Compact Deep Learning: Leveraging Quantization for Smaller Neural Networks