Meta-Learning-LR

Using a meta learning approach to predict the optimal learning rate for a network

Getting Started

Setup

For a quick setup follow the next steps:

conda create -n meta python=3.10.4

conda activate meta

git clone https://github.com/Shaier/Meta-Learning-LR.git

cd Meta-Learning-LR

pip install -r requirements.txt

Running the code

There are 2 main files:

1. simple neural network file which can be run using "python simple_neural_network.py" from the terminal. This is just a simple FC neural network. To get different saved predictions change the learning rate on line 60.
2. meta training file, which can be run by going into the file and running it (it's a jupyter notebook). This is the file that trains a meta learning model to output an optimal learning rate for the simple NN.