Trustworthy and interpretable ML
Automating decisions based on data is a multifaceted problem with many pitfalls. We tackle multiple ends of the spectrum, from the detection of anomalies to calibration of models and Bayesian decision-making.

Series in this AoI

Probabilistic Models

Uncertainty permeates all aspects of real-world agency: Perception is subject to uncertainty owing to partial observability and unreliable sensors; the effects of an agent’s own actions may have non-determinstic effects; and even the tasks an agent is given may be subject to ambiguity or incomplete specification. Probability theory is a mathematical framework for the conceptualisation of …

Simulation-based Inference

Simulation-based inference (SBI) offers a powerful framework for Bayesian parameter estimation in intricate scientific simulations where likelihood evaluations are not feasible. Recent advancements in neural network-based density estimation methods have broadened the horizons for SBI, enhancing its efficiency and scalability. While these novel methods show potential in deepening our understanding …

Uncertainty Quantification

Uncertainty quantification (UQ) in machine learning is the practice of measuring or estimating uncertainty in models. It is a set of tools to understand the limitations of models and predictions, and to make better decisions. UQ is a key component of trustworthy and interpretable machine learning.

Explainable AI

Large opaque models like neural networks require dedicated methods to study and interpret their behavior. In this series we review recent developments, analysing their relevance for business applications.

Classifier calibration

For many applications of probabilistic classifiers it is important that the predicted confidence vectors reflect true probabilities (one says that the classifier is calibrated). Recently, it has been shown that common models fail to satisfy this property, making reliable methods for measuring and improving calibration important tools. Unfortunately, obtaining these is far from trivial, especially …