virtuallyaverage/sin-approximation

find the minimum size network to approximate a sin, cos, or tan value.

sin-approximation

A small weekend project for estimating the sin of a value between 0-2pi radians. Mainly for learning the basics of publishing to github, managing versions, and learn the basics of maching learning. This is a quick and dirty project, so there are clear optimizations, bugs, and improvements to be made. But it has succeded in the goals of furthering my knowlege of ML, Github, PyTorch, Database creation, and python in general.

Setup on windows

• Install torch pip install torch then restart powershell (open a new window)
• Clone repository cd install/location git clone https://github.com/virtuallyaverage/sin-approximation.git
• Navigate to repository and run file cd install/location python get_a_sin.py

Models

You can use a different model from the models folder by putting -m "model name here" in front of the run command. An example using a more accurate, but less predictable model: python get_a_sin.py -m "model_(9.790426247491268e-07).mdl"

• model_(2.7785529255197616e-06).mdl (default) with a mean deviation of 0.00249rad and an mean error of 0.3074% it is not the lowest deviation, but it is the lowest error percentage. This discrepency suggests that it is more consistant across the evaluated values, which I decided would be best as the default. Even if it is off by a greater amount some times

• model_(9.790426247491268e-07).mdl (experimental) This is a model that has been trained over 10 hours, while previous model was trained in 30 minutes. With a mean deviation of .00069rad and an mean error of 1.142% it has the lowest deviation, but loses out in the mean error. This discrepency suggests that it is less consistant across the evaluated values but more accurate overall, making it a decent model, but possibly with unexpected results on occasion.