Transferring Inductive Biases through Knowledge Distillation
This addresses the challenge of efficiently adapting inductive biases for tasks with limited data or computational resources, though it is incremental as it builds on existing knowledge distillation methods.
The paper investigates using knowledge distillation to transfer inductive biases from models like LSTMs and CNNs to others such as Transformers and MLPs, showing that this approach can effectively transfer biases and improve model performance in data-limited or mismatched train-test scenarios.
Having the right inductive biases can be crucial in many tasks or scenarios where data or computing resources are a limiting factor, or where training data is not perfectly representative of the conditions at test time. However, defining, designing and efficiently adapting inductive biases is not necessarily straightforward. In this paper, we explore the power of knowledge distillation for transferring the effect of inductive biases from one model to another. We consider families of models with different inductive biases, LSTMs vs. Transformers and CNNs vs. MLPs, in the context of tasks and scenarios where having the right inductive biases is critical. We study the effect of inductive biases on the solutions the models converge to and investigate how and to what extent the effect of inductive biases is transferred through knowledge distillation, in terms of not only performance but also different aspects of converged solutions.