Fundamentals of Machine Learning with Python - Advanced Level À distance

Part 1 - Convolutional neural networks:

- The notion of invariance and how to cope with it

- Data augmentation

- Classical deep architecture - the example of Resnet18

Part 2 - Transformers and Learning on sequential data:

- Auto-regressive models 

- The notions of transformers and attention

- Text data: Tokenizers and how to use them

Part 3 - ML pipeline and Model Fine-tuning:

- Efficient baseline models: Random Forest, XGBoost

- Hyperparameter tuning with optuna

- Few shot learning, domain adaptation

- Example on images and/or text

- Finetuning with Low Rank Adaptation (LoRA)

Part 4 - Time series:

- Dynamic Time Warping

- Specific crossvalidation and evaluation procedures

- Clustering time series

- Dictionary learning

Part 5 - Physics-informed Neural Networks:

- Physics-informed loss

- Neural Operators

- Examples on ODEs, PDEs and data assimilation

Part 6 – Generative methods

- Unconditional, class-conditional, and embedding-conditional generative models

- Foundational techniques like normalizing flows

- Diffusion models and flow matching

- Text-to-image models such as Stable Diffusion or DALL·E

The learning of these concepts will be illustrated and put into practice through the handling of the scikit-learn and PyTorch libraries throughout the training.

• For a knowledgeable audience in Machine Learning, introduction and further exploration of some modern artificial intelligence methods
• Implementation of Deep Learning methodologies (and others) on various types of data
• Use of development frameworks such as scikit-learn and PyTorch

• Lectures based on interactive Jupyter notebooks, application exercises.