Quantum-Based Algorithm for Optimizing Artificial Neural Networks

Abstract

This paper presents a quantum-based algorithm for evolving artificial neural networks (ANNs). The aim is to design an ANN with few connections and high classification performance by simultaneously optimizing the network structure and the connection weights. Unlike most previous studies, the proposed algorithm uses quantum bit representation to codify the network. As a result, the connectivity bits do not indicate the actual links but the probability of the existence of the connections, thus alleviating mapping problems and reducing the risk of throwing away a potential candidate. In addition, in the proposed model, each weight space is decomposed into subspaces in terms of quantum bits. Thus, the algorithm performs a region by region exploration, and evolves gradually to find promising subspaces for further exploitation. This is helpful to provide a set of appropriate weights when evolving the network structure and to alleviate the noisy fitness evaluation problem. The proposed model is tested on four benchmark problems, namely breast cancer and iris, heart, and diabetes problems. The experimental results show that the proposed algorithm can produce compact ANN structures with good generalization ability compared to other algorithms.

Multiobjective Cognitive Cooperative Jamming Decision-Making Method Based on Tabu Search-Artificial Bee Colony Algorithm

“…In recent years, intelligent algorithms, such as genetic algorithm (GA) [15,16], improved ant colony (IAC) algorithm [17,18], and particle swarm optimization (PSO) [19], are widely used to solve the jamming decision-making problem. Among them, the IAC algorithm is the most widely used because it can greatly improve the optimization speed and reduce the running time to achieve the best decisionmaking.…”

“…Literature [15] used a genetic algorithm to solve multi-UAV (unmanned aerial vehicle) mission planning problems. Literature [16] proposed a quantum genetic algorithm that uses a qubit to encode communication parameters. It used quantum rotation gates to update the population and assign nonlinear characteristics to the algorithm.…”

Neural Network Approach for Automatic Landuse Classification of Satellite Images: One-Against-Rest and Multi-Class Classifiers

“…Generally, the final network is found through a trial-and-error procedure [6] [7] and depends on user experience as well as needs intensive human interaction and computational time [7]. The same approach has been applied in this research work to find the number of hidden layers and number of nodes in each of them.…”

Optimization of Deep Neural Networks Using SoCs with OpenCL

“…Niu et al [ 9 ] applied improved particle swarm optimization using optimal foraging theory (PSOOFT) to train the free parameters (weights and bias) of an ANN and used a binary PSO algorithm to evolve the network architecture. Lu et al [ 10 ] proposed a quantum-bit representation with which to codify a network, indicating not the actual links, but the probability of existence of the connections, thereby alleviating mapping problems and reducing the risk of discarding a potential candidate. Garro et al [ 11 ] presented a methodology for automatically designing an ANN using three variants of particle swarm optimization algorithms (PSO, second generation PSO (SGPSO) and Nelder–Mead PSO (NMPSO)) to evolve the three principal components of an ANN: the set of synaptic weights, the connections or architecture and the transfer functions for each neuron.…”