improved_DeepLOB

This repo contains my reimplementation and improvement of DeepLOB model. The original paper is "DeepLOB - Deep Convolutional Neural Networks".

I modified the model by using Dilated Convolution which could raise the accuracy by around 2.5% accuracy with only less than 1,000 parameters added. The model structure I use is depicted as follows:

The training process is visualized as follows: