Moritz Grosse-Wentrup

Gunnar König, Timo Freiesleben, Moritz Grosse-Wentrup

Algorithmic recourse recommendations, such as Karimi et al.'s (2021) causal recourse (CR), inform stakeholders of how to act to revert unfavourable decisions. However, some actions lead to acceptance (i.e., revert the model's decision) but do not lead to improvement (i.e., may not revert the underlying real-world state). To recommend such actions is to recommend fooling the predictor. We introduce a novel method, Improvement-Focused Causal Recourse (ICR), which involves a conceptual shift: Firstly, we require ICR recommendations to guide towards improvement. Secondly, we do not tailor the recommendations to be accepted by a specific predictor. Instead, we leverage causal knowledge to design decision systems that predict accurately pre- and post-recourse. As a result, improvement guarantees translate into acceptance guarantees. We demonstrate that given correct causal knowledge, ICR, in contrast to existing approaches, guides towards both acceptance and improvement.

Christoph Luther, Gunnar König, Moritz Grosse-Wentrup

The Shapley Additive Global Importance (SAGE) value is a theoretically appealing interpretability method that fairly attributes global importance to a model's features. However, its exact calculation requires the computation of the feature's surplus performance contributions over an exponential number of feature sets. This is computationally expensive, particularly because estimating the surplus contributions requires sampling from conditional distributions. Thus, SAGE approximation algorithms only take a fraction of the feature sets into account. We propose $d$-SAGE, a method that accelerates SAGE approximation. $d$-SAGE is motivated by the observation that conditional independencies (CIs) between a feature and the model target imply zero surplus contributions, such that their computation can be skipped. To identify CIs, we leverage causal structure learning (CSL) to infer a graph that encodes (conditional) independencies in the data as $d$-separations. This is computationally more efficient because the expense of the one-time graph inference and the $d$-separation queries is negligible compared to the expense of surplus contribution evaluations. Empirically we demonstrate that $d$-SAGE enables the efficient and accurate estimation of SAGE values.

Sebastian Weichwald, Moritz Grosse-Wentrup, Arthur Gretton

Causal inference concerns the identification of cause-effect relationships between variables, e.g. establishing whether a stimulus affects activity in a certain brain region. The observed variables themselves often do not constitute meaningful causal variables, however, and linear combinations need to be considered. In electroencephalographic studies, for example, one is not interested in establishing cause-effect relationships between electrode signals (the observed variables), but rather between cortical signals (the causal variables) which can be recovered as linear combinations of electrode signals. We introduce MERLiN (Mixture Effect Recovery in Linear Networks), a family of causal inference algorithms that implement a novel means of constructing causal variables from non-causal variables. We demonstrate through application to EEG data how the basic MERLiN algorithm can be extended for application to different (neuroimaging) data modalities. Given an observed linear mixture, the algorithms can recover a causal variable that is a linear effect of another given variable. That is, MERLiN allows us to recover a cortical signal that is affected by activity in a certain brain region, while not being a direct effect of the stimulus. The Python/Matlab implementation for all presented algorithms is available on https://github.com/sweichwald/MERLiN

Anja Meunier, Michal Robert Žák, Lucas Munz et al.

We introduce Brain-Artificial Intelligence Interfaces (BAIs) as a new class of Brain-Computer Interfaces (BCIs). Unlike conventional BCIs, which rely on intact cognitive capabilities, BAIs leverage the power of artificial intelligence to replace parts of the neuro-cognitive processing pipeline. BAIs allow users to accomplish complex tasks by providing high-level intentions, while a pre-trained AI agent determines low-level details. This approach enlarges the target audience of BCIs to individuals with cognitive impairments, a population often excluded from the benefits of conventional BCIs. We present the general concept of BAIs and illustrate the potential of this new approach with a Conversational BAI based on EEG. In particular, we show in an experiment with simulated phone conversations that the Conversational BAI enables complex communication without the need to generate language. Our work thus demonstrates, for the first time, the ability of a speech neuroprosthesis to enable fluent communication in realistic scenarios with non-invasive technologies.

