**Topic Models**

Probabilistic Topic Models. EM-algorithm and its application to topic modelling. Additive regularization topic model (ARTM). Probabilistic Latent Semantic Analysis (PLSA). Latent Dirichlet Allocation (LDA). Metrics for topic modelling.

**Ensemble methods, part 1**

Ensembles of classifiers. Simple voting, weighted voting, mixture of experts. AdaBoost, its learning guarantees. Gradient Boosting. Choosing basic algorithms, parameter tuning. XGBoost. Other boosting variants: GentleBoost, LogitBoost, BrownBoost

**Ensemble methods, part 2**

Bagging. Random subspace method. Random Forest.

Bias-variance decomposition of error. Why do boosting and bagging work?

Mixture of Experts algorithms.

Bias-variance decomposition of error. Why do boosting and bagging work?

Mixture of Experts algorithms.

**Learning to Rank**

Problem setting. Real case examples. Ranking features: PageRank, TF-IDF. Ranking metrics: Precision, MAP, AUC, DCG, NDCG, pFound. Point-wise, Pair-wise and List-wise ranking strategies. Ordinal classification SVM. Ranking SVM. RankNet. LambdaRank. Ranking in Yandex.

**Recommender Systems**

Problem setting. Practical cases. Collaborative Filtering. Memory-based models and Latent models. User-based and item-based strategies. Metrics and loss functions. RMSE. Using matrix decompositions. NNMF. Implicit ALS. Factorization Machines.

**Neural Networks, part 1**

Multilayer Neural Networks. Stochastic Gradient Descent and its modifications (a recap). BackPropagation. Speeding up convergence. Optimal Brain Damage.

**Neural Networks, part 2**

Vanishing gradient problem. Exploding gradient problem. Strategies of weight initialization. Batch Normalization. Dropout.

Convolutional Neural Networks. Pooling. Applications to Computer Vision.

Recurrent Layers. GRU and LSTM. Gradient Clipping.

Convolutional Neural Networks. Pooling. Applications to Computer Vision.

Recurrent Layers. GRU and LSTM. Gradient Clipping.

**Reinforced Learning**

Multi-armed bandits. Practical case examples. Greedy and ε-greedy strategies. Softmax strategy. Upper confidence bound method.

General problem setting. Environment, agent, observations and actions. On-policy and off-policy learning. Exploration and exploitation. Value function and Q-function. SARSA. Q-learning.

General problem setting. Environment, agent, observations and actions. On-policy and off-policy learning. Exploration and exploitation. Value function and Q-function. SARSA. Q-learning.

**Unsupervised Learning**

Clustering. Evaluation. Statistical setup. EM-algorithm for clustering. K-means and its modifications. Hierarchical clustering.

**Active Learning**

Problem setting. Object selection strategies: pool-based sampling, query synthesis, selective sampling, uncertainty sampling, query by committee, version space reduction, expected model change, expected error reduction, variance reduction, density-weighted methods. Model evaluation. Exploration and exploitation. ε-active algorithm. Exponential gradient. EG-active. Using RL.Thompson sampling.

**Recap**