Data valuation
Attributions of value to training samples can be used to examine data, improve data acquisition, debug and improve models or compensate providers. Recent developments in the field enable principled and useful definitions of value which overcome the exponential complexity involved in classical approaches.

All content in this series

Pill

Beyond neural scaling laws: beating power law scaling via data pruning

Large neural networks are very “data hungry”, but data pruning holds the promise to alleviate this dependence. A recent paper …

A photo of Fabio
Data Valuation
Pill

What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation

When data samples are difficult to learn, neural networks tend to memorize their labels rather than infer useful features. This has been …

A photo of Fabio
Data Valuation
Pill

CS-Shapley: Class-wise Shapley Values for Data Valuation in Classification

Using in-class accuracy to up-weight the value of a data point and out-of-class accuracy as a discounting factor, the authors define a new …

A picture of Miguel
Data Valuation
Pill

Data Banzhaf: A Robust Data Valuation Framework for Machine Learning

A definition of data value with much higher rank stability than Shapley values, proven to be most robust among all semi-values, mitigates …

A picture of Miguel
Data Valuation
Pill

Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning

In data valuation, relaxing the Shapley axioms leads to weighting schemes for marginal utilities that improve the quality of values

A picture of Miguel
Data Valuation
Software

pyDVL: the python Data Valuation Library

pyDVL strives to offer reference implementations of algorithms for data valuation with a common interface compatible with sklearn and a …

A picture of Anes A picture of Miguel A photo of Fabio A photo of Markus
Data Valuation
Pill

Improving cooperative game theory-based data valuation via data utility learning

Data valuation is typically limited by the exponential complexity of retraining for each possible subset of training points. This paper …

A picture of Miguel
Data Valuation
Pill

Energy-Based Learning for Cooperative Games, with Applications to Valuation Problems in Machine Learning

A theoretical framework connecting data valuation scores, game theory and energy based models is presented. Common criteria like Shapley or …

A photo of Fabio
Data Valuation
Seminar

Data (e)valuation and model interpretation: a game theoretic approach

Attributing a “fair” (for some definition of fair) value to training samples has multiple applications. It can be used to …

A picture of Miguel
Data Valuation
Review

Data valuation techniques. Applications, advances and problems

The attribution of value to data points can help in data collection, active learning, gauging model robustness, improving model performance …

A picture of Miguel
Data Valuation

Other series in Efficient Machine Learning

Representation Learning

Representation learning is concerned with finding semantically useful and compact, i.e., low dimensional, representations for inputs from selected data distributions. These representations can be efficiently used for downstream tasks, …

1 item

Data efficiency

Numerous applications of Machine Learning in business intelligence, process optimization, product development or sales do not benefit from true “big data”. Blind application of state-of-the-art methods can lead to catastrophic …

6 items

Efficient RL

Recent and popular advances in Reinforcement Learning are known to be data-hungry. Attempts to handle this deficit include developing complex simulators while improving the sim2real transition of models, explicitly modelling the …

9 items
Check all of our work