Xiaoxi Wei, A. Aldo Faisal, Moritz Grosse-Wentrup et al.

Transfer learning and meta-learning offer some of the most promising avenues to unlock the scalability of healthcare and consumer technologies driven by biosignal data. This is because current methods cannot generalise well across human subjects' data and handle learning from different heterogeneously collected data sets, thus limiting the scale of training data. On the other side, developments in transfer learning would benefit significantly from a real-world benchmark with immediate practical application. Therefore, we pick electroencephalography (EEG) as an exemplar for what makes biosignal machine learning hard. We design two transfer learning challenges around diagnostics and Brain-Computer-Interfacing (BCI), that have to be solved in the face of low signal-to-noise ratios, major variability among subjects, differences in the data recording sessions and techniques, and even between the specific BCI tasks recorded in the dataset. Task 1 is centred on the field of medical diagnostics, addressing automatic sleep stage annotation across subjects. Task 2 is centred on Brain-Computer Interfacing (BCI), addressing motor imagery decoding across both subjects and data sets. The BEETL competition with its over 30 competing teams and its 3 winning entries brought attention to the potential of deep transfer learning and combinations of set theory and conventional machine learning techniques to overcome the challenges. The results set a new state-of-the-art for the real-world BEETL benchmark.

Gunnar König, Timo Freiesleben, Moritz Grosse-Wentrup

Algorithmic recourse explanations inform stakeholders on how to act to revert unfavorable predictions. However, in general ML models do not predict well in interventional distributions. Thus, an action that changes the prediction in the desired way may not lead to an improvement of the underlying target. Such recourse is neither meaningful nor robust to model refits. Extending the work of Karimi et al. (2021), we propose meaningful algorithmic recourse (MAR) that only recommends actions that improve both prediction and target. We justify this selection constraint by highlighting the differences between model audit and meaningful, actionable recourse explanations. Additionally, we introduce a relaxation of MAR called effective algorithmic recourse (EAR), which, under certain assumptions, yields meaningful recourse by only allowing interventions on causes of the target.

Gunnar König, Timo Freiesleben, Bernd Bischl et al.

Global model-agnostic feature importance measures either quantify whether features are directly used for a model's predictions (direct importance) or whether they contain prediction-relevant information (associative importance). Direct importance provides causal insight into the model's mechanism, yet it fails to expose the leakage of information from associated but not directly used variables. In contrast, associative importance exposes information leakage but does not provide causal insight into the model's mechanism. We introduce DEDACT - a framework to decompose well-established direct and associative importance measures into their respective associative and direct components. DEDACT provides insight into both the sources of prediction-relevant information in the data and the direct and indirect feature pathways by which the information enters the model. We demonstrate the method's usefulness on simulated examples.

Alex Markham, Richeek Das, Moritz Grosse-Wentrup

We consider the problem of causal structure learning in the setting of heterogeneous populations, i.e., populations in which a single causal structure does not adequately represent all population members, as is common in biological and social sciences. To this end, we introduce a distance covariance-based kernel designed specifically to measure the similarity between the underlying nonlinear causal structures of different samples. Indeed, we prove that the corresponding feature map is a statistically consistent estimator of nonlinear independence structure, rendering the kernel itself a statistical test for the hypothesis that sets of samples come from different generating causal structures. Even stronger, we prove that the kernel space is isometric to the space of causal ancestral graphs, so that distance between samples in the kernel space is guaranteed to correspond to distance between their generating causal structures. This kernel thus enables us to perform clustering to identify the homogeneous subpopulations, for which we can then learn causal structures using existing methods. Though we focus on the theoretical aspects of the kernel, we also evaluate its performance on synthetic data and demonstrate its use on a real gene expression data set.

Gunnar König, Christoph Molnar, Bernd Bischl et al.

