Summary of Mixture Of Latent Experts Using Tensor Products, by Zhan Su et al.
Mixture of Latent Experts Using Tensor Products
by Zhan Su, Fengran Mo, Prayag Tiwari, Benyou Wang, Jian-Yun Nie, Jakob Grue Simonsen
First submitted to arxiv on: 26 May 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 In this paper, researchers explore ways to improve multi-task learning by minimizing negative transfer between tasks. They propose a novel modular language model called TensorPoly, which balances efficiency and routing methods. The authors also introduce LoRA (Low-Rank Adaptation) reparameterized as TLoRA, using tensor product operations. Two innovative routing functions are designed for fine-grained control: TensorPoly-I directs to each rank within the entangled tensor, while TensorPoly-II targets each order of the entangled tensor. Experimental results on the T0-benchmark show that modular LMs outperform dense approaches and achieve superior outcomes. The authors highlight the potential of their approach in multi-task transfer learning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making machine learning models better at doing multiple tasks at once. Right now, when we train a model to do many things, it can actually get worse if some tasks are similar. To fix this, the authors created a new type of model that breaks down complex tasks into smaller, more manageable parts. They tested their model on a benchmark and found that it performs better than previous models. This is important because it could lead to advancements in areas like natural language processing and computer vision. |
Keywords
» Artificial intelligence » Language model » Lora » Low rank adaptation » Machine learning » Multi task » Natural language processing » Transfer learning
