Learning ML

handwritten digits gradually turning into other numbers Plot showing training set on first two principal components a digit's bottom half being predicted by the model Digits showing learned distribution and samples taken on each digit

In this project, completed for STA414 during the winter semester of 2020, I learned how to train a model to classify some digits in the MNIST dataset, and to make predictions, samples, and interpolations from the latent space using the Julia programming language. The beauty of this very clean dataset lured me into a very messy space.

Learning to Read

visualization of ten lines of latin verse Fallon's permutations Vergil's syntax lines

Picture Vergil — Georgics 2.483-492

In this project, I attempted to represent ten lines (2.483-492) from Vergil’s famous farming poem visually, comparing syntactical, diction, and metrical (i.e. sound) qualities between the original and three increasingly modern translations (respectively by John Dryden, Cecil Day Lewis, and Peter Fallon). In this passage, the poet describes an idyllic landscape and laments the difficulty of really occupying the space. The same difficulty can be seen in the challenges that too literal-minded a visualization method results in. Read the original poem here. Learn more about my project here.

Learning to Draw

Inktober

ink drawing of park seen from a short distance away ink drawing of crows on a field under city sky ink drawing of a cherry tree ink drawing of a swing set ink drawing of a tree ink drawing of a park