Metaheuristics in the optimization of Laguerre-Volterra networks

The Laguerre-Volterra network (LVN) is a Volterra-equivalent connectionist architecture, which combines a bank of discrete Laguerre filters and a layer of polynomial activation functions. This architecture is designed to model nonlinear dynamic systems from input-output signals.

Here we optimize the continuous parameters of the LVN to model synthetic systems using different metaheuristics, with the purpose of performance evaluation.

Third party software versions

• Python 3.6.9
  • NumPy 1.17.3 (vector math)
  • Scipy 1.3.0 (Friedman significance test)
  • scikit-posthocs 0.6.1 (Nemenyi post-hoc significance test)
  • Matplotlib 3.0.3 (plotting)

List of modules

• base_metaheuristic.py
  • simulated_annealing.py
  • particle_swarm_optimization.py
  • ant_colony_for_continuous_domains.py
• laguerre_volterra_network_structure.py
• optimization_utilities.py
• simulated_systems.py
• data_handling.py

Scripts and their uses

• generate_datasets.py - Uses the data_handling module to generate synthetic train and test IO signals from simulated systems.
• optimize_LVN.py - Optimizes LVNs with arbitrary structure using different metaheuristics (mostly used for verification)
• results_collection.py - Runs some specified metaheuristic 30 times and stores the solutions found, along with their errors on test signals
• results_stats.py - With the results from 'results_collection.py', compute averages and standard deviations for train and test errors
• results_stats_significance.py - Compute the statistical significance of the results with the Friedman and Nemenyi tests
• plotting scripts

If this repository is valuable to you, consider citing:

Costa, V. O. and Müller, M. F. (2020). "Evaluation of Metaheuristics in the Optimization of Laguerre-Volterra Networks for Nonlinear Dynamic System Identification". 9th Brazilian Conference on Intelligent Systems, BRACIS (2020).