Various Deep Learning Projects (2022- 2023)
This repo demonstrates multiple Deep Learning Techniques including:
- introduction to Deep learning: Solving digit recognition problem by building a from scratch neural network and loss function.
- Deep dive into neural network structures, loss functions, backpropagation, and Training loops.
- Building self-made neural network for Computer vision tasks, solving image recognition problem on CIFAR dataset.
- Using pretrained model on ImageNet dataset for image classification and Dogs Vs Cats dataset.
- Perform Natural Language Processing using Encoder-Decoder structure to estimate the salary based on the job title.
- Introducing CBOW and Word2vec neural networks to solve embeddings tasks.
- Deep dive into Transformers pre-trained models.
- Deep models interpretability using SHapley and smooth grad
- Large Language Models and Their Implications, Prompt engineering, LLM fine-tuning.
- Introduction to Denoising Autoencoders and Image retrieval problem.
- Development of GAN (Generative Adversarial Networks) for human faces generation tasks.
- Speech processing pipeline using Deep learning.
Skills developed: pandas | pytorch | tensorflow | numpy | GANs | encoder-decoder | model-finetuning | LLM | Prompt engineering | models interpretability | SHapley | Transformers | HuggingFace | genism | nltk | python.
This repo is part of the DL course, HSE, Moscow, Russia.