Multirate Training of Neural Networks
This addresses computational bottlenecks for researchers and practitioners in vision and NLP by offering a speed-up in transfer learning, though it is incremental as it builds on existing training methods.
The paper tackles the problem of computational inefficiency in neural network training by proposing multirate training, which partitions parameters into fast and slow parts updated at different time scales, resulting in almost half the time for fine-tuning deep networks without reducing generalization performance.
We propose multirate training of neural networks: partitioning neural network parameters into "fast" and "slow" parts which are trained on different time scales, where slow parts are updated less frequently. By choosing appropriate partitionings we can obtain substantial computational speed-up for transfer learning tasks. We show for applications in vision and NLP that we can fine-tune deep neural networks in almost half the time, without reducing the generalization performance of the resulting models. We analyze the convergence properties of our multirate scheme and draw a comparison with vanilla SGD. We also discuss splitting choices for the neural network parameters which could enhance generalization performance when neural networks are trained from scratch. A multirate approach can be used to learn different features present in the data and as a form of regularization. Our paper unlocks the potential of using multirate techniques for neural network training and provides several starting points for future work in this area.