I developed Multi-Layer Neural Network including functions for activation, data preprocessing, training, evaluation, and forward and backward propagation algorithms.
I also utilised the PyTorch library to implement and train a neural network that is capable of predicting house prices in California. Achieving this required me to infer the median house value of a block group from the value of all other provided attributes. The data was provided to me via a csv file containing examples of houses with several attributes of varying data types. Making use of specialised ML Python libraries, I created and fitted a regression-based neural network. I then evaluated the performance of my network by using it to make predictions on a held-out test set. The accuracy achieved by the model was good, with the RMSE of the model being $51,000 - significantly lower than the specified $90,000 constraint.
I generated a Decision Tree using Python and performed evaluation and pruning – designed to locate the position of a device using signal strength data from multiple Wi-Fi emitters