Connectionists: XVII Madrid UPM Machine Learning and Advanced Statistics Summer School (June 16th - June 27th, 2025)

[mlas at fi.upm.es]

Dear colleagues,

The Technical University of Madrid (UPM) will once more organize the 'Madrid UPM Machine Learning and Advanced Statistics' summer school. The summer school will be held in Boadilla del Monte, near Madrid, from June 16th to June 27th. This year's edition comprises 12 week-long courses (15 lecture hours each), given during two weeks (six courses each week). Attendees may register in each course independently. No restrictions, besides those imposed by timetables, apply on the number or choice of courses.

Early registration is now *OPEN*. Extended information on course programmes, price, venue, accommodation and transport is available at the school's website:

https://www.dia.fi.upm.es/MLAS

There is a 25% discount for members of Spanish AEPIA and SEIO societies.  

Please, forward this information to your colleagues, students, and whomever you think may find it interesting.

Best regards,

Pedro Larrañaga, Concha Bielza, Bojan Mihaljević and Laura Gonzalez Veiga.
-- School coordinators.

*** List of courses and brief description ***

# Week 1 (June 16th - June 20th, 2025) 

## 1st session: 9:45-12:45

### Course 1: Bayesian Networks (15 h)
      Basics of Bayesian networks. Inference in Bayesian networks. Learning Bayesian networks from data. Real applications. Practical demonstration: R.

### Course 2: Time Series(15 h)
      Basic concepts in time series. Linear models for time series. Time series clustering. Practical demonstration: R.
      
## 2nd session: 13:45-16:45

### Course 3: Supervised Classification (15 h)
      Introduction. Assessing the performance of supervised classification algorithms. Preprocessing. Classification techniques. Combining multiple classifiers. Comparing supervised classification algorithms. Practical demonstration: python. 

### Course 4: Reinforcement learning (15 h)
      Introduction. Dynamic programming methods. Temporal-difference learning. Policy gradient methods. Causal reinforcement learning. Practical demonstration: R.  

## 3rd session: 17:00 - 20:00

### Course 5: Deep Learning (15 h)
      Introduction. Learning algorithms. Learning in deep networks. Deep Learning for Computer Vision. Deep Learning for Language. Practical session: Python notebooks with Google Colab with keras, Pytorch and Hugging Face Transformers.

### Course 6: Bayesian Inference (15 h)
      Introduction: Bayesian basics. Conjugate models. MCMC and other simulation methods. Regression and Hierarchical models. Model selection. Practical demonstration: R and WinBugs.
      

# Week 2 (June 23rd - June 27th, 2025)

## 1st session: 9:45-12:45 

### Course 7: Causality (15 h)
    Introduction. Causal graphs. Mediation analysis. Sensitivity analysis to unmeasured confounding. Counterfactual reasoning. Practical sessions: R.

### Course 8: Clustering (15 h)
      Introduction to clustering. Data exploration and preparation. Prototype-based clustering. Density-based clustering. Graph-based clustering. Cluster evaluation. Miscellanea. Conclusions and final advice. Practical session: R.

## 2nd session: 13:45-16:45

### Course 9: Gaussian Processes and Bayesian Optimization (15 h)
      Introduction to Gaussian processes. Sparse Gaussian processes. Deep Gaussian processes. Introduction to Bayesian optimization. Bayesian optimization in complex scenarios. Practical demonstration: python using GPytorch and BOTorch.
      
### Course 10: Explainable Machine Learning (15 h)
      Introduction. Inherently interpretable models. Post-hoc interpretation of black box models. Basics of causal inference. Beyond tabular and i.i.d. data. Other topics. Practical demonstration: Python with Google Colab. 
         
## 3rd session: 17:00-20:00

### Course 11:  Generative AI (15 h)
     Introduction to the course. Neural networks and deep learning. Generative AI for images. Generative AI for language. Hands-on session: Pytorch, VAEs, GANs, diffusion models, LLMs, aligning a generative LLM, using an open-source image generation model.
      
### Course 12: Feature Subset Selection (15 h)
      Introduction. Filter approaches. Embedded methods. Wrapper methods. Additional topics. Hands-on sessions: R and python.