Interpretable Machine Learning (IML) methods are used to gain insight into the relevance of a feature of interest for the performance of a model. Commonly used IML methods differ in whether they consider features of interest in isolation, e.g., Permutation Feature Importance (PFI), or in relation to all remaining feature variables, e.g., Conditional Feature Importance (CFI). As such, the perturbation mechanisms inherent to PFI and CFI represent extreme reference points. We introduce Relative Feature Importance (RFI), a generalization of PFI and CFI that allows for a more nuanced feature importance computation beyond the PFI versus CFI dichotomy. With RFI, the importance of a feature relative to any other subset of features can be assessed, including variables that were not available at training time. We derive general interpretation rules for RFI based on a detailed theoretical analysis of the implications of relative feature relevance, and demonstrate the method's usefulness on simulated examples.

Christoph Molnar, Gunnar König, Julia Herbinger et al.

An increasing number of model-agnostic interpretation techniques for machine learning (ML) models such as partial dependence plots (PDP), permutation feature importance (PFI) and Shapley values provide insightful model interpretations, but can lead to wrong conclusions if applied incorrectly. We highlight many general pitfalls of ML model interpretation, such as using interpretation techniques in the wrong context, interpreting models that do not generalize well, ignoring feature dependencies, interactions, uncertainty estimates and issues in high-dimensional settings, or making unjustified causal interpretations, and illustrate them with examples. We focus on pitfalls for global methods that describe the average model behavior, but many pitfalls also apply to local methods that explain individual predictions. Our paper addresses ML practitioners by raising awareness of pitfalls and identifying solutions for correct model interpretation, but also addresses ML researchers by discussing open issues for further research.

Matthias R. Hohmann, Lisa Konieczny, Michelle Hackl et al.

Neurophysiological studies are typically conducted in laboratories with limited ecological validity, scalability, and generalizability of findings. This is a significant challenge for the development of brain-computer interfaces (BCIs), which ultimately need to function in unsupervised settings on consumer-grade hardware. We introduce MYND: A framework that couples consumer-grade recording hardware with an easy-to-use application for the unsupervised evaluation of BCI control strategies. Subjects are guided through experiment selection, hardware fitting, recording, and data upload in order to self-administer multi-day studies that include neurophysiological recordings and questionnaires. As a use case, we evaluate two BCI control strategies ("Positive memories" and "Music imagery") in a realistic scenario by combining MYND with a four-channel electroencephalogram (EEG). Thirty subjects recorded 70.4 hours of EEG data with the system at home. The median headset fitting time was 25.9 seconds, and a median signal quality of 90.2% was retained during recordings.Neural activity in both control strategies could be decoded with an average offline accuracy of 68.5% and 64.0% across all days. The repeated unsupervised execution of the same strategy affected performance, which could be tackled by implementing feedback to let subjects switch between strategies or devise new strategies with the platform.

Alex Markham, Moritz Grosse-Wentrup

We consider the task of causal structure learning over measurement dependence inducing latent (MeDIL) causal models. We show that this task can be framed in terms of the graph theoretic problem of finding edge clique covers,resulting in an algorithm for returning minimal MeDIL causal models (minMCMs). This algorithm is non-parametric, requiring no assumptions about linearity or Gaussianity. Furthermore, despite rather weak assumptions aboutthe class of MeDIL causal models, we show that minimality in minMCMs implies some rather specific and interesting properties. By establishing MeDIL causal models as a semantics for edge clique covers, we also provide a starting point for future work further connecting causal structure learning to developments in graph theory and network science.

Jiachen Xu, Moritz Grosse-Wentrup, Vinay Jayaram

Methods based on Riemannian geometry have proven themselves to be good models for decoding in brain-computer interfacing (BCI). However, one major drawback of these methods is that it is not possible to determine what aspect of the signal the classifier is built on, leaving open the possibility that artifacts drive classification performance. In areas where artifactual control is problematic, specifically neurofeedback and BCIs in patient populations, this has led people to continue to rely on spatial filters as a way of generating features that are provably brain-related. Furthermore, these methods also suffer from the curse of dimensionality and are almost infeasible in high-density online BCI systems. To tackle these drawbacks, we introduce here a method for computing spatial filters from any linear function in the Riemannian tangent space, which allows for more efficient classification as well as the removal of artifact sources from classifiers built on Riemannian methods. We first prove a fundamental relationship between certain tangent spaces and spatial filtering methods, including an explanation of common spatial patterns within this framework, and then validate our proposed approach using an open-access BCI analysis framework.

