Tasks given By Prof. Nipun Batra
clone the repository.
All the files except Neural Network for MNIST dataset.ipynb will run directly in google colab. For Neural Network for MNIST dataset.ipynb, extract the contents from mnist_train.csv.zip to the same directory where Neural Network for MNIST dataset.ipynb is present and run the code.
Run the code Bivariate Gaussian Visualizer.ipynb.
Use the widgets to change mean and Covariance to visualize bivariate guassian for yourself.
Multivariate Gaussian Sampling.ipynb is for drawing samples from MVN distribution. It can generate Samples from MVN for any dimensions.
3. Implement two hidden layers neural network classifier from scratch in JAX
Implemented neural network to classify the data Got a test set accuray of 91.50% on the test set by training the neural network.
Generated synthetic 1d linear dataset with added noise.
Used Baysian linear regression to get the distribution for the Parameters using
- analytical solution
- Sampling using BlackJAX.
plotted the results.