arXiv subjects

Martin Jullum

Publications and source records attributed to Martin Jullum.

23 records · Page 2Linked to original sources

arXiv2021

groupShapley: Efficient prediction explanation with Shapley values for feature groups

Shapley values has established itself as one of the most appropriate and theoretically sound frameworks for explaining predictions from complex machine learning models. The popularity of Shapley values in the explanation setting is probably due to its unique theoretical properties. The main drawback with Shapley values, however, is that its computational complexity grows exponentially in the number of input features, making it unfeasible in many real world situations where there could be hundreds or thousands of features. Furthermore, with many (dependent) features, presenting/visualizing and interpreting the computed Shapley values also becomes challenging. The present paper introduces groupShapley: a conceptually simple approach for dealing with the aforementioned bottlenecks. The idea is to group the features, for example by type or dependence, and then compute and present Shapley values for these groups instead of for all individual features. Reducing hundreds or thousands of features to half a dozen or so, makes precise computations practically feasible and the presentation and knowledge extraction greatly simplified. We prove that under certain conditions, groupShapley is equivalent to summing the feature-wise Shapley values within each feature group. Moreover, we provide a simulation study exemplifying the differences when these conditions are not met. We illustrate the usability of the approach in a real world car insurance example, where groupShapley is used to provide simple and intuitive explanations.

stat.ML↗

arXiv2021

Explaining predictive models using Shapley values and non-parametric vine copulas

The original development of Shapley values for prediction explanation relied on the assumption that the features being described were independent. If the features in reality are dependent this may lead to incorrect explanations. Hence, there have recently been attempts of appropriately modelling/estimating the dependence between the features. Although the proposed methods clearly outperform the traditional approach assuming independence, they have their weaknesses. In this paper we propose two new approaches for modelling the dependence between the features. Both approaches are based on vine copulas, which are flexible tools for modelling multivariate non-Gaussian distributions able to characterise a wide range of complex dependencies. The performance of the proposed methods is evaluated on simulated data sets and a real data set. The experiments demonstrate that the vine copula approaches give more accurate approximations to the true Shapley values than its competitors.

stat.ME↗

arXiv2020

Explaining predictive models with mixed features using Shapley values and conditional inference trees

It is becoming increasingly important to explain complex, black-box machine learning models. Although there is an expanding literature on this topic, Shapley values stand out as a sound method to explain predictions from any type of machine learning model. The original development of Shapley values for prediction explanation relied on the assumption that the features being described were independent. This methodology was then extended to explain dependent features with an underlying continuous distribution. In this paper, we propose a method to explain mixed (i.e. continuous, discrete, ordinal, and categorical) dependent features by modeling the dependence structure of the features using conditional inference trees. We demonstrate our proposed method against the current industry standards in various simulation studies and find that our method often outperforms the other approaches. Finally, we apply our method to a real financial data set used in the 2018 FICO Explainable Machine Learning Challenge and show how our explanations compare to the FICO challenge Recognition Award winning team.

stat.ML↗

arXiv2020

Explaining individual predictions when features are dependent: More accurate approximations to Shapley values

Explaining complex or seemingly simple machine learning models is an important practical problem. We want to explain individual predictions from a complex machine learning model by learning simple, interpretable explanations. Shapley values is a game theoretic concept that can be used for this purpose. The Shapley value framework has a series of desirable theoretical properties, and can in principle handle any predictive model. Kernel SHAP is a computationally efficient approximation to Shapley values in higher dimensions. Like several other existing methods, this approach assumes that the features are independent, which may give very wrong explanations. This is the case even if a simple linear model is used for predictions. In this paper, we extend the Kernel SHAP method to handle dependent features. We provide several examples of linear and non-linear models with various degrees of feature dependence, where our method gives more accurate approximations to the true Shapley values. We also propose a method for aggregating individual Shapley values, such that the prediction can be explained by groups of dependent variables.

stat.ML↗

arXiv2019

Estimating seal pup production in the Greenland Sea using Bayesian hierarchical modeling

The Greenland Sea is an important breeding ground for harp and hooded seals. Estimates of the annual seal pup production are critical factors in the abundance estimation needed for management of the species. These estimates are usually based on counts from aerial photographic surveys. However, only a minor part of the whelping region can be photographed, due to its large extent. To estimate the total seal pup production, we propose a Bayesian hierarchical modeling approach motivated by viewing the seal pup appearances as a realization of a log-Gaussian Cox process using covariate information from satellite imagery as a proxy for ice thickness. For inference, we utilize the stochastic partial differential equation (SPDE) module of the integrated nested Laplace approximation (INLA) framework. In a case study using survey data from 2012, we compare our results with existing methodology in a comprehensive cross-validation study. The results of the study indicate that our method improves local estimation performance, and that the increased prediction uncertainty of our method is required to obtain calibrated count predictions. This suggests that the sampling density of the survey design may not be sufficient to obtain reliable estimates of the seal pup production.

stat.AP↗