Genetic-Algorithm-for-feature-subset-selection

Implementation of Genetic Algorithm for finding the best subset of features for emotion detection from an image using the CK+ dataset and Openface toolkit.

Dataset extraction

• This model was trained on The Extended Cohn-Kanade Dataset (CK+) which is a complete dataset for action unit and emotion-specified expression
• The images from the dataset were then processed using the OpenFace toolkit to obtain the final dataset
• Manually included labels to the dataset in order to train the model for the respective emotion

Features implemented

• Random initialization of population chromosomes
• Fitness for the individuals (feature subset) was defined using the mean accuracy for classifying emotions
• Selection methods:
  1. Roulette-wheel selection
  2. Rank-based selection
  3. Tournament selection
• Crossover methods:
  1. k-point crossover
  2. Uniform crossover
• Mutation methods:
  1. Bit-swap mutation
  2. Bit-flip mutation
• Population update methods:
  1. Generational update
  2. Weak-parents update

Models used for training

• Implemented SVM and Logistic Regression from the sklearn Python library
• Implemented Neural Network using Keras
• SVM was found to give the best accuracy with the fastest convergence