Summary of Tlasdi: Thermodynamics-informed Latent Space Dynamics Identification, by Jun Sur Richard Park et al.
tLaSDI: Thermodynamics-informed latent space dynamics identification
by Jun Sur Richard Park, Siu Wun Cheung, Youngsoo Choi, Yeonjong Shin
First submitted to arxiv on: 9 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Dynamical Systems (math.DS)
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 method called tLaSDI that combines thermodynamics with machine learning techniques. The approach embeds the first and second principles of thermodynamics into a latent space dynamics identification framework, using an autoencoder as a non-linear dimension reduction model. The method also utilizes a neural network-based model to construct the latent dynamics while preserving certain structures for the thermodynamic laws through the GENERIC formalism. An abstract error estimate is established, providing a new loss formulation that involves Jacobian computation of the autoencoder. The autoencoder and latent dynamics are simultaneously trained to minimize this new loss. The paper demonstrates the effectiveness of tLaSDI through computational examples, showing robust generalization ability even in extrapolation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to understand how things change over time using math and computer models. They create a special method called tLaSDI that helps us see patterns in data that follow certain rules from physics. This method uses something called an autoencoder, which is like a special kind of map that helps us find the underlying patterns in the data. The authors also use a neural network to help them understand how these patterns change over time. They show that their method can be very good at making predictions and even works when we’re not sure what’s going to happen next. |
Keywords
* Artificial intelligence * Autoencoder * Generalization * Latent space * Machine learning * Neural network
