Looking for physics based simulation for data generation

Click For Summary

Discussion Overview

The discussion revolves around the search for physics-based simulations that can generate data for training physics-based neural networks, particularly focusing on the integration of physics-based loss functions. Participants explore the need for tools that can provide data resembling real experimental results without being overly complex.

Discussion Character

  • Exploratory
  • Technical explanation
  • Debate/contested

Main Points Raised

  • One participant is seeking online tools for physics-based simulations to generate data for proof of concept (POC) in their neural network research.
  • Another participant suggests writing a Python program to generate test data by applying theoretical models and adding noise for realism.
  • Several participants inquire about the specifics of the datasets needed, including the number of channels and precision required.
  • Some participants propose alternative data sources, such as lottery data, weather data, or stock/bond price trading data, as potential datasets for analysis.
  • A link to a related thread is provided, which may contain relevant information for the original poster's research.

Areas of Agreement / Disagreement

Participants express varying degrees of understanding regarding the requirements for the data generation, with some seeking clarification on the specifics of the datasets while others suggest different approaches or data sources. The discussion does not reach a consensus on a specific simulation tool or method.

Contextual Notes

The discussion lacks detailed specifications regarding the desired characteristics of the simulation tools, such as the complexity of the simulations or the exact nature of the data required. There are also unresolved questions about the precision and structure of the datasets needed.

Who May Find This Useful

Researchers and practitioners in the fields of physics, machine learning, and data science who are interested in generating experimental-like data for neural network training may find this discussion relevant.

up2front
Messages
2
Reaction score
1
Good day,

I am working on physics based neural networks and I am investigating the role of integrating physics based loss function on the network's training.
So I am looking for physics based simulation for data generation (online tools) to generate data from experiments.

I know I may generate data using the explicit laws on excel but I am looking for relatively close to real experiments simulation. Please note that I am not looking for complicated simulation I just need data for POC.

So if anyone aware of such simulation software or tool please let me know.
 
Technology news on Phys.org
Welcome to PhysicsForums. :smile:

What kind of datasets? How many channels? How many bits of precision for each channel?
 
I would write a program in python to generate your test data.

use the theory to generate expected data and then use the math.rand function to add noise to the data.

we used this approach to create FFT test data that the FFT function could analyze and generate the frequency spectrum.
 
  • Like
Likes   Reactions: up2front
up2front said:
Good day,

I am working on physics based neural networks and I am investigating the role of integrating physics based loss function on the network's training.
So I am looking for physics based simulation for data generation (online tools) to generate data from experiments.

I know I may generate data using the explicit laws on excel but I am looking for relatively close to real experiments simulation. Please note that I am not looking for complicated simulation I just need data for POC.

So if anyone aware of such simulation software or tool please let me know.
Not exactly what you are looking for (means I don't know how complicate an implementation is), but it fits to your research: https://www.physicsforums.com/threa...ving-electronic-schroedinger-equation.997917/
 
  • Like
Likes   Reactions: up2front
berkeman said:
Welcome to PhysicsForums. :smile:

What kind of datasets? How many channels? How many bits of precision for each channel?
Well, any data set with enough variables that can be modeled into equation or loss function
 
up2front said:
Well, any data set with enough variables that can be modeled into equation or loss function
Sorry, please answer my questions explicitly and technically so that we can try to help you. Thank you.
 
You could always look for lottery data or weather data. Weather in particular could be interesting to analyze.

Or even stock/bond price trading data.
 

Similar threads

Computational Physics Simulations
  • · Replies 1 ·
Replies
1
Views
3K
Undergrad Using AI for Gravitational Wave Generation
  • · Replies 3 ·
Replies
3
Views
1K
C# Questions on locating and characterizing facial features in C#
  • · Replies 1 ·
Replies
1
Views
2K
Thoughts on neural networks "discovering" physical concepts
  • · Replies 1 ·
Replies
1
Views
2K
Understanding Grids and Units in Computer Graphics and Physics Simulations
  • · Replies 48 ·
2
Replies
48
Views
5K
How Do I Calculate Actual Frequency from Wind Turbine Simulation Data?
  • · Replies 11 ·
Replies
11
Views
2K
ABC reference frame synchronous generator modelling in Ltspice
  • · Replies 1 ·
Replies
1
Views
3K
Fixing Error: (vlog-7) Failed to Open Design Unit File
  • · Replies 4 ·
Replies
4
Views
7K
Getting Started with Application Development
  • · Replies 6 ·
Replies
6
Views
3K
I need help please for Simulation of Xenon Oscillations in a Nuclear Reactor
  • · Replies 4 ·
Replies
4
Views
2K