Articles

• char-based Joke LSTM: Ongoing experiments in building a char based LSTM model on the OpenAI jokes Corpus. Medium article here.Tried various methods and heuristics for text generation (greedy prediction, beam search etc)
• A first look at Pytorch 1.0 - My experience at the Pytorch Decelopers Conference, and a discussion of new developments in Pytorch 1.0.
• Matrix Calculus for Deep Learning - A collection of notes based on ‘The Matrix Calculus You Need For Deep Learning’ by Terence Parr and Jeremy Howard.
• Is the human wearing glasses? - Example of training a classifier to recognize if the human is wearing glasses or not, using resnet34, and a dataset with <200 images
• Fun with small image data-sets - Demonstration of how pre-trained image models with a custom head can be used to quickly prototype classifiers, even with small dataset sizes.
• Image Data Pre-Processing for Neural Networks - An overview of common pre-processing steps taken to prepare image datasets for computer vision based ML projects
• Netflix for Genes - Collaborative filtering model to help identifying candidate biomarkers for pRCC cancer. One of 3 winning teams in SVAI cancer research hackathon competition. PPT