacing EDC with the notion of 1-pass preorder typing (1PPT): schemas that allow one to determine the type of an element of a streaming document when its opening tag is met. This notion can be defined in terms of grammars with restrained competition regular expressions and there is again an equivalent syntactical formalism based on contextual patterns. Finally, algorithms for recognition, simplification, and inclusion of schemas for the various classes are given.

/pdf/expressiveness-and-complexity-of-xml-schema-qouxg4ykks.pdf

Expressiveness and complexity of XML Schema

The common abstraction of XML Schema by unranked regular tree languages is not entirely accurate. To shed some light on the actual expressive power of XML Schema, intuitive semantical characterizations of the Element Declarations Consistent (EDC) rule are provided. In particular, it is obtained that schemas satisfying EDC can only reason about regular properties of ancestors of nodes. Hence, with respect to expressive power, XML Schema is closer to DTDs than to tree automata. These theoretical results are complemented with an investigation of the XML Schema Definitions (XSDs) occurring in practice, revealing that the extra expressiveness ofXSDs over DTDs is only used to a very limited extent. As this might be due to the complexity of the XML Schema specification and the difficulty of understanding the effect of constraints on typing and validation of schemas, a simpler formalism equivalent to XSDs is proposed. It is based on contextual patterns rather than on recursive types and it might serve as a light-weight front end for XML Schema. Next, the effect of EDC on the way XML documents can be typed is discussed. It is argued that a cleaner, more robust, larger but equally feasible class is obtained by replacing EDC with the notion of 1-pass preorder typing (1PPT): schemas that allow one to determine the type of an element of a streaming document when its opening tag is met. This notion can be defined in terms of grammars with restrained competition regular expressions and there is again an equivalent syntactical formalism based on contextual patterns. Finally, algorithms for recognition, simplification, and inclusion of schemas for the various classes are given.

Expressiveness and complexity of XML schema

We consider the problem to infer a concise Document Type Definition (DTD) for a given set of XML-documents, a problem which basically reduces to learning of concise regular expressions from positive example strings. We identify two such classes: single occurrence regular expressions (SOREs) and chain regular expressions (CHAREs). Both classes capture the far majority of the regular expressions occurring in practical DTDs and are succinct by definition. We present the algorithm iDTD (infer DTD) that learns SOREs from strings by first inferring an automaton by known techniques and then translating that automaton to a corresponding SORE, possibly by repairing the automaton when no equivalent SORE can be found. In the process, we introduce a novel automaton to regular expression rewrite technique which is of independent interest. We show that iDTD outperforms existing systems in accuracy, conciseness and speed. In a scenario where only a very small amount of XML data is available, for instance when generated by Web service requests or by answers to queries, iDTD produces regular expressions which are too specific. Therefore, we introduce a novel learning algorithm CRX that directly infers CHAREs (which form a subclass of SOREs) without going through an automaton representation. We show that CRX performs very well within its target class on very small data sets. Finally, we discuss incremental computation, noise, numerical predicates, and the generation of XML Schemas.

http://www.vldb.org/conf/2006/p115-bex.pdf

Inference of concise DTDs from XML data

Although the presence of a schema enables many optimizations for operations on XML documents, recent studies have shown that many XML documents in practice either do not refer to a schema, or refer to a syntactically incorrect one. It is therefore of utmost importance to provide tools and techniques that can automatically generate schemas from sets of sample documents. While previous work in this area has mostly focused on the inference of Document Type Definitions (DTDs for short), we will consider the inference of XML Schema Definitions (XSDs for short) --- the increasingly popular schema formalism that is turning DTDs obsolete. In contrast to DTDs where the content model of an element depends only on the element's name, the content model in an XSD can also depend on the context in which the element is used. Hence, while the inference of DTDs basically reduces to the inference of regular expressions from sets of sample strings, the inference of XSDs also entails identifying from a corpus of sample documents the contexts in which elements bear different content models. Since a seminal result by Gold implies that no inference algorithm can learn the complete class of XSDs from positive examples only, we focus on a class of XSDs that captures most XSDs occurring in practice. For this class, we provide a theoretically complete algorithm that always infers the correct XSD when a sufficiently large corpus of XML documents is available. In addition, we present a variant of this algorithm that works well on real-world (and therefore incomplete) data sets.

/pdf/inferring-xml-schema-definitions-from-xml-data-168zn1cxpa.pdf

Geert Jan Bex

Papers

DTDs versus XML schema: a practical study

Expressiveness and complexity of XML Schema

Expressiveness and complexity of XML schema

Inference of concise DTDs from XML data

Inferring XML schema definitions from XML data