Machine Learning

Automatic Speech Recognition

A coursework from my master's degree. We experimented with classic HMM-GMM methods for speech recognition. We later expanded our model by using a DNNs for phone classification, which improves the word error rate.

Learning Algorithms and Regularization

A coursework from my master's degree. I experimented with the adaptive learning algorithms SGD, RMSProp and Adam on training a DNN for classification. I also looked into the effects of using a cosine annealing learning rate scheduler.

Optimising The Boids Algorithm With Unsupervised Learning

I investigated the famous Boids algorithm and increased its performance by using unsupervised learning.