Machine Learning on

SOFSEM2025

Mon, 20 Jan 2025 00:00:00 +0000

This paper is concerned with learning some context-free languages (cfls) over infinite alphabets from membership and equivalence queries. Argyros and D’Antoni (2018) proposed a query learning algorithm for regular languages over infinite alphabets using…

SOFSEM2024

Fri, 23 Feb 2024 00:00:00 +0000

ICGI2023

Wed, 12 Jul 2023 00:00:00 +0000

This paper is concerned with the identification in the limit from positive data of substitutable context-free languages CFLs over infinite alphabets. [ClarkE07] showed that substitutable CFLs over finite alphabets are learnable in this learning paradigm. We show that substitutable CFLs generated by grammars whose production rules may have predicates that represent sets of potentially infinitely many terminal symbols in a compact manner are learnable if the terminal symbol sets represented by those predicates are learnable, under a certain condition. This can be seen as a result parallel to [ArgyrosDA2018]’s work (2018) that amplifies the query learnability of predicate classes to that of symbolic automata classes. Our result is the first that shows such amplification is possible for identifying some CFLs in the limit from positive data.

ICGI2023-Kanazawa

Wed, 12 Jul 2023 00:00:00 +0000

We consider an extension of distributional learning of context-free languages (from positive data and membership queries), where nonterminals are represented by extended regular expressions (allowi…

ICGI2021-Kambe

Mon, 23 Aug 2021 00:00:00 +0000

We propose an inside-outside algorithm for stochastic macro grammars. Our approach is based on types, which has been inspired by type-based approaches to reasoning about functional programs and higher-order grammars. By considering type derivations instead of ordinary word derivation sequences, we can naturally extend the standard inside-outside algorithm for stochastic context-free grammars to obtain the algorithm for stochastic macro grammars. We have implemented the algorithm and confirmed its effectiveness through an application to the learning of macro grammars.

ICGI2023-Kaito

Mon, 23 Aug 2021 00:00:00 +0000

We propose a query learning algorithm for an extension of weighted finite automata (WFAs), named symbolic weighted finite automata (SWFAs), which can handle strings over infinite alphabets more efficiently. Based on the idea of symbolic finite automata, SWFAs generalize WFAs by allowing transitions to be functions from a possibly infinite alphabet to weights. Our algorithm can learn SWFAs if functions in transitions are also learnable by queries. We also investigate minimization and equivalence checking for SWFAs.

ICGI2023-kanazawa

Mon, 23 Aug 2021 00:00:00 +0000

We consider a generalization of the “dual” approach to distributional learning of context-free grammars, where each nonterminal $A$ is associated with a string set $X_A$ characterized by a finite set $C$ of contexts. Rather than letting $X_A$ be the set of all strings accepted by all contexts in $C$ as in previous works, we allow more flexible uses of the contexts in $C$, using some of them positively (contexts that accept the strings in $X_A$) and others negatively (contexts that do not accept any strings in $X_A$). The resulting more general algorithm works in essentially the same way as before, but on a larger class of context-free languages.

TCS2020-Narisada-gap

Mon, 06 Apr 2020 00:00:00 +0000

In this paper, we introduce new types of approximate palindromes called single-arm-gapped palindromes (shortly SAGPs). A SAGP contains a gap in either its left or right arm, which is in the form of either $wgucu^Rw^R$ or $wucu^Rgw^R$, where $w$ and $u$ are non-empty strings, $w^R$ and $u^R$ are respectively the reversed strings of $w$ and $u$, $g$ is a string called a gap, and $c$ is either a single character or the empty string. Here we call $wu$ and $u^Rw^R$ the arm of the SAGP, and $|uv|$ the length of the arm. We classify SAGPs into two groups: those which have $ucu^R$ as a maximal palindrome (type-1), and the others (type-2). We propose several algorithms to compute type-1 SAGPs with longest arms occurring in a given string, based on suffix arrays. Then, we propose a linear-time algorithm to compute all type-1 SAGPs with longest arms, based on suffix trees. Also, we show how to compute type-2 SAGPs with longest arms in linear time. We also perform some preliminary experiments to show practical performances of the proposed methods.

