Diploma thesis - Neural Network Application in Physical Problems

In summary, the speaker is a student studying Computational Physics and is interested in Machine Learning and Neural Networks. They are looking for ideas for their diploma thesis that can combine their interest in neural networks and their physics background. They mention the possibility of using neural networks for spectral analysis of galaxies or quasars, or for clustering particle collisions at CERN. They also mention wanting to build and train their own neural network using Monte Carlo data and evaluating it on real data. They ask for guidance or any useful links. The conversation also brings up the field of symbolic regression, where ML techniques are used to discover equations for physical systems or data. The speaker mentions a project by Cornell that became a commercial product and links to more information on the topic.
  • #1
dirac26
2
0
Hello everybody. I am currently on my last year of Computational Physics education. More and more I am interested in Machine Learning and Neural Network. Time has come for me to choose diploma work thesis, so I am searching for interesting ideas where I can merge my interest in neural networks and my background in physics.

I am looking for something like spectral analysis of galaxies or quasars using neural networks. Or maybe, use ML algorithm for clustering particle collision at CERN(divide ggH and VBF for example).
In a nutshell, I want to build my neural network, feed it with some Monte Carlo data, train it, and then evaluate it on real data.

Do you have some guidance, some ideas, useful links, or anything that can help.
Thank you in advance, I really appreciate it.
 
Last edited by a moderator:
Physics news on Phys.org
  • #2
There's a field called symbolic regression where ML techniques are used to discover the equations describing a physical system or collection of data.

Cornell had done a project that eventually became the commercial product Eureqa. The 2009 Cornell project discovers the equations of motion for a compound pendulum system using data collected about position and time of the moving pendulum bob. Wired magazine did an article on it and they published a couple of papers on the scheme.

https://en.wikipedia.org/wiki/Symbolic_regression
 
  • Like
Likes berkeman

1. What is a neural network and how does it work?

A neural network is a type of artificial intelligence that is modeled after the structure and function of the human brain. It is composed of interconnected nodes or neurons that process information and make decisions based on input data. The network learns and improves through a process called training, where it adjusts the connections between neurons to better predict outcomes.

2. How is a neural network applied in physical problems?

In physical problems, a neural network can be used to model complex systems and make predictions based on input data. For example, it can be used to predict the behavior of a physical system, such as the movement of particles in a fluid, based on variables such as temperature, pressure, and velocity. The network is trained on a large dataset of known outcomes and can then make predictions for new input data.

3. What are the advantages of using a neural network in physical problems?

One advantage of using a neural network is its ability to handle complex and nonlinear relationships between variables. In physical problems, there are often many variables that affect the outcome, and a neural network can capture these relationships and make accurate predictions. Additionally, neural networks can learn and adapt to new data, making them useful for solving problems with changing conditions.

4. Are there any limitations to using a neural network in physical problems?

While neural networks have many benefits, they also have some limitations. One limitation is the need for a large amount of training data to accurately model a system. Additionally, the network may struggle with extrapolating beyond the range of the training data, so it is important to have a diverse and representative dataset. Finally, the network's decisions may not always be explainable, which can be a challenge for some applications.

5. What are some real-world applications of using a neural network in physical problems?

There are many real-world applications of using a neural network in physical problems. For example, it can be used to predict weather patterns, optimize energy usage in buildings, and improve the efficiency of industrial processes. It is also commonly used in fields such as physics, chemistry, and engineering to model complex systems and make predictions for various scenarios.

Similar threads

Python Why backpropagation dominates neural networks instead of interpolation
    • Programming and Computer Science
Replies
4
Views
637
Why is my multilayered, feedforward neural network not working?
    • Programming and Computer Science
2
Replies
50
Views
3K
Admissions Comments on my physics PhD applicant profile
    • STEM Academic Advising
Replies
4
Views
818
Thoughts on neural networks "discovering" physical concepts
    • Mechanics
Replies
6
Views
741
Thoughts on neural networks "discovering" physical concepts
    • Programming and Computer Science
Replies
1
Views
918
Python Understanding Graph Neural Networks (GNNs) Tutorials & Examples
    • Programming and Computer Science
Replies
2
Views
1K
Neural networks - uses in Physics and Astrophysics?
    • Programming and Computer Science
Replies
22
Views
3K
Courses Which of these Machine Learning courses should I attend?
    • STEM Academic Advising
Replies
5
Views
853
Studying Self Study roadmap for Machine Learning
    • STEM Academic Advising
Replies
4
Views
1K
Need help coming up with ideas (bachelor's thesis)
    • STEM Academic Advising
Replies
5
Views
942

Hot Threads

Recent Insights

Back
Top