Quantization Aware Factorization for Deep Neural Network Compression