Author name cluster

Deepak Mehta 0001

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

1 author row

ECAI Conference 2014 Conference Paper

A Decomposition Approach for Discovering Discriminative Motifs in a Sequence Database

David Lesaint
Deepak Mehta 0001
Barry O'Sullivan
Vincent Vigneron

This paper addresses the discovery of discriminative nary motifs in databases of labeled sequences. We consider databases made up of positive and negative sequences and define a motif as a set of patterns embedded in all positive sequences and subject to alignment constraints. We formulate constraints to eliminate redundant motifs and present a general constraint optimization framework to compute motifs that are exclusive to the positive sequences. We cast the discovery of closed and replication-free motifs in this framework and propose a two-stage approach whose last stage reduces to a minimum set covering problem. Experiments on protein sequence datasets demonstrate its efficiency.

Details

SoCS Conference 2013 Conference Paper

Evolving Instance Specific Algorithm Configuration

Yuri Malitsky
Deepak Mehta 0001
Barry O'Sullivan

Combinatorial problems are ubiquitous in artificial intelligence and related areas. While there has been a significant amount of research into the design and implementation of solvers for combinatorial problems, it is well-known that there is still no single solver that performs best across a broad set of problem types and domains. This has motivated the development of portfolios of solvers. A portfolio typically comprises either many different solvers, instances of the same solver tuned in different ways, or some combination of these. However, current approaches to portfolio design take a static view of the process. Specifically, the design of the portfolio is determined offline, and then deployed in some setting. In this paper we propose an approach to evolving the portfolio over time based on the problems instances that it encounters. We study several challenges raised by such a dynamic approach, such as how to re-tune the portfolio over time. Our empirical results demonstrate that our evolving portfolio approach significantly out-performed the standard static approach in the case when the type of instances observed change over time.

Details

ECAI Conference 2010 Conference Paper

Improving the Global Constraint SoftPrec

David Lesaint
Deepak Mehta 0001
Barry O'Sullivan
Luis Quesada 0001
Nic Wilson

A soft global constraint SOFTPREC has been proposed recently for solving optimisation problems involving precedence relations. In this paper we present new pruning rules for this global constraint. We introduce a pruning rule that improves propagation from the objective variable to the decision variables, which is believed to be harder to achieve. We further introduce a pruning rule based on linear programming, and thereby make SOFTPREC a hybrid of constraint programming and linear programming. We present results demonstrating the efficiency of the pruning rules.

Details

ECAI Conference 2008 Conference Paper

A BDD Approach to the Feature Subscription Problem

Tarik Hadzic
David Lesaint
Deepak Mehta 0001
Barry O'Sullivan
Luis Quesada 0001
Nic Wilson

Modern feature-rich telecommunications services offer significant opportunities to human users. To make these services more usable, facilitating personalisation is very important since it enhances the users' experience considerably. However, regardless how service providers organise their catalogues of features, they cannot achieve complete configurability due to the existence of feature interactions. Distributed Feature Composition (DFC) provides a comprehensive methodology, underpinned by a formal architecture model to address this issue. In this paper we present an approach based on using Binary Decision Diagrams (BDD) to find optimal reconfigurations of features when a user's preferences violate the technical constraints defined by a set of DFC rules. In particular, we propose hybridizing constraint programming and standard BDD compilation techniques in order to scale the construction of a BDD for larger size catalogues. Our approach outperforms the standard BDD techniques by reducing the memory requirements by as much as five orders-of-magnitude and compiles the catalogues for which the standard techniques ran out of memory.

Details