2019-09-fit

Thu, 05 Sep 2019 00:00:00 +0000

2019-01-GandALF

Tue, 01 Jan 2019 00:00:00 +0000

We propose a query learning algorithm for residual symbolic finite automata (RSFAs). Symbolic finite automata (SFAs) are finite automata whose transitions are labeled by predicates over a Boolean algebra, in which a big collection of characters leading the same transition may be represented by a single predicate. Residual finite automata (RFAs) are a special type of non-deterministic finite automata which can be exponentially smaller than the minimum deterministic finite automata and have a favorable property for learning algorithms. RSFAs have both properties of SFAs and RFAs and can have more succinct representation of transitions and fewer states than RFAs and deterministic SFAs accepting the same language. The implementation of our algorithm efficiently learns RSFAs over a huge alphabet and outperforms an existing learning algorithm for deterministic SFAs. The result also shows that the benefit of non-determinism in efficiency is even larger in learning SFAs than non-symbolic automata.

2019-01-J-Comput-Syst-Sci

Tue, 01 Jan 2019 00:00:00 +0000

Abstract Approaches based on the idea generically called distributional learning have been making great success in the algorithmic learning of several rich subclasses of context-free languages and their extensions. Those language classes are defined by properties concerning string-context relation. In this paper, we present a distributional learning algorithm for conjunctive grammars with the k -finite context property ( k - fcp ) for each natural number k . We also compare our result with the closely related work by Clark et al. (JMLR 2010) [5] on learning some context-free grammars using contextual binary feature grammars ( cbfg s). We prove that the context-free grammars targeted by their algorithm have the k - fcp . Moreover, we show that every exact cbfg has the k - fcp , too, while not all of them are learnable by their algorithm. Clark et al. conjectured a learning algorithm for exact cbfg s should exist. This paper answers their conjecture in a positive way.

2017-01-IEEE-BigData

Fri, 01 Dec 2017 00:00:00 +0000

2017-11-gpw-ohkawa

Sat, 11 Nov 2017 00:00:00 +0000

2017-01-ALT

Sun, 01 Oct 2017 00:00:00 +0000

A ZDD is a directed acyclic graph that represents a family of sets over a fixed universe set. In this paper, we propose an algorithm that learns zero-suppressed binary decision diagrams (ZDDs) usin…

2017-01-LATA

Wed, 01 Mar 2017 00:00:00 +0000

We identify the properties of context-free grammars that exactly correspond to the behavior of the dual and primal versions of Clark and Yoshinaka’s distributional learning algorithm and call them the very weak finite context/kernel property. We show that the…

2017-01-Ann-Math-Artif-Intell

Sun, 01 Jan 2017 00:00:00 +0000

This paper is concerned with a sufficient condition under which a concept class is learnable in Gold’s classical model of identification in the limit

2016-01-paper-021

Thu, 21 Jan 2016 00:00:00 +0000

2016-01-Fundam-Informaticae

Fri, 01 Jan 2016 00:00:00 +0000

Abstract could not be automatically retrieved from the linked source. Please add the official abstract manually if available.

2015-11-paper-022

Sun, 08 Nov 2015 00:00:00 +0000

2015-11-paper-023

Sun, 08 Nov 2015 00:00:00 +0000

2015-11-qbf

Sun, 08 Nov 2015 00:00:00 +0000

2015-01-LATA

Sun, 01 Mar 2015 00:00:00 +0000

Approaches based on the idea generically called distributional learning have been making great success in the algorithmic learning of context-free languages and their extensions. We in this paper show that conjunctive grammars are also learnable by a distributional…

2015-01-FG

Thu, 01 Jan 2015 00:00:00 +0000

We study tree-generating almost linear second-order ACGs that admit bounded nonlinearity either on the context side or on the substructure side, and give distributional learning algorithms for them.

2014-07-paper-031

Thu, 17 Jul 2014 00:00:00 +0000

2014-01-ICAART-1

Sat, 01 Mar 2014 00:00:00 +0000

Most existing reinforcement learning algorithms require many trials until they obtain optimal policies. In this study, we apply transfer learning to reinforcement learning to realize greater efficiency. We propose a new algorithm called TR-MAX, based on the R-MAX algorithm. TR-MAX transfers the transition and reward probabilities from a source task to a target task as prior knowledge. We theoretically analyze the sample complexity of TR-MAX. Moreover, we show that TR-MAX performs much better in practice than R-MAX in maze tasks.

