Մեքենայական ուսուցման դասընթաց

Փուլ 1 - Introduction to ML

• Machine Learning history

• Intuition

• Common use cases

Փուլ 2 - Math refresher block

• Logarithms

• Linear algebra refresher

• Calculus refresher

• Prob and stats refresher

• Gradients, chain rule, autodiff

Փուլ 3 - Quiz

• Classical ML block

• Linear regression, Loss functions (MSE), gradient descent

• L1, L2, intro to regularization and feature selection, cross validation

• Logistic regression, locally weighted regression

• Intro to NLP: Naive Bayes, TF-IDF

• Teaching linear models do non-linear stuff (feature cross w linear regression, polynomial regression, SVM, kernel trick)

• Clustering: K means, Spectral, (H)DBSCAN, GMM

• Decision trees, intro to ensemble methods

• Random forest, GBDT

• Dimensionality reduction, PCA, uses with clustering

Փուլ 4 - Midterm

• Deep learning block (taught alongside pytorch)

• Neural network intuition, MLP

• Backprop, better optimizers

• CNN

• Common CNN architectures, their evolution

• Training neural networks, useful tricks (batchnorm, dropout), activation functions

• RNN, vanishing/exploding gradients, truncated backprop

• LSTM, GRU, LSTM + attention

• Autoencoders

• Generative models: VAE, intro to GANs

• Making sense of text: word vectors, word2vec, Glove

• Making sense of text: tokenization, attention is all you need, encoder-only models, decoder-only models

Փուլ 5 - Bonus modules

• Intro to reinforcement learning

• Common problems and how to solve them

• Scaling and deployment (pruning, quantization, FastAPI, docker)

• Interesting research areas: alignment/RLHF, text to image, multimodal, retrieval-oriented learning

Փուլ 6 - Group project

• TBD. Will cover things like version control and working on a shared codebase within a team