Stochastic Shared Embeddings: Data-driven Regularization of Embedding Layers
arXiv:1905.10630
Abstract
In deep neural nets, lower level embedding layers account for a large portion of the total number of parameters. Tikhonov regularization, graph-based regularization, and hard parameter sharing are approaches that introduce explicit biases into training in a hope to reduce statistical complexity. Alternatively, we propose stochastically shared embeddings (SSE), a data-driven approach to regularizing embedding layers, which stochastically transitions between embeddings during stochastic gradient descent (SGD). Because SSE integrates seamlessly with existing SGD algorithms, it can be used with only minor modifications when training large scale neural networks. We develop two versions of SSE: SSE-Graph using knowledge graphs of embeddings; SSE-SE using no prior information. We provide theoretical guarantees for our method and show its empirical effectiveness on 6 distinct tasks, from simple neural networks with one hidden layer in recommender systems, to the transformer and BERT in natural languages. We find that when used along with widely-used regularization methods such as weight decay and dropout, our proposed SSE can further reduce overfitting, which often leads to more favorable generalization results.
Accepted to 2019 Conference on Neural Information Processing Systems
References in corpus (7)
- Adam: A Method for Stochastic Optimization
- Distributed Representations of Words and Phrases and their Compositionality
- Improving neural networks by preventing co-adaptation of feature detectors
- BPR: Bayesian Personalized Ranking from Implicit Feedback
- On the difficulty of training Recurrent Neural Networks
- Self-Attentive Sequential Recommendation
- SQL-Rank: A Listwise Approach to Collaborative Ranking
Cited by in corpus (4)
- CARCA: Context and Attribute-Aware Next-Item Recommendation via Cross-Attention
- A social context-aware graph-based multimodal attentive learning framework for disaster content classification during emergencies: a benchmark dataset and method
- Proxy-based Item Representation for Attribute and Context-aware Recommendation
- Temporal Collaborative Ranking Via Personalized Transformer