2014-01-ISAIM

Wed, 01 Jan 2014 00:00:00 +0000

2014-01-KDIR

Wed, 01 Jan 2014 00:00:00 +0000

2014-01-Mach-Learn

Wed, 01 Jan 2014 00:00:00 +0000

Natural languages require grammars beyond context-free for their description. Here we extend a family of distributional learning algorithms for context-fre

2013-11-paper-033

Wed, 13 Nov 2013 00:00:00 +0000

2013-04-n-57-2-03696

Wed, 24 Apr 2013 00:00:00 +0000

2013-10-ALT

Tue, 01 Jan 2013 00:00:00 +0000

In recent years different interesting subclasses of cfls have been found to be learnable by techniques generically called distributional learning. The theoretical study on the exact learning of cfls by those techniques under different learning scheme is now quite…

2012-01-ICGI

Sat, 01 Sep 2012 00:00:00 +0000

Semilinearity is widely held to be a linguistic invariant but, controversially, some linguistic phenomena in languages like Old Georgian and Yoruba seem to violate this constraint. In this paper we…

2012-01-ICGI-2

Sat, 01 Sep 2012 00:00:00 +0000

A Bayesian manner which marginalizes transition probabilities can be generally applied to various kinds of probabilistic finite state machine models. Based on such a Bayesian manner, we implemented…

2012-01-TAG

Sat, 01 Sep 2012 00:00:00 +0000

Ryo Yoshinaka. Proceedings of the 11th International Workshop on Tree Adjoining Grammars and Related Formalisms (TAG+11). 2012.

2012-01-ICAART-1

Wed, 01 Feb 2012 00:00:00 +0000

2012-03-LATA

Sun, 01 Jan 2012 00:00:00 +0000

Recently several “distributional learning algorithms” have been proposed and have made great success in learning different subclasses of context-free grammars. The distributional learning models and exploits the relation between strings and contexts that…

2011-01-ALT

Sat, 01 Oct 2011 00:00:00 +0000

This paper demonstrates how existing distributional learning techniques for context-free grammars can be adapted to simple context-free tree grammars in a straightforward manner once the necessary notions and properties for string languages have been redefined for…

2011-01-ISCIS

Thu, 01 Sep 2011 00:00:00 +0000

Angluin (1980) showed that there is a consistent and conservative learner for the class of non-erasing pattern languages; however, most of these learners are NP-hard. In the current work, the complexity of consistent polynomial time learners for the class of…

2011-08-satori

Thu, 04 Aug 2011 00:00:00 +0000

2011-07-paper-049

Tue, 19 Jul 2011 00:00:00 +0000

2011-01-Developments-in-Language-Theory

Fri, 01 Jul 2011 00:00:00 +0000

Recent studies on grammatical inference have demonstrated the benefits of “distributional learning” for learning context-free and context-sensitive languages. Distributional learning models and exploits the relation between strings and contexts in the…

2011-01-LACL-2

Fri, 01 Jul 2011 00:00:00 +0000

Recent studies on grammatical inference have demonstrated the benefits of the learning strategy called “distributional learning” for context-free and multiple context-free languages. This paper gives a comprehensive view of distributional learning of…

2011-01-Theor-Comput-Sci

Sat, 01 Jan 2011 00:00:00 +0000

Abstract could not be automatically retrieved from the linked source. Please add the official abstract manually if available.

2010-01-FG

Fri, 01 Jan 2010 00:00:00 +0000

We present an algorithm for the inference of some Multiple Context-Free Grammars from Membership and Equivalence Queries, using the Minimally Adequate Teacher model of Angluin. This is an extension of the congruence based methods for learning some Context-Free…

2009-01-COLT

Mon, 01 Jun 2009 00:00:00 +0000

2009-01-AAMAS-2

Fri, 01 May 2009 00:00:00 +0000

2009-03-paper-064

Fri, 20 Mar 2009 00:00:00 +0000

2009-02-paper-065

Sun, 01 Feb 2009 00:00:00 +0000

2009-01-IEICE-Trans-Inf-Syst

Thu, 01 Jan 2009 00:00:00 +0000

Access full-text academic articles: J-STAGE is an online platform for Japanese academic journals.

