This week, I spent time refining and testing the CNN model. I employed a series of Conv1D layers with LeakyReLU activations and a final Dense layer for classification. I spent time adjusting the hyperparameters and experimented with various learning rates for the Adam optimizer. Through experimentation, the optimal balance between convergence rate and stability was achieved with the learning rate of 0.001. Also, I experimented with different combinations of columns as input features and decide on a better choice for the combination.
There is a trade-off when we need to balance model complexity and overfitting. I found that larger batch sizes have better training stability, but they require more memory in order to cache intermediate results during training.
Next week, I will perform the final fine-tuning of our model.
