1,721,014 research outputs found
Terminal Attractor Algorithms and the Class of Unimodal Loading Problems
No full textThe effectiveness of connectionist models in emulating intelligent behaviour and solving significant practical problems is strictly related to the capability of the learning algorithms to find
optimal or near–optimal solutions, and to generalise to new examples. This paper deals with optimal learning and provides a unified viewpoint of most significant results in the field. We briefly review proposals for discovering optimal solutions and give some general guidelines for performing successful optimisation. Most importantly, we show some intriguing links between optimal
learning and the computational complexity of loading problems. We prove that all problems giving rise to unimodal error functions have O(1) as a complexity upper bound, thus suggesting that they belong to the same class, defined on the basis of computational requirements
Optimal Learning and the Class of Unimodal Error Surface Loading Problems
No full textThe effectiveness of connectionist models in emulating intelligent behaviour and solving significant practical problems is strictly related to the capability of the learning algorithms to find optimal or near-optimal solutions, and to generalize to new examples. This paper deals with optimal learning and provides a unified viewpoint of most significant results in the field. In particular, we exhibit a computational model such that the solution of all loading problems giving rise to unimodal error functions require the same time, thus suggesting that they belong to the same computational class
Non−suspiciousness: A Generalisation of Convexity in the Frame of Foundations of Numerical Analysis and Learning
No full textThe effectiveness of connectionist models in emulating intelligent behaviour is strictly related to the capability of the learning algorithms to find optimal or near-optimal solutions. In this paper, a canonical reduction of gradient descent dynamics is proposed, allowing the formulation of the neural network learning as a finite continuous optimisation problem, under some non-suspiciousness conditions. In the linear case, the non-suspect nature of the problem guarantees the implementation of an iterative method with O(n^2) as computational complexity. Finally, since non-suspiciousness is a generalisation of the concept of convexity, it is possible to apply this theory to the resolution of nonlinear problems
A Heuristic Global Optimisation Algorithm for Training Recurrent Neural Networks
No full textRecurrent neural networks are an efficient tool for the solution of problems of automatic speech recognition. On the other hand, using the Reduced Gradient (RG) algorithm for isolated word recognition is known to be better than Backpropagation in locally recurrent architectures. The implementation of RG guarantees the convergence to a Kuhn-Tucker point which, however, is in general neither the global minimum nor a satisfactory sub-optimal solution. In this paper we have experimented a new heuristic method based on a combined use of RG and a deterministic global algorithm for unconstrained optimisation. The preliminary numerical experiences of such method have evidenced very promising results
Toward a unified approach for the classification of NP-complete optimization problems
No full textAbstractTwo notions which have been introduced with the aim of classifying NP-complete optimization problems are compared: the notion of strong NP-completeness, due to Garey and Johnson, and that of simple and rigid problems, due to Paz and Moran. In particular, we show under what conditions reductions preserve rigidity, simplicity, strong simplicity and p-simplicity and we show that under reasonable hypothesis, p-simple problems are solved by pseudo-polynomial algorithms and strong NP-complete problems are weakly rigid
On Learning Monotone DNF Formulae under Uniform Distributions
No full textWe show how to learn in polynomial time monotone d-term DNF formulae (formulae in disjunctive normal form with at most d terms) using positive examples drawn from a distribution that is a generalization of the uniform distribution. © 1994 Academic Press. All rights reserved
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
