register
data-science
data-science
- From Chaos to Signal Fifteen thousand JavaScript components in NEAR BOS, with little agreement on what any of them did. Raw code and noisy tags defeated the first classifiers. Staged compression and human selection made the collection searchable.
- Signal Meets Structure A recommender for NEAR's decentralised social graph. Embeddings and second-degree connections made profiles easier to find; sparse network structure set the limit on what the system could recommend.
- Twitter on Health Social Media Sentiment Analysis. We conducted an analysis of emotional responses to prominent diseases on Twitter in the post-pandemic era.
- APTOS Retinopathy Detection A Kaggle experiment classifying retinal images into five grades of diabetic retinopathy. DenseNet and mixup handle a dataset with uneven image quality; quadratic weighted kappa measures agreement with the labels.
- Blockchain Approach to Medical Data Ownership A blockchain prototype for auctioning access to anonymised medical data. BioDataHero started with the patient as seller; its successor Sicherlich moved toward keeping the record in the patient's own vault.
- Classifiying Gene Mutations Classifying gene mutations from medical literature in a Kaggle challenge. TF-IDF, word embeddings, and several classifiers were tested. The word clouds were easier to look at than to learn from.