Summary of Preventing Conflicting Gradients in Neural Marked Temporal Point Processes, by Tanguy Bosser and Souhaib Ben Taieb
Preventing Conflicting Gradients in Neural Marked Temporal Point Processes
by Tanguy Bosser, Souhaib Ben Taieb
First submitted to arxiv on: 11 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 The paper proposes a new approach to learning Neural Marked Temporal Point Processes (MTPP) models that capture complex temporal inter-dependencies between labeled events. The authors frame MTPP learning as a two-task problem, where both tasks share trainable parameters optimized jointly. This leads to conflicting gradients during training, which can be detrimental to task performance. To overcome this issue, the authors introduce novel parametrizations for neural MTPP models that allow separate modeling and training of each task. The framework is evaluated on multiple real-world event sequence datasets, demonstrating improved performance compared to original model formulations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper teaches us about a new way to understand events happening in time. It’s like trying to predict when and what will happen next. The old way was complicated because it tried to do two things at once: predict the timing of events and what type of event it would be. This caused problems during training, making it hard for the model to learn correctly. To fix this, the authors created a new way to train these models that lets them focus on each task separately. They tested their method on real-world data and showed that it works better than the old way. |