2009-01-Theor-Comput-Sci

Thu, 01 Jan 2009 00:00:00 +0000

Abstract could not be automatically retrieved from the linked source. Please add the official abstract manually if available.

2008-01-RoboCup

Tue, 01 Jul 2008 00:00:00 +0000

This paper describes an interactive experimental environment for autonomous soccer robots, which is a soccer field augmented by utilizing camera input and projector output. This environment, in a sense, plays an intermediate role between simulated environments and…

2008-03-paper-070

Sat, 01 Mar 2008 00:00:00 +0000

2007-01-Discovery-Science

Mon, 01 Oct 2007 00:00:00 +0000

In this paper, we propose a new concept, thinning-out, for reducing the number of trials in skill discovery. Thinning-out means to skip over such trials that are unlikely to improve discovering results, in the same way as “pruning” in a search tree. We…

JSAI-SIG-2007

Tue, 01 May 2007 00:00:00 +0000

2006-01-RoboCup

Sun, 01 Jan 2006 00:00:00 +0000

This paper describes an autonomous learning method used with real robots in order to acquire ball trapping skills in the four-legged robot league. These skills involve stopping and controlling an oncoming ball and are essential to passing a ball to each other. We…

2004-01-ALT

Fri, 01 Oct 2004 00:00:00 +0000

Finding a good pattern which discriminates one set of strings from the other set is a critical task in knowledge discovery. In this paper, we review a series of our works concerning with the string pattern discovery. It includes theoretical analyses of learnabilities…

2004-01-Discovery-Science

Fri, 01 Oct 2004 00:00:00 +0000

We consider the problem of discovering the optimal pair of substring patterns with bounded distance α, from a given set S of strings. We study two kinds of pattern classes, one is in form…

2003-01-Discovery-Science

Wed, 01 Jan 2003 00:00:00 +0000

The classificatory power of a pattern is measured by how well it separates two given sets of strings. This paper gives practical algorithms to find the fixed/variable-length-don’t-care pattern (FVLDC pattern) and approximate FVLDC pattern which are most…

2003-01-Theor-Comput-Sci

Wed, 01 Jan 2003 00:00:00 +0000

We study the problem of learning nearly $(s,\epsilon)$-sparse unitaries, meaning that the Pauli spectrum is concentrated on at most $s$ components with at most $\epsilon$ residual mass in Pauli $\ell_1$-norm. This class generalizes well-studied families, including sparse unitaries, quantum $k$-juntas, $2^k$-Pauli dimensional channels, and compositions of depth $O(\log\log n)$ circuits with near-Clifford circuits. Given query access to an unknown nearly sparse unitary $U$, our goal is to efficiently (both in time and query complexity) construct a quantum channel that is close in diamond distance to $U$. We design a learning algorithm achieving this guarantee using $\tilde{O}(s^6/\epsilon^4)$ forward queries to $U$, and running time polynomial in relevant parameters. A key contribution is an efficient quantum algorithm that, given query access to an arbitrary unknown unitary $U$, estimates all Pauli coefficients (up to a shared global phase) whose magnitude exceeds a given threshold $\theta$, extending existing sparse recovery techniques to general unitaries. We also study the broader class of unitaries with bounded Pauli $\ell_1$-norm. For that class, we prove an exponential query lower bound $\Omega(2^{n/2})$. We introduce a more relaxed accuracy metric which is the diamond distance restricted to a set of input states. Then, we show that, under this metric, unitaries with Pauli $\ell_1$-norm uniformly bounded by $L_1$ are learnable with $\tilde{O}(L_1^8/\epsilon^{16})$.

