Rune D. Kjærsgaard

Line H. Clemmensen, Rune D. Kjærsgaard

Data representativity is crucial when drawing inference from data through machine learning models. Scholars have increased focus on unraveling the bias and fairness in models, also in relation to inherent biases in the input data. However, limited work exists on the representativity of samples (datasets) for appropriate inference in AI systems. This paper reviews definitions and notions of a representative sample and surveys their use in scientific AI literature. We introduce three measurable concepts to help focus the notions and evaluate different data samples. Furthermore, we demonstrate that the contrast between a representative sample in the sense of coverage of the input space, versus a representative sample mimicking the distribution of the target population is of particular relevance when building AI systems. Through empirical demonstrations on US Census data, we evaluate the opposing inherent qualities of these concepts. Finally, we propose a framework of questions for creating and documenting data with data representativity in mind, as an addition to existing dataset documentation templates.

Rune Kjærsgaard, Ahcène Boubekki, Line Clemmensen

Prototypical self-explainable classifiers have emerged to meet the growing demand for interpretable AI systems. These classifiers are designed to incorporate high transparency in their decisions by basing inference on similarity with learned prototypical objects. While these models are designed with diversity in mind, the learned prototypes often do not sufficiently represent all aspects of the input distribution, particularly those in low density regions. Such lack of sufficient data representation, known as representation bias, has been associated with various detrimental properties related to machine learning diversity and fairness. In light of this, we introduce pantypes, a new family of prototypical objects designed to capture the full diversity of the input distribution through a sparse set of objects. We show that pantypes can empower prototypical self-explainable models by occupying divergent regions of the latent space and thus fostering high diversity, interpretability and fairness.

Rune D. Kjærsgaard, Aaron Bello-Arufe, Alexander D. Rathcke et al.

The absorption of light by molecules in the atmosphere of Earth is a complication for ground-based observations of astrophysical objects. Comprehensive information on various molecular species is required to correct for this so called telluric absorption. We present a neural network autoencoder approach for extracting a telluric transmission spectrum from a large set of high-precision observed solar spectra from the HARPS-N radial velocity spectrograph. We accomplish this by reducing the data into a compressed representation, which allows us to unveil the underlying solar spectrum and simultaneously uncover the different modes of variation in the observed spectra relating to the absorption of $\mathrm{H_2O}$ and $\mathrm{O_2}$ in the atmosphere of Earth. We demonstrate how the extracted components can be used to remove $\mathrm{H_2O}$ and $\mathrm{O_2}$ tellurics in a validation observation with similar accuracy and at less computational expense than a synthetic approach with molecfit.

Rune D. Kjærsgaard, Manja G. Grønberg, Line K. H. Clemmensen

Data imbalance is common in production data, where controlled production settings require data to fall within a narrow range of variation and data are collected with quality assessment in mind, rather than data analytic insights. This imbalance negatively impacts the predictive performance of models on underrepresented observations. We propose sampling to adjust for this imbalance with the goal of improving the performance of models trained on historical production data. We investigate the use of three sampling approaches to adjust for imbalance. The goal is to downsample the covariates in the training data and subsequently fit a regression model. We investigate how the predictive power of the model changes when using either the sampled or the original data for training. We apply our methods on a large biopharmaceutical manufacturing data set from an advanced simulation of penicillin production and find that fitting a model using the sampled data gives a small reduction in the overall predictive performance, but yields a systematically better performance on underrepresented observations. In addition, the results emphasize the need for alternative, fair, and balanced model evaluations.