TransferLab courses

This is a one-day workshop introducing the concept of probabilistic model checking and its applications.

Goal of the training, including:

• Understanding the theory behind probabilistic model checking.
• Getting to know the most important algorithms.
• Learning how to model systems as Markov chains and Markov decision processes.
• Learning how to formulate queries about the behavior of a system.

This is a two-day workshop for ML practitioners who want to make their models more understandable to themselves and decision makers.

Goal of the training, including:

• Get an overview of the potential applications of explainability in machine learning.
• Learn the taxonomy of explainability methods and their applicability to different types of models.
• Learn how to integrate explainability considerations into the machine learning pipeline.
• Learn about model-independent and model-specific explainability methods.

This is a two-day workshop introducing Bayesian modeling using practical examples and probabilistic programming.

Goal of the training, including:

• Learn Bayesian methodology and understand how Bayesian inference can be used as a general machine learning tool.
• Understand the computational challenges in Bayesian inference and how to overcome them.
• Become familiar with the basics of approximate Bayesian inference.
• Understanding the first steps in probabilistic programming with Pyro.

An overview of model-based planning and control for AI engineers interested in solving real-world decision and control problems using efficient methods.

Aim of the training, including:

• Fundamentals of decision making: environments, trajectories, actors, and rewards.
• Typical and less typical control problems.
• Planning and classical control.
• From simulation to reality.

This 2-day Deep Dive into Deep RL is suitable for engineers who want to solve real-world control problems using efficient methods.

Aims of the training, among others:

• Participants will learn the most important recent advances in Deep RL and gain a sense of when and how RL techniques should (and perhaps more importantly, when not to) be used. Learning content includes:
• Fundamentals of deep, model-free, and model-based RL.
• What problems are appropriate for an RL approach?
• Exploration vs. exploitation
• On-policy, off-policy, and offline learning.

A two-day introduction to unsupervised ML techniques for anomaly detection and their strengths and weaknesses in various application domains.

Goal of the training, including:

• Understand qualitative and quantitative definitions of anomalies.
• Overview of the theoretical foundations and practical implementations of several anomaly detection algorithms.
• Understand which algorithms are appropriate for which application domains.
• Learn how to evaluate and compare the performance of different algorithms.

An introduction to the pitfalls of uncalibrated classifiers and modern techniques for (re)calibration.

Goal of the training, including:

• Quantitative understanding of how optimal decisions depend on all probabilities, not just the predicted class.
• Learning to measure the (mis)calibration of models.
• Recalibration of classifiers during training: loss functions, regularization.
• Recalibration of classifiers a posteriori: non-parametric, parametric, Bayesian.