Summary of Layer-wise Regularized Dropout For Neural Language Models, by Shiwen Ni et al.
Layer-wise Regularized Dropout for Neural Language Models
by Shiwen Ni, Min Yang, Ruifeng Xu, Chengming Li, Xiping Hu
First submitted to arxiv on: 26 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel regularization technique called Layer-wise Regularized Dropout (LR-Drop) for Transformer-based language models. The goal is to address the inconsistency between training and inference caused by dropout’s randomness. LR-Drop layer-wise regularizes each Transformer layer using consistency training, forcing hidden states, attention matrices, and output distributions of siamese sub-models to be consistent. Through experiments on 15 natural language understanding, machine translation, and summarization datasets, LR-Drop achieves state-of-the-art performances. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to help language models work better. It’s called Layer-wise Regularized Dropout (LR-Drop). The problem is that some things used in training don’t happen when we use the model later. LR-Drop fixes this by making sure each part of the model works together consistently. The researchers tested it on many kinds of tasks, like understanding language and translating texts, and it did really well. |
Keywords
» Artificial intelligence » Attention » Dropout » Inference » Language understanding » Regularization » Summarization » Transformer » Translation
