Neural Networks in Optics

Repository for applications of neural networks in optics.

---

Tested with Python3.8 and packages listed in requirements.txt file

MNIST Diffusion Experiments

The mnist_ directories contain scripts for training a denoising diffusion model on the MNIST dataset (handwritten digits). These scripts are illustrative, with a focus on the overall structure and workflow rather than achieving optimal performance.

Running the Code

Simple Model

python model/mnist_diffusion/train_model_cnn.py \
  configs/mnist_diffusion/config_unet_simple.yaml \
  data_directory results_directory -c

Simple Model

python model/mnist_diffusion/train_model_cnn.py \
  configs/mnist_diffusion/config_unet_advanced.yaml \
  data_directory results_directory -c

---

Some experiments exploring different activation functions for a fully connected neural network—designed to predict the mean and standard deviation of 10 variables sampled from a standard normal distribution ( N(0, 1) ) - can be found in experiments/fully_connected_mean_std.py.