Peter James Winn's Projects
Coding a neural network from scratch using Numpy in Python
Material associated with our PLoS One paper on DeepCDpred. DeepCDpred: Inter-residue distance and contact prediction for improved prediction of protein structure: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0205214 Please see the README.md, or the PlosOne publication for author contributions. I did not write this code, but rather the students in my research group.
Practising different ways of coding the evaluation of the Fibonnaci sequence in Python and benchmarking them. Including recursion with and without memoisation, using a loop and memoisation, and an object based coding solution.
Evolutionary Scale Modeling (esm): Pretrained language models for proteins
A robot powered training repository :robot:
Practicing using Altair to produce .html files with interactive plots. For this example I used cellular location data. More information about Altair here: https://altair-viz.github.io/getting_started/overview.html
Some basic exercises in SQL
Some exercises to learn Keras and Tensor flow. The code here loosely follows the teaching on the free Udacity course accessable from tensorflow.org, and also includes some exercises borrowed from Justin Johnson's lectures on Deep Learning for Vision, delivered at Michigan University, although these latter are originally set in PyTorch so require some adaptation for Keras.
RNN and LSTM models for generating text
microsoft mlops course
Examples of simple PyTorch uses including geometric deep learning in PyG
some basic scripts for using RDKit for manipulating chemical structures.
My clone repository
My clone repository
Some experiments to practice transfer learning
Coding a generative self attention transformer follow Andrej Karpathy's YouTube tutorial