Research Article
Ensemble Convolution Neural Network for Robust Video Emotion Recognition Using Deep Semantics
| Data: Training Set | | | | BC: Number of base classifiers | | SR: Ratio of samples that need replacement | | : Parameter used to reduce the distance between training and synthetic data | | : ith attribute | | : standard deviation | | r: normal distribution’s sampling value N(0, 1) | | Training Phase: | | For a = 1: BC | | Copy the original dataset i.e. Data | | Identify the number of training samples that need replacement, i.e., | | For b = 1: TS | | Randomly pick “z” samples from | | If x is a majority class sample, then | | Generate a neighborhood of z based on and replace z exists in | | Else if | | Check z is a minority sample, then compute m = Round | | Replace m neighbourhoods of z in | | End For | | Build base classifier from | | End For | | Classification Phase: | | For a given z | | evaluate ensemble to classify the sample z based on the majority voting strategy |
|