Dropping Networks for Transfer Learning

TL;DR

This paper introduces Dropping Networks, a novel ensemble-based transfer learning method combining Dropout and Bagging to improve transferability and reduce negative transfer in natural language understanding tasks, especially in few-shot learning.

Contribution

It proposes a new approach that integrates source Dropping Networks with target tasks, using a decaying parameter based on error curve analysis to enhance transfer learning performance.

Findings

Outperforms hard and soft parameter sharing methods in few-shot learning.

Achieves comparable results to state-of-the-art with less target data.

Mitigates negative transfer effectively.

Abstract

Many tasks in natural language understanding require learning relationships between two sequences for various tasks such as natural language inference, paraphrasing and entailment. These aforementioned tasks are similar in nature, yet they are often modeled individually. Knowledge transfer can be effective for closely related tasks. However, transferring all knowledge, some of which irrelevant for a target task, can lead to sub-optimal results due to \textit{negative} transfer. Hence, this paper focuses on the transferability of both instances and parameters across natural language understanding tasks by proposing an ensemble-based transfer learning method. \newline The primary contribution of this paper is the combination of both \textit{Dropout} and \textit{Bagging} for improved transferability in neural networks, referred to as \textit{Dropping} herein. We present a straightforward yet novel approach for incorporating source \textit{Dropping} Networks to a target task for few-shot learning that mitigates \textit{negative} transfer. This is achieved by using a decaying parameter chosen according to the slope changes of a smoothed spline error curve at sub-intervals during training. We compare the proposed approach against hard parameter sharing and soft parameter sharing transfer methods in the few-shot learning case. We also compare against models that are fully trained on the target task in the standard supervised learning setup. The aforementioned adjustment leads to improved transfer learning performance and comparable results to the current state of the art only using a fraction of the data from the target task.