evolutionary-network-arc-search

find a good neural network architecture by an evolutionary algorithm

the project goal is to find a neural network architecture that fits cifar10 data the best

• evolutionary algorithm should search among
  • different feature extractors
  • different layers count
  • different neuron count for each layer
  • different activation function for each layer such as ReLU, sigmoid, linear
• used pytorch to extract feature
• used tensorflow to evaluate each model

Installation

1. Create a virtual environment for the project:

python -m venv venv

2. Activate the virtual environment:

• On Windows:

venv\Scripts\activate

• On macOS and Linux:

source venv/bin/activate

3. Install the required dependencies:

pip install -r requirements.txt

4. run:

python main.py