Neural Networks: Backprop Hyperparameters
Budget: $20 – $50 USD
This project will require you to (i) make modifications to a NNDL code file (tuning various hyperparameters) and (ii) write a small application code.
The original NNDL code file hard-codes several things, including the sigmoid as the activation function, L2 regularization and the input data format (for MNIST). For the application code, you will do some systematic experiments to test various combinations of the hyper-parameter values.
The original NNDL code file hard-codes several things, including the sigmoid as the activation function, L2 regularization and the input data format (for MNIST). For the application code, you will do some systematic experiments to test various combinations of the hyper-parameter values.