Schedule

Detailed information about the activities

Statistics for data science (Dan Nicolae)

• A data science case study
• Foundations of data analysis
• Statistical inference with resampling methods

Statistical learning (Dan Nicolae)

• Probability and simulations
• Regression models and inference
• Model Complexity
• Prediction and classification

Basics of machine learning (ML) (Razvan Bunescu)

• Feature vector representations
• Occam’s razor for ML, intelligence, and science
• Overfitting, underfitting, generalization, and regularization
• ML experiments: training, validation, and testing

Linear models for classification (Razvan Bunescu)

• Logistic regression, softmax, and temperature
• ML algorithms in Python: the sklearn library
• Linear vs. non-linear classification and deep learning

Applied Deep Learning (Gabriel Terejanu)

• From linear models to neural networks
• Why deep neural networks?
• What is an embedding? 
• How to make use of pre-trained models?

Basics of Large Language Models (LLMs) (Razvan Bunescu)

• Subword tokenization, word embeddings, and neural language models (LMs)
• Encoder, encoder-decoder, and decoder LMs
• Pre-training and fine-tuning

Prompting techniques and LLM APIs (Razvan Bunescu)

• Zero-shot and few-shot in-context learning
• Chain-of-thought prompting, retrieval augmented generation, and ReAct
• The chat completion API and LangChain

Data-driven decision making (Gabriel Terejanu)

• Introduction to A/B testing for decision making
• Designing effective A/B tests
• Analytical techniques in A/B testing

Causal AI (Gabriel Terejanu)

• Importance of causality in AI
• What is a causal model?
• What is an intervention?
• How to estimate causal effects?

Time series analysis and forecasting (Jože Rožanec)

• Introduction to time series
• Analysis tools and real-world examples
• Time series forecasting 

Time series: Forecasting, XAI, and databases (Jože Rožanec)

• Using network models to represent and forecast time series
• Introduction to explainability methods
• Introduction to time series databases

Graph data management (Dumitru Roman)

• Intro to graph data structure
• Knowledge Graphs
• Graph data management (graph databases with Noe4j, graph data model, graph construction and querying)

Graph data analytics (Daniel Schroeder, Dumitru Roman)

• Complex Network and Graph Analysis in IGraph
• Intro to Graph Neural Networks in PyG
• Graph data visualization

Data enrichment (Roberto Avogadro, Dumitru Roman)

• Data linking
• Tabular data enrichment 
• Human-in-the-Loop (HITL) for data enrichment

LLMs-based data processing (Ioan Toma)

• Introduction to LLM-based Data Processing
• Knowledge Extraction using LLMs
• Document Classification, Summarization and Comparison using LLMs

Conversational AI (Ioan Toma)

• Conversational AI setup and designing a chatbot interface
• Semantic Knowledge Graphs and their role in Conversational AI
• Building a chatbot using Onlim Conversational AI framework

LLM-based agents for data science (Hui Song)

• Use of ChatGPT Data Analyst to process data files, generate data processing code
• Development of LLM-based agents for multi-phase data processing tasks
• Multi-agents for complex and collaborative data processing

Findable, Accessible, Interoperable, Reusable (FAIR) data (Anna Fensel)

• Introduction to FAIR data. Examples from agri-food and health domains
• How to make data FAIR? Open data, closed data and everything in between
• Research data infrastructures

Best practices in data sharing (Anna Fensel)

• Legal compliance (GDPR, AI Act, Data Act)
• Consent, contracts and licenses, empowered with knowledge graphs
• Incentivising data sharing

High performance data processing (Radu Prodan)

• Parallel computing architectures
• Multiprocessing
• Parallel algorithms
• Parallel computing for AI and data science

Data pipeline scheduling on the computing continuum (Nikolay Nikolov)

• Introduction to pipeline scheduling in the context of big data and distributed applications 
• The importance and challenges of pipeline scheduling
• Solutions and practical approaches to pipeline scheduling

Operationalizing data and ML pipelines (Wiktor Sowinski-Mydlarz)

• Contemporary Data Processing
• GATE Institute Data Platform
• Alternatives and Decisions
• The Lifecycle

Deployment, orchestration, monitoring of data and ML pipelines (Wiktor Sowinski-Mydlarz)

• Data Spaces: Decentralized Supply and Consumption of Data Services
• Private Cloud for Big Data Processing
• Platform support
• Resources: Free products and software bibles