Deep Mutual Learning Optimizes Multi-Task Recommender Systems with Cross Task Feature Mining

Table of Links

Abstract and 1 Introduction

2. Methodology
3. Experiments
4. Conclusion and References

2. METHODOLOGY

In this section, we first introduce the problem of multi-objective ranking for recommender systems. Second, we describe the general design of DML. Finally, we elaborate on the introduced components.

2.1 Multi-Objective Ranking for Recsys

2.2 Overall Design of DML

Please refer to figure 1 for the overall network architecture of DML. With the existing MTL algorithms[25], the equation (1) can be further decomposed as below. For simplicity, we omit the subscript 𝑛 in this section.

In this research, rather than 𝐺, we focus on the enhancement of upper-level networks for improved prediction performance. First, the shared component of CTFM is introduced, which leverages the attention mechanism to extract relevant information from the inputs of other task towers (the results of Equation (3)) as a complement to the target task. Please note that this attention is welldesigned to solve the task-awareness missing issue, for which the excessive encouragement of knowledge sharing is not conducive to the extraction of task-specific knowledge. With our design, the gradients computed from the target task’s loss will not impact the inputs of other task towers.

2.3 Cross Task Feature Mining