Giampietro Tecchiolli

The continuous reactive tabu search: Blending combinatorial optimization and stochastic search for global optimization

A novel algorithm for the global optimization of functions (C-RTS) is presented, in which a combinatorial optimization method cooperates with a stochastic local minimizer. The combinatorial optimization component, based on the Reactive Tabu Search recently proposed by the authors, locates the most promising “boxes”, in which starting points for the local minimizer are generated. In order to cover a wide spectrum of possible applications without user intervention, the method is designed with adaptive mechanisms: the box size is adapted to the local structure of the function to be optimized, the search parameters are adapted to obtain a proper balance of diversification and intensification. The algorithm is compared with some existing algorithms, and the experimental results are presented for a variety of benchmark tasks.

Totem:. a Highly Parallel Chip for Triggering Applications with Inductive Learning Based on the Reactive Tabu Search

The training of a Multi-Layer Perceptron (MLP) classifier is considered as a Combinatorial Optimization task and solved using the Reactive Tabu Search (RTS) method. RTS needs only forward passes (no derivatives) and does not require high precision network parameters. TOTEM, a special-purpose VLSI chip, was developed to take advantage of the limited memory and processing requirements of RTS: the final system effects a very close match between hardware and training algorithm. The RTS algorithm and the design of TOTEM are discussed, together with the operational characteristics of the VLSI chip and some preliminary training and generalization tests on triggering tasks.

Advances in the design of the TOTEM neurochip

Abstract The TOTEM neurochip has proved its viability as a system for real-time computation in HEP and space applications requiring high performance for event classification, data mining, and signal processing. ISA and VME boards integrating the TOTEM chip as a coprocessor have been made available to selected experimental groups which reported satisfactory results. This paper presents a new architectural solution yielding higher performance and reduced silicon area. The on-chip computational structures have been entirely redesigned to take advantage of a novel approach to number representation that, at the cost of a provably bounded approximation, leads to a much-reduced silicon area, lower power dissipation, and faster computation. This approach is validated by simulation results on experimental data, as presented in the paper.

A parallel processor for neural networks

A deeply-pipelined digital parallel processor for the implementation of Multi-Layer Perceptrons is presented. It employs high-speed limited-precision integer arithmetic and allows good recognition performance in combination with a novel training algorithm. Internal dynamic RAM is provided for storage of the weights. The chip achieves a performance of 600 million multiply-and-accumulate operations per second and requires a silicon area of 70 mm/sup 2/ in a 1.2-/spl mu/m CMOS technology.

On the problem of stability for higher-order derivative Lagrangian systems

The problem of stability for dynamical systems whose Lagrangian function depends on the derivatives of a higher order than one is studied. The difficulty of this analysis arises from the indefiniteness of the Hamiltonian, so that the well-known Lagrange-Dirichlet theorem cannot be used and the methods of the canonical perturbation theory (KAM theory) must be employed. We show, with an example, that the indefiniteness of the energy does not forbid the stability.

Vector Quantization with the Reactive Tabu Search

A novel application of the Reactive Tabu Search to Vector Quantization (RTS-VQ) is presented. The results obtained on benchmark tasks demonstrate that a performance similar to that of traditional techniques can be obtained even if the code vectors are represented with small integers. The result is of interest for application-specific VLSI circuits.

Stochastic quantization of linearized gravitational theories

Abstract A generalization of stochastic quantization is proposed in order to deal with higher order derivative gravitational theories. In the linear approximation the correct graviton propagator is recovered, while for the higher order derivative case, the resulting propagator has pole structure, which exhibits the impossibility of Wick rotation. The unitarity of the models is briefly discussed.