Results 1  10
of
25
Three Main Concerns in Sketch Recognition and an Approach to Addressing Them
, 2002
"... curvilinear configurations to handdrawn sketches. It collects observations from our own recent research, which focused initially on the domain of sketched human stick figures in diverse postures, as well as related computer vision literature. Sketch recognition, i.e., labeling strokes in the i ..."
Abstract

Cited by 29 (0 self)
 Add to MetaCart
curvilinear configurations to handdrawn sketches. It collects observations from our own recent research, which focused initially on the domain of sketched human stick figures in diverse postures, as well as related computer vision literature. Sketch recognition, i.e., labeling strokes in the input with the names of the model parts they depict, would be a key component of higherlevel sketch understanding processes that reason about the recognized configurations. A sketch recognition technology must meet three main requirements. It must cope reliably with the pervasive variability of hand sketches, provide interactive performance, and be easily extensible to new configurations. We argue that useful sketch recognition may be within the grasp of current research, if these requirements are addressed systematically and in concert.
Active coevolutionary learning of deterministic finite automata
 Journal of Machine Learning Research
, 2005
"... This paper describes an active learning approach to the problem of grammatical inference, specifically the inference of deterministic finite automata (DFAs). We refer to the algorithm as the estimationexploration algorithm (EEA). This approach differs from previous passive and active learning appro ..."
Abstract

Cited by 29 (9 self)
 Add to MetaCart
This paper describes an active learning approach to the problem of grammatical inference, specifically the inference of deterministic finite automata (DFAs). We refer to the algorithm as the estimationexploration algorithm (EEA). This approach differs from previous passive and active learning approaches to grammatical inference in that training data is actively proposed by the algorithm, rather than passively receiving training data from some external teacher. Here we show that this algorithm outperforms one version of the most powerful set of algorithms for grammatical inference, evidence driven state merging (EDSM), on randomlygenerated DFAs. The performance increase is due to the fact that the EDSM algorithm only works well for DFAs with specific balances (percentage of positive labelings), while the EEA is more consistent over a wider range of balances. Based on this finding we propose a more general method for generating DFAs to be used in the development of future grammatical inference algorithms.
Grammar Inference, Automata Induction, and Language Acquisition
 Handbook of Natural Language Processing
, 2000
"... The natural language learning problem has attracted the attention of researchers for several decades. Computational and formal models of language acquisition have provided some preliminary, yet promising insights of how children learn the language of their community. Further, these formal models als ..."
Abstract

Cited by 21 (1 self)
 Add to MetaCart
The natural language learning problem has attracted the attention of researchers for several decades. Computational and formal models of language acquisition have provided some preliminary, yet promising insights of how children learn the language of their community. Further, these formal models also provide an operational framework for the numerous practical applications of language learning. We will survey some of the key results in formal language learning. In particular, we will discuss the prominent computational approaches for learning different classes of formal languages and discuss how these fit in the broad context of natural language learning.
Learning DFA from Simple Examples
, 1997
"... Efficient learning of DFA is a challenging research problem in grammatical inference. It is known that both exact and approximate (in the PAC sense) identifiability of DFA is hard. Pitt, in his seminal paper posed the following open research problem: "Are DFAPACidentifiable if examples are drawn ..."
Abstract

Cited by 18 (4 self)
 Add to MetaCart
Efficient learning of DFA is a challenging research problem in grammatical inference. It is known that both exact and approximate (in the PAC sense) identifiability of DFA is hard. Pitt, in his seminal paper posed the following open research problem: "Are DFAPACidentifiable if examples are drawn from the uniform distribution, or some other known simple distribution?" [25]. We demonstrate that the class of simple DFA (i.e., DFA whose canonical representations have logarithmic Kolmogorov complexity) is efficiently PAC learnable under the Solomonoff Levin universal distribution. We prove that if the examples are sampled at random according to the universal distribution by a teacher that is knowledgeable about the target concept, the entire class of DFA is efficiently PAC learnable under the universal distribution. Thus, we show that DFA are efficiently learnable under the PACS model [6]. Further, we prove that any concept that is learnable under Gold's model for learning from characteristic samples, Goldman and Mathias' polynomial teachability model, and the model for learning from example based queries is also learnable under the PACS model.
Learning to Verify Safety Properties
 In LNCS 3308, Proc. of ICFEM’04
, 2004
"... We present a novel approach for verifying safety properties of finite state machines communicating over unbounded FIFO channels that is based on applying machine learning techniques. We assume that we are given a model of the system and learn the set of reachable states from a sample set of exec ..."
Abstract

Cited by 11 (4 self)
 Add to MetaCart
We present a novel approach for verifying safety properties of finite state machines communicating over unbounded FIFO channels that is based on applying machine learning techniques. We assume that we are given a model of the system and learn the set of reachable states from a sample set of executions of the system, instead of attempting to iteratively compute the reachable states. The learnt set of reachable states is then used to either prove that the system is safe or to produce a valid execution of the system leading to an unsafe state (i.e. a counterexample). We have implemented this method for verifying FIFO automata in a tool called Lever that uses a regular language learning algorithm called RPNI. We apply our tool to a few case studies and report our experience with this method. We also demonstrate how this method can be generalized and applied to the verification of other infinite state systems.
Regular inference for state machines with parameters
 In FASE
, 2006
"... Abstract. Techniques for inferring a regular language, in the form of a finite automaton, from a sufficiently large sample of accepted and nonaccepted input words, have been employed to construct models of software and hardware systems, for use, e.g., in test case generation. We intend to adapt thes ..."
Abstract

