* Design Philosophy
We would like to do something similar to what CONEX does.
We build a recursive neural network which essentially tiles to find the function f in,
y(x+dx) = f (y(x)). which would be the solution to the cascade equation. We will call f the stepping function from now.
* Short term goal (Couple of Weeks)
- CONEX doesnt write the intermediate steps to a file.
We generate data by making CONEX write the intermediate steps to a file in a purely electromagnetic cascade.
- We then train the neural network with these intermediate steps and hope the neural network can find the stepping function.
- We chose to do this with CONEX, because EM Cascade is relatively simple and we can get the data faster than using C8.
* Mid Term Goal (A month or so)
- If we are able to get the neural network to step similar to CONEX, then we are on the right track.
- Second goal, is to code a piping setup which takes slices from an actual shower from C8 and generates the source functions.
- Once we have the data and the source functions, we can redo the RNN with actual C8 data this time.
- Sparcity
* Long Term Goal
** There are 2 ways we can go here, both of it requires design.
*** Method 1 (3D showers)
- Think and modify the source function (which is specialized for 1D) to be able to be applicable for 3D.
*** Method 2 (More complicated Showers)
- Add additional interactions into C8 and check if the RNN is able to pickup the stepping function.
- We would need to modify the network so that we can introduce new stuff in between.
