🐱artificial neural network project
Jupyter Notebook Mathematica Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
.ipynb_checkpoints
MMA
ipynb
notes
py
.gitignore
LICENSE
README.md

README.md

ANN - Artificial Neural Network Method

Artificial neural network for differential equation solving.

Cotents

I gave a talk about this idea a while ago. A Physicist's Crash Course on Artificial Neural Network

We did not use the simple back-prop method for the project because it's aweful. (I do not have the original code for it but I rewrote an example using PyTorch.) We really need much better cost minimization method. So we tested the best minimization algorithms here.

Structure of this repository:

.
β”œβ”€β”€ LICENSE
β”œβ”€β”€ MMA
β”‚Β Β  β”œβ”€β”€ homogeneousGas.nb
β”‚Β Β  └── vac.nb
β”œβ”€β”€ README.md
β”œβ”€β”€ ipynb
β”‚Β Β  β”œβ”€β”€ Basics.ipynb
β”‚Β Β  β”œβ”€β”€ Basics.ipynb.bak
β”‚Β Β  β”œβ”€β”€ HomogeneousModel.ipynb
β”‚Β Β  β”œβ”€β”€ NetworkConstructor.ipynb
β”‚Β Β  β”œβ”€β”€ Untitled.ipynb
β”‚Β Β  β”œβ”€β”€ Untitled1.ipynb
β”‚Β Β  β”œβ”€β”€ ann_julia.ipynb
β”‚Β Β  β”œβ”€β”€ assets
β”‚Β Β  β”œβ”€β”€ test.ipynb
β”‚Β Β  β”œβ”€β”€ vacOsc4Comp.ipynb
β”‚Β Β  β”œβ”€β”€ vacOsc4CompSSConvention.ipynb
β”‚Β Β  β”œβ”€β”€ vacOsc4Fourier.ipynb
β”‚Β Β  β”œβ”€β”€ vacOsc4Piecewise.ipynb
β”‚Β Β  β”œβ”€β”€ vacuum-Copy1.ipynb
β”‚Β Β  β”œβ”€β”€ vacuum-Copy2.ipynb
β”‚Β Β  β”œβ”€β”€ vacuum.ipynb
β”‚Β Β  β”œβ”€β”€ vacuum4Component.ipynb
β”‚Β Β  └── vacuumClean.ipynb
β”œβ”€β”€ notes
β”‚Β Β  └── note-2015S.pdf
└── py
    β”œβ”€β”€ functionvalue-moretol.txt
    β”œβ”€β”€ functionvalue.txt
    β”œβ”€β”€ ss
    β”œβ”€β”€ timespent-moretol.txt
    β”œβ”€β”€ timespent.txt
    β”œβ”€β”€ vacOsc4CompSSConvention-moretol.py
    β”œβ”€β”€ vacOsc4CompSSConvention-verify.py
    β”œβ”€β”€ xresult-1.txt
    β”œβ”€β”€ xresult-moretol.txt
    └── xresult.txt
  1. notes is the notes for the project. I explained some of the conventions and the preliminary results. I pulled this file from my private repo of the project. I think it can made public now.
  2. The folder MMA is for my Mathematica code related to this problem.
  3. ipynb contains the Jupyter Notebooks.
    1. Basics.ipynb: the basics of the idea. quite similar to the talk mentioned above.
    2. HomogeneousModel.ipynb: solving Homogeneous gas model of neutrino oscillations.
    3. NetworkConstructor.ipynb: example of network constructor for differential equations.
    4. ann_julia.ipynb: Julia code example.
    5. test.ipynb: testing different methods, benchmarking functions.
    6. vacOsc4Comp.ipynb: Solving neutrino vacuum oscillations.
    7. vacOsc4CompSSConvention.ipynb: vacuum oscillations using Shanshak's convention
    8. vacOsc4Fourier.ipynb: Using Fourier as the internal network structure, aka, Fourier analysis as approximators.
    9. vacOsc4Piecewise.ipynb: Using piecewise functions as approximators
    10. vacuumClean.ipynb: Vacuum oscillations cleaned up
    11. vacuum4Component.ipynb: Vacuum oscillations with 4-component conventions
  4. py folder is for the python code.