In this repository I have added my machine learning and deep learning homeworks, that can help you get a comprehension of my skills and my knowledges in these topics. Most of the notebooks were originally made by me in Russian and translated with DeepL after. Don't be surprised by the choice of words.
You can find classical machine learning notebooks and notebooks with basic tools/libraries in Machine Learning folder. Kind of tasks: gradient descent, random forest, knn, feature engineering, etc.
Stack: Pandas, Numpy, sklearn, matplotlib.
There are deep learning notebooks and notebooks with basic dl libraries in this folder. The first notebook is about basic FC structure (regression task) and Pytorch, the second one is about image classification with my own models (hand-made architecture with BasicBlocks from resnet but adjusted to 64x64 images) and pretrained large models (vision transformers). The third notebook is about text classification with embeddings.
Stack: Pytorch, pytorchLightning, wandb, torchvision.
There is also a report of the university project attached. In this project we built efficient Random Forest models from scratch in java to solve a complex regression problem of predicting students grades. The complexity was due to the diffrence between available features for various students. Theoretical aspects of statistics behind Random Forest is discussed in the report, as well as the most efficient implementation for our specific task.
I have also competed in EEG classification competition on Kaggle with a friend, where we used EfficientNet and spectograms as inputs to solve the problem. We weren't seeking high place in the leaderboard (u can search team named dmitrysakharov in the Leaderboard serchbar), but still the basic pipeline is implemented. https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/overview
Stack: Pytorch, sklearn