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

Data Mining - Concepts and Techniques.

Social Network Analysis

We study the problem of representation learning in heterogeneous networks. Its unique challenges come from the existence of multiple types of nodes and links, which limit the feasibility of the conventional network embedding techniques. We develop two scalable representation learning models, namely metapath2vec and metapath2vec++. The metapath2vec model formalizes meta-path-based random walks to construct the heterogeneous neighborhood of a node and then leverages a heterogeneous skip-gram model to perform node embeddings. The metapath2vec++ model further enables the simultaneous modeling of structural and semantic correlations in heterogeneous networks. Extensive experiments show that metapath2vec and metapath2vec++ are able to not only outperform state-of-the-art embedding models in various heterogeneous network mining tasks, such as node classification, clustering, and similarity search, but also discern the structural and semantic correlations between diverse network objects.

https://dl.acm.org/doi/pdf/10.1145/3097983.3098036

metapath2vec: Scalable Representation Learning for Heterogeneous Networks

The Analysis of Time Series: An Introduction, 4th edn. By C. Chatfield. ISBN 0 412 31820 2. Chapman and Hall, London, 1989. 242 pp. £13.50.

The Analysis of Time Series: An Introduction

We propose a framework for representing conditioned probabilistic relational data. In this framework the existence of tuples in possible worlds is determined by Boolean expressions composed from elementary events. The probability of a possible world is computed from the probabilities associated with these elementary events. In addition, a set of global constraints conditions the database. Conditioning is the formalization of the process of adding knowledge to a database. Some worlds may be impossible given the constraints and the probabilities of possible worlds are accordingly re-defined. The new constraints can come from the observation of the existence or non-existence of a tuple, from the knowledge of a specific rule, such as the existence of an exclusive set of tuples, or from the knowledge of a general rule, such as a functional dependency. We are therefore interested in computing a concise representation of the possible worlds and their respective probabilities after the addition of new constraints, namely an equivalent probabilistic database instance without constraints after conditioning. We devise and present a general algorithm for this computation. Unfortunately, the general problem involves the simplification of general Boolean expressions and is NP-hard. We therefore identify specific practical families of constraints for which we devise and present efficient algorithms.

A Framework for Conditioning Uncertain Relational Data

The dissemination of messages to a vast number of mobile users has raised a lot of attention. This issue is inherent in emerging applications, such as location-based targeted advertising, selective information disseminating, and ride sharing. In this paper, we examine how to support location-based message dissemination in an effective and efficient manner. Our main idea is to develop a location-aware version of the Pub/Sub model, which was designed for message dissemination. While a lot of studies have successfully used this model to match the interest of subscriptions (e.g., the properties of potential customers) and events (e.g., information of casual users), the issues of incorporating the location information of subscribers and publishers have not been well addressed. We propose to model subscriptions and events by boolean expressions and location data. This allows complex information to be specified. However, since the number of publishers and subscribers can be enormous, the time cost for matching subscriptions and events can be prohibitive. To address this problem, we have developed the \(R^I\)-tree. This data structure is an integration of the R-tree and the dynamic interval-tree. Together with our novel pruning strategy on \(R^I\)-tree, our solution can effectively and efficiently return the top-k subscriptions with respect to an event. We have performed extensive evaluations to verify our approach.

/pdf/efficient-top-k-subscription-matching-for-location-aware-1dbxm0aqp1.pdf

Efficient Top-k Subscription Matching for Location-Aware Publish/Subscribe

In a social tagging system, resources (such as photos, video and web pages) are associated with tags. These tags allow the resources to be effectively searched through tag-based keyword matching using traditional IR techniques. We note that in many such systems, tags of a resource are often assigned by a diverse audience of causal users (taggers). This leads to two issues that gravely affect the effectiveness of resource retrieval: (1) Noise: tags are picked from an uncontrolled vocabulary and are assigned by untrained taggers. The tags are thus noisy features in resource retrieval. (2) A multitude of aspects: different taggers focus on different aspects of a resource. Representing a resource using a flattened bag of tags ignores this important diversity of taggers. To improve the effectiveness of resource retrieval in social tagging systems, we propose CubeLSI — a technique that extends traditional LSI to include taggers as another dimension of feature space of resources. We compare CubeLSI against a number of other tag-based retrieval models and show that CubeLSI significantly outperforms the other models in terms of retrieval accuracy. We also prove two interesting theorems that allow CubeLSI to be very efficiently computed despite the much enlarged feature space it employs.

/pdf/cubelsi-an-effective-and-efficient-method-for-searching-24tbr2alcw.pdf

CubeLSI: An effective and efficient method for searching resources in social tagging systems

A common approach to improve the reliability of query results based on error-prone sensors is to introduce redundant sensors. However, using multiple sensors to generate the value for a data item can be expensive, especially in wireless environments where continuous queries are executed. Moreover, some sensors may not be working properly and their readings need to be discarded. In this paper, we propose a statistical approach to decide which sensor nodes to be used to answer a query. In particular, we propose to solve the problem with the aid of continuous probabilistic query (CPQ), which is originally used to manage uncertain data and is associated with a probabilistic guarantee on the query result. Based on the historical data values from the sensor nodes, the query type, and the requirement on the query, we present methods to select an appropriate set of sensors and provide reliable answers for several common aggregate queries. Our statistics-based sensor node selection algorithm is demonstrated in a number of simulation experiments, which shows that a small number of sensor nodes can provide accurate and robust query results.

/pdf/a-statistics-based-sensor-selection-scheme-for-continuous-2ppxgdhagy.pdf

A statistics-based sensor selection scheme for continuous probabilistic queries in sensor networks

In applications like sensor network monitoring and location-based services, due to limited network bandwidth and battery power, a system cannot always acquire accurate and fresh data from the external environment. To capture data errors in these environments, recent researches have proposed to model uncertainty as a probability distribution function (pdf), as well as the notion of probabilistic queries, which provide statistical guarantees on answer correctness. In this paper, we present an entropy-based metric to quantify the degree of ambiguity of probabilistic query answers due to data uncertainty. Based on this metric, we develop a new method to improve the query answer quality. The main idea of this method is to acquire (or probe) data from a selected set of sensing devices, in order to reduce data uncertainty and improve the quality of a query answer. Given that a query is assigned a limited number of probing resources, we investigate how the quality of a query answer can attain an optimal improvement. To improve the efficiency of our solution, we further present heuristics which achieve near-to-optimal quality improvement. We generalize our solution to handle multiple queries. An experimental simulation over a realistic dataset is performed to validate our approaches.

/pdf/quality-aware-probing-of-uncertain-data-with-resource-1lmjd6519i.pdf

Reynold Cheng

Papers

A Framework for Conditioning Uncertain Relational Data

Efficient Top-k Subscription Matching for Location-Aware Publish/Subscribe

CubeLSI: An effective and efficient method for searching resources in social tagging systems

A statistics-based sensor selection scheme for continuous probabilistic queries in sensor networks

Quality-Aware Probing of Uncertain Data with Resource Constraints