Data Analysis with Deep Learning Methods

Detailed syllabus of lectures:

• Types of data analysis problems. Data preprocessing and machine learning methods in the data analysis process.
• Binary and multi-class classification. Determining decision boundaries. Support vector machines, one-vs-all algorithm, cross-validation.
• Explainable machine learning methods. Lazy learning methods. K-nearest neighbor method.
• Linear models. Linear discriminant analysis (LDA).
• Clustering, k-means (Forgy method), k-means++ algorithm. Internal and external evaluation indicators, cluster analysis. Hierarchical clustering. Density-based clustering (e.g., DBSCAN).
• The role of dimensionality reduction in data analysis. Dimensionality reduction methods.
• Deep learning, deep neural networks (DNN). Activation and loss functions. Training neural networks, backpropagation. Regularization against overfitting (early stopping, dropout, batch normalization). Optimization procedures, hyperparameter optimization.
• DNN types: Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Bidirectional RNN (BRNN).
• Data analysis on structured and unstructured corporate data. Text datasets. Classification and segmentation of corporate data.
• Specificities of time series data. Time series analysis. Analysis of economic data, forecasting methods.
• Time series analysis with deep learning methods. Sequence-to-sequence methods.
• Multimedia analysis, deep learning for audio data classification.
• Video analysis. Object tracking in videos. Tracking-by-detection approach. Simple IOU-based tracking, position estimation procedures, Kalman filter. Multiple Object Tracking (MOT). Simple Online and Realtime Tracking (SORT) and DeepSORT algorithm.
• Classification with limited data. Zero-shot, one-shot, few-shot learning, Siamese networks. Deception methods. Face spoofing, deep fake.

Detailed syllabus of practices/labs: During the semester, the practices take place in a Python environment, which always relate to the lecture material and serve to deepen the knowledge.

• Python basics (ipython notebook, numpy, scikit-learn). Data preprocessing steps in practice.
• Model validation and feature engineering. Comparison of KNN and SVM.
• K-means and hierarchical clustering on corporate data. Determining the number of clusters.
• Dimensionality reduction in practice.
• Neural network implementation.
• Deep learning frameworks. Implementing convolutional layers.
• Python OpenCV library.
• Classification with linear models.
• Time series analysis. Analysis of macroeconomic data.
• Analysis of text data.
• Speech recognition.
• Analysis of traffic data.
• Tracking vehicles using IOU tracker and SORT, based on YOLO.
• Presentation and discussion of homework assignments.