AI Training

Day 1: Introduction to Machine Learning with Python
Morning Session: Python Fundamentals

• Lecture: Basics, NumPy, Pandas
• Workshop: Cleaning and preprocessing genomic data

Afternoon Session: ML Fundamentals

• Lecture: Supervised/unsupervised learning, model evaluation
• Workshop: Building ML models with scikit-learn

Day 2: Data Visualization and Scikit-Learn
Morning Session: Data Visualization

• Lecture: Matplotlib, Seaborn, PCA, t-SNE
• Workshop: Visualizing genomic datasets

Afternoon Session: Mastering Scikit-Learn

• Lecture: Feature engineering, pipelines, hyperparameter tuning
• Workshop: Optimizing ML pipelines for sequence analysis

Day 3: Introduction to Neural Networks
Morning Session: Neural Network Basics

• Lecture: Neurons, activation functions, Keras/TensorFlow
• Workshop: Building a simple neural network

Afternoon Session: Deep Learning Best Practices

• Lecture: Overfitting, regularization, optimizers
• Workshop: Improving and regularizing deep learning models

Day 4: Convolutional Neural Networks (CNNs)
Morning Session: CNN Concepts

• Lecture: CNN architectures, transfer learning
• Workshop: Advanced CNN techniques

Afternoon Session: Deep Visualization

• Lecture: Data augmentation, model explainability
• Workshop: Visualizing model decisions

Day 5: Variational Autoencoders (VAEs) and Beyond
Morning Session: VAEs

• Lecture: VAE concepts and applications
• Workshop: Implementing a VAE

Afternoon Session: Future Concepts

• Lecture: Emerging trends, self-supervised learning
• Workshop: Project brainstorming and discussion of future projects

These intensive half-day tutorials are designed to equip post-docs with AI skills directly relevant to their research.

Tutorial 1: Explainability and Fairness in Machine Learning (XAI)

This tutorial provides a deep dive into techniques for understanding, interpreting, and ensuring fairness in complex AI models, especially those considered "black boxes."

• Model Interpretability: Global and local explanation methods (feature importance, LIME, SHAP, etc.)
• Visualizing Model Behavior: Techniques to understand internal representations and decision-making processes of AI models, particularly deep learning architectures.
• Fairness in AI: Definitions of fairness (group, individual), metrics for measuring bias, and strategies for bias mitigation (pre-processing, in-processing, post-processing techniques).
• Case Studies: Exploration of real-world scenarios where XAI techniques have been critical in ensuring fairness and trustworthiness across different research applications.

Tutorial 2: Generative AI and Natural Language Processing (NLP)

In-depth look at novel architectures that power Large Language Models, to understand their role and impact in AI research.

• Transformer Architectures: Understanding self-attention mechanisms and their role in NLP.
• Text Generation: Text summarization, writing assistance, question-answering systems, dialogue generation.
• Machine Translation: Advancements in neural machine translation, challenges in handling low-resource languages.
• Fine-tuning Generative Models: Tailoring language models to specific domains/tasks to improve performance and control generation.

Day 1: Machine Learning Basics

• Morning:
  • Lecture: Introduction to Machine Learning, including supervised and unsupervised learning.
  • Workshop: Hands-on practice with basic algorithms, such as regression and classification.
• Afternoon:
  • Lecture: Model evaluation techniques and validation strategies.
  • Workshop: Practical exercises on model tuning and evaluation.

Day 2: Application of ML Technologies to Bioscience

• Morning:
  • Lecture: Review of existing machine learning algorithms, including SVMs, decision trees, and ensemble methods, and their general applications.
  • Workshop: Hands-on coding session with these algorithms, followed by a discussion on how they can be applied to bioscience problems.
• Afternoon:
  • Lecture: Introduction to unsupervised learning methods like clustering and dimensionality reduction, with an emphasis on their relevance to gene expression analysis or other biological data.
  • Workshop: Applying unsupervised learning techniques to a public bioscience dataset.

Day 3: Deep Learning

• Morning:
  • Lecture: Introduction to deep learning and neural networks.
  • Workshop: Hands-on experience with building simple deep learning models.
• Afternoon:
  • Lecture: Overview of cutting-edge models like GPT and their applications.
  • Workshop: Experimentation with pre-trained GPT models and customization for specific tasks.

Day 1 (Optional)

• Lecture 1: Basics of NNs
• Workshop 1: Implementing a Simple NN
• Lecture 2: Regularization and Hyperparameter Tuning
• Workshop 2: Fine-tuning a NN

Day 2

• Lecture 1: CNNs
• Workshop 1: Building a CNN for image classification
• Lecture 2: RNNs and LSTMs
• Workshop 2: Implementing an RNN for text generation

Day 3

• Lecture 1: Advanced Deep Learning
• Workshop 1: Implementing advanced techniques
• Lecture 2: NLP and LLMs
• Workshop 2: Using LLMs for NLP tasks