Improving Knowledge Distillation in Transfer Learning with Layer-wise Learning Rates
This work addresses performance degradation in transfer learning for complex tasks, offering an incremental improvement over existing methods.
The paper tackles the problem of poor performance in transfer learning when task complexity increases by proposing a layer-wise learning rate scheme based on differences in Jacobian/Attention/Hessian of output activations, resulting in improved learning performance and stability across datasets, with gains becoming more significant as task difficulty rises.
Transfer learning methods start performing poorly when the complexity of the learning task is increased. Most of these methods calculate the cumulative differences of all the matched features and then use them to back-propagate that loss through all the layers. Contrary to these methods, in this work, we propose a novel layer-wise learning scheme that adjusts learning parameters per layer as a function of the differences in the Jacobian/Attention/Hessian of the output activations w.r.t. the network parameters. We applied this novel scheme for attention map-based and derivative-based (first and second order) transfer learning methods. We received improved learning performance and stability against a wide range of datasets. From extensive experimental evaluation, we observed that the performance boost achieved by our method becomes more significant with the increasing difficulty of the learning task.