Thoughts on neural networks "discovering" physical concepts

In summary: They introduce a new network structure, called SciNet, and use unsupervised representation learning techniques to encourage the network to find a minimal representation of experimental data without any prior assumptions. They believe that this could lead to a better understanding of physical laws without relying on previous knowledge. However, their approach might not be as reliable as traditional methods and may be more useful in areas such as molecular biology rather than physics.
  • #1
sphyrch
37
9
(sorry i post this here now because i didn't know about this sub forum earlier)

I came across an interesting paper from which I'll quote parts of the intro:
[...] the physical theories we know may not necessarily be the simplest ones to explain the experimental data, but rather the ones that most naturally followed from a previous theory at the time. The formalism of quantum theory, for instance, is built upon classical mechanics; it has been impressively successful, but leads to conceptually challenging consequences [...]

This raises an interesting question: are the laws of quantum physics, and other physical theories more generally, the most natural ones to explain data from experiments if we assume no prior knowledge of physics? [...] we investigate whether neural networks can be used to discover physical concepts in classical and quantum mechanics from experimental data, without imposing prior assumptions and restrictions on the space of possible concepts.
and the conclusion:
[...] we have shown that neural networks can be used to recover physical variables from experimental data. To do so, we have introduced a new network structure, SciNet, and employed techniques from unsupervised representation learning to encourage the network to and a minimal uncorrelated representation of experimental data. [...] Furthermore, the analogy between the process of reasoning of a physicist and representation learning provides insight about ways to formalize physical reasoning without adding prior knowledge about the system.
obviously the authors are more competent in Physics than I am - so are most PF member. what are your opinions on this? can the work of these authors "discover" physical laws in any way or is the paper title misleading?

would stuff like special relativity or quantum mechanics be discovered purely from experimental data without using prior knowledge? even if physical laws were encoded by these networks as "black box" models, would they be a good replacement of closed-form expressions or differential equation solutions?
 
  • Like
Likes FactChecker and lomidrevo
Technology news on Phys.org
  • #2
I am not sure what the authors are seeking to achieve. It seems to me that some very good neural networks were actively engaged in developing quantum theory.

Experience shows that human neural networks are good at creating mathematical models of the physical world. It will likely be even better with the assistance of fast computational machines that can process large volumes of data and find patterns and relationships. Molecular biology rather than physics might be a better area to focus on

AM
 

Related to Thoughts on neural networks "discovering" physical concepts

What is the purpose of "discovering" physical concepts through neural networks?

The purpose of using neural networks to "discover" physical concepts is to gain a deeper understanding of complex physical phenomena that may be difficult for humans to comprehend or predict. By training neural networks on large datasets, we can potentially uncover patterns and relationships in the data that can help us better understand the underlying physical principles at play.

How accurate are the physical concepts discovered by neural networks?

The accuracy of the physical concepts discovered by neural networks depends on the quality and quantity of the data used for training. If the dataset is diverse and comprehensive, the neural network may be able to accurately uncover underlying physical concepts. However, if the dataset is limited or biased, the discovered concepts may not be entirely accurate or representative of the true physical phenomena.

What are the potential applications of using neural networks to discover physical concepts?

The applications of using neural networks to discover physical concepts are vast and varied. They can range from improving our understanding of complex systems, such as weather patterns, to developing more accurate and efficient models for predicting physical phenomena. This knowledge can also be applied to various industries, such as transportation, energy, and healthcare, to optimize processes and improve outcomes.

What are the limitations of using neural networks for discovering physical concepts?

One of the main limitations of using neural networks for discovering physical concepts is that they require large amounts of high-quality data for training. This can be a challenge in fields where obtaining data is difficult, such as in astrophysics or particle physics. Additionally, neural networks may not always uncover the most accurate or complete physical concepts, as they are limited by the data they are trained on and may not be able to capture all the complexities of a given system.

How can we ensure the ethical use of neural networks for discovering physical concepts?

As with any emerging technology, it is important to consider the ethical implications of using neural networks to discover physical concepts. This includes being transparent about the data used for training, addressing potential biases in the data, and ensuring that the insights gained from these networks are used for the benefit of society. It is also crucial to continuously monitor and evaluate the impact of these discoveries to ensure they are not being used in a harmful or discriminatory manner.

Similar threads

Replies
6
Views
840
  • Programming and Computer Science
Replies
2
Views
986
  • Programming and Computer Science
Replies
1
Views
765
  • Computing and Technology
Replies
4
Views
2K
  • Engineering and Comp Sci Homework Help
Replies
1
Views
688
  • Programming and Computer Science
Replies
15
Views
1K
  • Programming and Computer Science
Replies
7
Views
1K
  • STEM Academic Advising
Replies
4
Views
2K
  • STEM Academic Advising
Replies
6
Views
1K
  • Quantum Interpretations and Foundations
5
Replies
169
Views
8K
Back
Top