Cited by 10 (3 self)
 Add to MetaCart
Abstract. Techniques for inferring a regular language, in the form of a finite automaton, from a sufficiently large sample of accepted and nonaccepted input words, have been employed to construct models of software and hardware systems, for use, e.g., in test case generation. We intend to adapt these techniques to construct state machine models of entities of communication protocols. The alphabet of such state machines can be very large, since a symbol typically consists of a protocol data unit type with a number of parameters, each of which can assume many values. In typical algorithms for regular inference, the number of needed input words grows with the size of the alphabet and the size of the minimal DFA accepting the language. We therefore modify such an algorithm (Angluin’s algorithm) so that its complexity grows not with the size of the alphabet, but only with the size of a certain symbolic representation of the DFA. The main new idea is to infer, for each state, a partitioning of input symbols into equivalence classes, under the hypothesis that all input symbols in an equivalence class have the same effect on the state machine. Whenever such a hypothesis is disproved, equivalence classes are refined. We show that our modification retains the good properties of Angluin’s original algorithm, but that its complexity grows with the size of our symbolic DFA representation rather than with the size of the alphabet. We have implemented the algorithm; experiments on synthesized examples are consistent with these complexity results. 1
The QSM algorithm and its application to software behavior model induction
 APPLIED ARTIFICIAL INTELLIGENCE
, 2008
"... This article presents a novel application of grammatical inference techniques to the synthesis of behavior models of software systems. This synthesis is used for the elicitation of software requirements. This problem is formulated as a deterministic finitestate automaton induction problem from pos ..."
Abstract

Cited by 10 (2 self)
 Add to MetaCart
This article presents a novel application of grammatical inference techniques to the synthesis of behavior models of software systems. This synthesis is used for the elicitation of software requirements. This problem is formulated as a deterministic finitestate automaton induction problem from positive and negative scenarios provided by an end user of the softwaretobe. A querydriven state merging (QSM) algorithm is proposed. It extends the Regular Positive and Negative Inference (RPNI) and bluefringe algorithms by allowing membership queries to be submitted to the end user. State merging operations can be further constrained by some prior domain knowledge formulated as fluents, goals, domain properties, and models of external software components. The incorporation of domain knowledge both reduces the number of queries and guarantees that the induced model is consistent with such knowledge. The proposed techniques are implemented in the ISIS tool and practical evaluations on standard requirements engineering test cases and synthetic data illustrate the interest of this approach.
Learning Communicating Automata from MSCs
, 2009
"... This paper is concerned with bridging the gap between requirements and distributed systems. Requirements are defined as basic message sequence charts (MSCs) specifying positive and negative scenarios. Communicating finitestate machines (CFMs), i.e., finite automata that communicate via FIFO buffer ..."
Abstract

Cited by 8 (2 self)
 Add to MetaCart
This paper is concerned with bridging the gap between requirements and distributed systems. Requirements are defined as basic message sequence charts (MSCs) specifying positive and negative scenarios. Communicating finitestate machines (CFMs), i.e., finite automata that communicate via FIFO buffers, act as system realizations. The key contribution is a generalization of Angluin’s learning algorithm for synthesizing CFMs from MSCs. This approach is exact—the resulting CFM precisely accepts the set of positive scenarions and rejects all negative ones—and yields fully asynchronous implementations. The paper investigates for which classes of MSC languages CFMs can be learned, presents an optimization technique for learning partial orders, and provides substantial empirical evidence indicating the practical feasibility of the approach.
Incremental LearningBased Testing for Reactive Systems
"... Abstract. We show how the paradigm of learningbased testing (LBT) can be applied to automate specificationbased blackbox testing of reactive systems. Since reactive systems can be modeled as Kripke structures, we introduce an efficient incremental learning algorithm IKL for such structures. We sh ..."
Abstract

Cited by 5 (2 self)
 Add to MetaCart
Abstract. We show how the paradigm of learningbased testing (LBT) can be applied to automate specificationbased blackbox testing of reactive systems. Since reactive systems can be modeled as Kripke structures, we introduce an efficient incremental learning algorithm IKL for such structures. We show how an implementation of this algorithm combined with an efficient model checker such as NuSMV yields an effective learningbased testing architecture for automated test case generation (ATCG), execution and evaluation, starting from temporal logic requirements. 1
Learning to Verify Systems
, 2006
"... Making high quality and reliable software systems remains a difficult problem. One approach to address this problem is automated verification which attempts to demonstrate algorithmically that a software system meets its specification. However, verification of software systems is not easy: such sys ..."
Abstract

Cited by 3 (1 self)
 Add to MetaCart
Making high quality and reliable software systems remains a difficult problem. One approach to address this problem is automated verification which attempts to demonstrate algorithmically that a software system meets its specification. However, verification of software systems is not easy: such systems are often modeled using abstractions of infinite structures such as unbounded integers, infinite memory for allocation, unbounded space for call stack, unrestricted queue sizes and so on. It can be shown that for most classes of such systems, the verification problem is actually undecidable (there exists no algorithm which will always give the correct answer for arbitrary inputs). In spite of this negative theoretical result, techniques have been developed which are successful on some practical examples although they are not guaranteed to always work. This dissertation is in a similar spirit and develops a new paradigm for automated verification of large or infinite state systems. We observe that even if the state space of a system is infinite, for practical examples, the set of reachable states (or other fixpoints needed for verification) is often expressible in a simple representation. Based on this observation, we propose an entirely new approach to verification: the idea is to use techniques from computational learning theory to identify the reachable states (or other fixpoints) and then verify the property of interest. To use learning techniques, we solve key problems of