surrogate-neuroevolution

Distributed Bayesian Optimization - NeuroEvolution

Code

We have three different sets of problems:

• First set includes Iris, Cancer and Chess problems : pso_distributed.

  • Run the file [pso_dist.py] using [run_pso.sh] for DSNE versions.
  • Run the file [surr_revamp_syncswap.py] using [run_surr_revamp_syncswap.sh] for surrogate version- BONE.
  • Run the file [surr_sch.py] using [run_surr_sch.sh] for surrogate version- BONE*.

• Second set features the MNIST problem using CNN : pso_cnn.

  • Run the file [pso_cnn.py] using [run_pso_cnn.sh] for DSNE versions.
  • Run the file [surr_sampled_cnn.py] using [run_surr_sampled_cnn.sh] for surrogate version- BONE.
  • Run the file [surr_cnn_sch.py] using [run_surr_cnn_sch.sh] for surrogate version- BONE*.

• Third set features the Time-Series problem : pso_time_series.

  • Run the file [pso_timeseries.py] using [run_pso_timeseries.sh] for DSNE versions.
  • Run the file [surr_pso_timeseries.py] using [run_surr_pso_timeseries.sh] for surrogate version- BONE.
  • Run the file [surr_pso_ts_sch.py] using [run_surr_pso_ts_sch.sh] for surrogate version- BONE*.

Data

The Data used in Experiments can be found here: DATA

Prerequisites

Installation of libraries such as Keras, Tensorflow and scikitlearn is required for surrogate training, Pytorch is required for running the experiments for MNIST and Time-Series problems.

Experiments

Sample results for all the problems can be found here: results. The files named "final.txt" report the final results including mean and standard deviation for different versions for different problems.