[Python] Improve Validation Accuracy (Smoothing the Curve)

Any Comments Please, how can I improve my training process. I'm using a batch size of 8. Notice: Training accuracy and test accuracy are acceptable for my problem.

# Train the model
history = model3.fit(
XTrain_reshaped,
YTrain_one_hot,
epochs=200,
batch_size=8,
validation_split=0.2)

## XTrain_reshaped -> 499*62*800*1 (499 records, 62 channels, 800 timesteps)
## YTrain_one_hot -> 499*3


I'm using a deep convolution network to classify eeg signals into three classes.

Continue reading...