FleXOR: Trainable Fractional Quantization
This addresses the problem of restricted compression-accuracy trade-offs in quantization for neural network deployment, offering a novel approach for efficient model storage and inference.
The paper tackles the limitation of integer-bit quantization by proposing FleXOR, a method that achieves fractional bits per weight through encryption and XOR-gate networks, enabling higher compression and accuracy than binary neural networks on datasets like MNIST, CIFAR-10, and ImageNet.
Quantization based on the binary codes is gaining attention because each quantized bit can be directly utilized for computations without dequantization using look-up tables. Previous attempts, however, only allow for integer numbers of quantization bits, which ends up restricting the search space for compression ratio and accuracy. In this paper, we propose an encryption algorithm/architecture to compress quantized weights so as to achieve fractional numbers of bits per weight. Decryption during inference is implemented by digital XOR-gate networks added into the neural network model while XOR gates are described by utilizing $\tanh(x)$ for backward propagation to enable gradient calculations. We perform experiments using MNIST, CIFAR-10, and ImageNet to show that inserting XOR gates learns quantization/encrypted bit decisions through training and obtains high accuracy even for fractional sub 1-bit weights. As a result, our proposed method yields smaller size and higher model accuracy compared to binary neural networks.