Paul K. Rubenstein, Sebastian Weichwald, Stephan Bongers et al.

Complex systems can be modelled at various levels of detail. Ideally, causal models of the same system should be consistent with one another in the sense that they agree in their predictions of the effects of interventions. We formalise this notion of consistency in the case of Structural Equation Models (SEMs) by introducing exact transformations between SEMs. This provides a general language to consider, for instance, the different levels of description in the following three scenarios: (a) models with large numbers of variables versus models in which the `irrelevant' or unobservable variables have been marginalised out; (b) micro-level models versus macro-level models in which the macro-variables are aggregate features of the micro-variables; (c) dynamical time series models versus models of their stationary behaviour. Our analysis stresses the importance of well specified interventions in the causal modelling process and sheds light on the interpretation of cyclic SEMs.

Anastasia-Atalanti Mastakouri, Sebastian Weichwald, Ozan Özdenizci et al.

We propose to fuse two currently separate research lines on novel therapies for stroke rehabilitation: brain-computer interface (BCI) training and transcranial electrical stimulation (TES). Specifically, we show that BCI technology can be used to learn personalized decoding models that relate the global configuration of brain rhythms in individual subjects (as measured by EEG) to their motor performance during 3D reaching movements. We demonstrate that our models capture substantial across-subject heterogeneity, and argue that this heterogeneity is a likely cause of limited effect sizes observed in TES for enhancing motor performance. We conclude by discussing how our personalized models can be used to derive optimal TES parameters, e.g., stimulation site and frequency, for individual patients.

Leila Wehbe, Anwar Nunez-Elizalde, Marcel van Gerven et al.

This workshop explores the interface between cognitive neuroscience and recent advances in AI fields that aim to reproduce human performance such as natural language processing and computer vision, and specifically deep learning approaches to such problems. When studying the cognitive capabilities of the brain, scientists follow a system identification approach in which they present different stimuli to the subjects and try to model the response that different brain areas have of that stimulus. The goal is to understand the brain by trying to find the function that expresses the activity of brain areas in terms of different properties of the stimulus. Experimental stimuli are becoming increasingly complex with more and more people being interested in studying real life phenomena such as the perception of natural images or natural sentences. There is therefore a need for a rich and adequate vector representation of the properties of the stimulus, that we can obtain using advances in machine learning. In parallel, new ML approaches, many of which in deep learning, are inspired to a certain extent by human behavior or biological principles. Neural networks for example were originally inspired by biological neurons. More recently, processes such as attention are being used which have are inspired by human behavior. However, the large bulk of these methods are independent of findings about brain function, and it is unclear whether it is at all beneficial for machine learning to try to emulate brain function in order to achieve the same tasks that the brain achieves.

Sebastian Weichwald, Tatiana Fomina, Bernhard Schölkopf et al.

While the channel capacity reflects a theoretical upper bound on the achievable information transmission rate in the limit of infinitely many bits, it does not characterise the information transfer of a given encoding routine with finitely many bits. In this note, we characterise the quality of a code (i. e. a given encoding routine) by an upper bound on the expected minimum error probability that can be achieved when using this code. We show that for equientropic channels this upper bound is minimal for codes with maximal marginal entropy. As an instructive example we show for the additive white Gaussian noise (AWGN) channel that random coding---also a capacity achieving code---indeed maximises the marginal entropy in the limit of infinite messages.

Sebastian Weichwald, Arthur Gretton, Bernhard Schölkopf et al.

