/NN_evolution_with_PSO

Demonstrating how the optimal architecture for a FFNN can found using PSO algorithm

Primary LanguageJupyter Notebook

NN_evolution_with_PSO

Demonstrating how the optimal architecture for a FFNN to predict earthquake magnitude can found using PSO algorithm. Uses earthquake data from: https://www.kaggle.com/usgs/earthquake-database/data

Implemented using three classes:

  1. NN_model(): Abstracts a Feed forward NN model.
  2. PSO_NN_Particle(): Abstracts PSO particle.
  3. PSO_NN_Swarm(): Abstracts the PSO swarm.

Implementation details:

  1. NN built using Keras with TensorFLow backend.
  2. tf.data API used for feeding data into keras model.