Scientific machine learning for knowledge discovery
For centuries, the process of formulating new knowledge from observations has driven scientific discoveries. Moreover, the incorporation of that knowledge has led to several practical applications. With rapid advancements in machine learning, it is natural to question the possibility of automating k...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/163595 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | For centuries, the process of formulating new knowledge from observations has driven scientific discoveries. Moreover, the incorporation of that knowledge has led to several practical applications. With rapid advancements in machine learning, it is natural to question the possibility of automating knowledge discovery and incorporation in the scientific field. A benchmark task for automated knowledge discovery is called symbolic regression. The task aims to predict a mathematical equation that best describes the observational data. On the other hand, known governing equations are incorporated into the learning process of deep neural networks to predict the system’s behaviour. The advancements in both these tasks have significant potential to aid research in understanding and predicting unexplored systems’ dynamics and properties. Although the application of machine learning for knowledge discovery and incorporation has rich and active research, there is a lack of substantial research on integrating both tasks. Therefore, the aim of this paper are twofold. First, a comprehensive literature survey of existing algorithms for knowledge discovery and incorporation. Second, integrating the knowledge discovery and incorporation methods as a single process to predict the solution of Burger’s system. In addition, conduct a comparative study to understand the strengths and limitations of each method. And develop a web-based dashboard to interact with the knowledge discovery algorithms. Finally, highlighting the key challenges in the current methods and future research directions in using machine learning to automate scientific knowledge discovery and incorporation. |
|---|