Causal inference concerns the identification of cause-effect relationships between variables. However, often only linear combinations of variables constitute meaningful causal variables. For example, recovering the signal of a cortical source from electroencephalography requires a well-tuned combination of signals recorded at multiple electrodes. We recently introduced the MERLiN (Mixture Effect Recovery in Linear Networks) algorithm that is able to recover, from an observed linear mixture, a causal variable that is a linear effect of another given variable. Here we relax the assumption of this cause-effect relationship being linear and present an extended algorithm that can pick up non-linear cause-effect relationships. Thus, the main contribution is an algorithm (and ready to use code) that has broader applicability and allows for a richer model class. Furthermore, a comparative analysis indicates that the assumption of linear cause-effect relationships is not restrictive in analysing electroencephalographic data.

Sebastian Weichwald, Bernhard Schölkopf, Tonio Ball et al.

Pattern recognition in neuroimaging distinguishes between two types of models: encoding- and decoding models. This distinction is based on the insight that brain state features, that are found to be relevant in an experimental paradigm, carry a different meaning in encoding- than in decoding models. In this paper, we argue that this distinction is not sufficient: Relevant features in encoding- and decoding models carry a different meaning depending on whether they represent causal- or anti-causal relations. We provide a theoretical justification for this argument and conclude that causal inference is essential for interpretation in neuroimaging.

Sebastian Weichwald, Timm Meyer, Bernhard Schölkopf et al.

While invasively recorded brain activity is known to provide detailed information on motor commands, it is an open question at what level of detail information about positions of body parts can be decoded from non-invasively acquired signals. In this work it is shown that index finger positions can be differentiated from non-invasive electroencephalographic (EEG) recordings in healthy human subjects. Using a leave-one-subject-out cross-validation procedure, a random forest distinguished different index finger positions on a numerical keyboard above chance-level accuracy. Among the different spectral features investigated, high $β$-power (20-30 Hz) over contralateral sensorimotor cortex carried most information about finger position. Thus, these findings indicate that finger position is in principle decodable from non-invasive features of brain activity that generalize across individuals.

Vinay Jayaram, Morteza Alamgir, Yasemin Altun et al.

The performance of brain-computer interfaces (BCIs) improves with the amount of available training data, the statistical distribution of this data, however, varies across subjects as well as across sessions within individual subjects, limiting the transferability of training data or trained models between them. In this article, we review current transfer learning techniques in BCIs that exploit shared structure between training data of multiple subjects and/or sessions to increase performance. We then present a framework for transfer learning in the context of BCIs that can be applied to any arbitrary feature space, as well as a novel regression estimation method that is specifically designed for the structure of a system based on the electroencephalogram (EEG). We demonstrate the utility of our framework and method on subject-to-subject transfer in a motor-imagery paradigm as well as on session-to-session transfer in one patient diagnosed with amyotrophic lateral sclerosis (ALS), showing that it is able to outperform other comparable methods on an identical dataset.

Sebastian Weichwald, Timm Meyer, Ozan Özdenizci et al.

Causal terminology is often introduced in the interpretation of encoding and decoding models trained on neuroimaging data. In this article, we investigate which causal statements are warranted and which ones are not supported by empirical evidence. We argue that the distinction between encoding and decoding models is not sufficient for this purpose: relevant features in encoding and decoding models carry a different meaning in stimulus- and in response-based experimental paradigms. We show that only encoding models in the stimulus-based setting support unambiguous causal interpretations. By combining encoding and decoding models trained on the same data, however, we obtain insights into causal relations beyond those that are implied by each individual model type. We illustrate the empirical relevance of our theoretical findings on EEG data recorded during a visuo-motor learning task.

Pouyan R. Fard, Moritz Grosse-Wentrup

Understanding the relationship between the decoding accuracy of a brain-computer interface (BCI) and a subject's subjective feeling of control is important for determining a lower limit on decoding accuracy for a BCI that is to be deployed outside a laboratory environment. We investigated this relationship by systematically varying the level of control in a simulated BCI task. We find that a binary decoding accuracy of 65% is required for users to report more often than not that they are feeling in control of the system. Decoding accuracies above 75%, on the other hand, added little in terms of the level of perceived control. We further find that the probability of perceived control does not only depend on the actual decoding accuracy, but is also in influenced by whether subjects successfully complete the given task in the allotted time frame.