2002-01-Discovery-Science

Fri, 01 Nov 2002 00:00:00 +0000

A variable-length-don’t-care pattern (VLDC pattern) is an element of set Π = (∑∪{⋆}), where ∑ is an alphabet and ⋆ is a wildcard matching any string in ∑. Given two sets of strings, we consider the problem of finding…

2002-01-Progress-in-Discovery-Science

Tue, 01 Jan 2002 00:00:00 +0000

Finding a pattern which separates two sets is a critical task in discovery. Given two sets of strings, consider the problem to find a subsequence that is common to one set but never appears in the other set. The problem is known to be NP-complete. Episode pattern is…

2001-01-Discovery-Science

Thu, 01 Nov 2001 00:00:00 +0000

The class of pattern languages was introduced by Angluin (1980), and a lot of studies have been undertaken on it from the theoretical viewpoint of learnabilities. However, there have been few practical studies except for the one by Shinohara (1982), in which patterns…

2001-01-Discovery-Science-2

Thu, 01 Nov 2001 00:00:00 +0000

Episode pattern is a generalized concept of subsequence pattern where the length of substring containing the subsequence is bounded. Given two sets of strings, consider an optimization problem to find a best episode pattern that is common to one set but not common in…

2000-01-New-Gener-Comput

Sat, 01 Jan 2000 00:00:00 +0000

An elementary formal system (EFS) is a logic program consisting of definite clauses whose arguments have patterns instead of first-order terms. We investig

1999-03-paper-091

Mon, 01 Mar 1999 00:00:00 +0000

1999-01-Discovery-Science

Fri, 01 Jan 1999 00:00:00 +0000

JSPS1993

Thu, 01 Oct 1998 00:00:00 +0000

EuroCOLT97

Tue, 18 Mar 1997 00:00:00 +0000

A pattern is a finite string of constant and variable symbols. For k≥1, we denote by kμΠ the set of all patterns in which each variable symbol occurs at most k times. In particular, we abbreviate μΠ for k=1. The language L(π) of a…

1995-01-ISMB

Sat, 01 Jul 1995 00:00:00 +0000

1995-01-Machine-Intelligence-15

Sun, 01 Jan 1995 00:00:00 +0000

1994-01-AII-ALT

Sat, 01 Oct 1994 00:00:00 +0000

We propose a notion of the refutably PAC learning, which formalizes the refutability of hypothesis spaces in the PAC learning model. Intuitively, the refutably PAC learning for a concept class F requires that the learning algorithm should refute F with high…

1994-01-ECML

Fri, 01 Apr 1994 00:00:00 +0000

In the PAC-learning model, the Vapnik-Chervonenkis (VC) dimension plays the key role to estimate the polynomial-sample learnability of a class of binary functions. For a class of {0,…, m}-valued functions, the notion has been generalized in various ways. This paper…

1994-01-HICSS-5

Sat, 01 Jan 1994 00:00:00 +0000

1993-07-bonsai

Sun, 11 Jul 1993 00:00:00 +0000

1993-01-ALT

Fri, 01 Jan 1993 00:00:00 +0000

The Vapnik-Chervonenkis (VC) dimension is known to be the crucial measure of the polynomial-sample learnability in the PAC-learning model. This paper investigates the complexity of computing VC-dimension of a concept class over a finite learning domain. We consider a…

1993-01-New-Gener-Comput

Fri, 01 Jan 1993 00:00:00 +0000

This paper describes a machine learning system that discovered a “negative motif”, in transmembrane domain identification from amino acid seque

1991-07-pac-mat

Thu, 11 Jul 1991 00:00:00 +0000

1991-07-efs

Mon, 01 Jul 1991 00:00:00 +0000

1991-02-paper-116

Thu, 21 Feb 1991 00:00:00 +0000

1991-01-New-Gener-Comput

Tue, 01 Jan 1991 00:00:00 +0000

This paper considers computational learning from the view-point of teaching. We introduce a notion of teachability with which we establish a relationship b

1991-01-Nonmonotonic-and-Inductive-Logic

Tue, 01 Jan 1991 00:00:00 +0000

Elementary formal system (EFS for short) is a kind of logic program directly dealing with character strings. In 1989, we proposed the class of variable-bounded EFS’s as a unifying framework for language learning. Responding to the proposal, several works have…

1990-07-jsai-alt

Mon, 23 Jul 1990 00:00:00 +0000