Search

In the field of Explainable AI (XAI), counterfactual (CF) explanations are one prominent method to interpret a black-box model by suggesting changes to the input that would alter a prediction. In real-world applications, the input is predominantly in tabular form and comprised of mixed data types and complex feature interdependencies. These unique data characteristics are difficult to model, and we empirically show that they lead to bias towards specific feature types when generating CFs. To overcome this issue, we introduce TABCF, a CF explanation method that leverages a transformer-based Variational Autoencoder (VAE) tailored for modeling tabular data. Our approach uses transformers to learn a continuous latent space and a novel Gumbel-Softmax detokenizer that enables precise categorical reconstruction while preserving end-to-end differentiability. Extensive quantitative evaluation on five financial datasets demonstrates that TABCF does not exhibit bias toward specific feature types, and outperforms existing methods in producing effective CFs that align with common CF desiderata., Comment: Paper accepted at ICAIF '24: 5th ACM International Conference on AI in Finance, Brooklyn, NY, USA, November 2024

Recent advances in out-of-distribution (OOD) detection on image data show that pre-trained neural network classifiers can separate in-distribution (ID) from OOD data well, leveraging the class-discriminative ability of the model itself. Methods have been proposed that either use logit information directly or that process the model's penultimate layer activations. With "WeiPer", we introduce perturbations of the class projections in the final fully connected layer which creates a richer representation of the input. We show that this simple trick can improve the OOD detection performance of a variety of methods and additionally propose a distance-based method that leverages the properties of the augmented WeiPer space. We achieve state-of-the-art OOD detection results across multiple benchmarks of the OpenOOD framework, especially pronounced in difficult settings in which OOD samples are positioned close to the training set distribution. We support our findings with theoretical motivations and empirical observations, and run extensive ablations to provide insights into why WeiPer works.

Many European citizens become targets of the Kremlin propaganda campaigns, aiming to minimise public support for Ukraine, foster a climate of mistrust and disunity, and shape elections (Meister, 2022). To address this challenge, we developed ''Check News in 1 Click'', the first NLP-empowered pro-Kremlin propaganda detection application available in 7 languages, which provides the lay user with feedback on their news, and explains manipulative linguistic features and keywords. We conducted a user study, analysed user entries and models' behaviour paired with questionnaire answers, and investigated the advantages and disadvantages of the proposed interpretative solution., Comment: Accepted to EACL 2024 System Demonstrations

Written reflective practice is a regular exercise pre-service teachers perform during their higher education. Usually, their lecturers are expected to provide individual feedback, which can be a challenging task to perform on a regular basis. In this paper, we present the first open-source automated feedback tool based on didactic theory and implemented as a hybrid AI system. We describe the components and discuss the advantages and disadvantages of our system compared to the state-of-art generative large language models. The main objective of our work is to enable better learning outcomes for students and to complement the teaching activities of lecturers., Comment: 2 figures

The full-scale conflict between the Russian Federation and Ukraine generated an unprecedented amount of news articles and social media data reflecting opposing ideologies and narratives. These polarized campaigns have led to mutual accusations of misinformation and fake news, shaping an atmosphere of confusion and mistrust for readers worldwide. This study analyses how the media affected and mirrored public opinion during the first month of the war using news articles and Telegram news channels in Ukrainian, Russian, Romanian and English. We propose and compare two methods of multilingual automated pro-Kremlin propaganda identification, based on Transformers and linguistic features. We analyse the advantages and disadvantages of both methods, their adaptability to new genres and languages, and ethical considerations of their usage for content moderation. With this work, we aim to lay the foundation for further development of moderation tools tailored to the current conflict., Comment: 9 pages, 3 figures

Saliency methods attempt to explain deep neural networks by highlighting the most salient features of a sample. Some widely used methods are based on a theoretical framework called Deep Taylor Decomposition (DTD), which formalizes the recursive application of the Taylor Theorem to the network's layers. However, recent work has found these methods to be independent of the network's deeper layers and appear to respond only to lower-level image structure. Here, we investigate the DTD theory to better understand this perplexing behavior and found that the Deep Taylor Decomposition is equivalent to the basic gradient$\times$input method when the Taylor root points (an important parameter of the algorithm chosen by the user) are locally constant. If the root points are locally input-dependent, then one can justify any explanation. In this case, the theory is under-constrained. In an empirical evaluation, we find that DTD roots do not lie in the same linear regions as the input - contrary to a fundamental assumption of the Taylor theorem. The theoretical foundations of DTD were cited as a source of reliability for the explanations. However, our findings urge caution in making such claims., Comment: Published in Transactions on Machine Learning Research (TMLR)

K-nearest neighbors (KNN) is one of the earliest and most established algorithms in machine learning. For regression tasks, KNN averages the targets within a neighborhood which poses a number of challenges: the neighborhood definition is crucial for the predictive performance as neighbors might be selected based on uninformative features, and averaging does not account for how the function changes locally. We propose a novel method called Differential Nearest Neighbors Regression (DNNR) that addresses both issues simultaneously: during training, DNNR estimates local gradients to scale the features; during inference, it performs an n-th order Taylor approximation using estimated gradients. In a large-scale evaluation on over 250 datasets, we find that DNNR performs comparably to state-of-the-art gradient boosting methods and MLPs while maintaining the simplicity and transparency of KNN. This allows us to derive theoretical error bounds and inspect failures. In times that call for transparency of ML models, DNNR provides a good balance between performance and interpretability., Comment: published at ICML 2022

A variety of methods exist to explain image classification models. However, whether they provide any benefit to users over simply comparing various inputs and the model's respective predictions remains unclear. We conducted a user study (N=240) to test how such a baseline explanation technique performs against concept-based and counterfactual explanations. To this end, we contribute a synthetic dataset generator capable of biasing individual attributes and quantifying their relevance to the model. In a study, we assess if participants can identify the relevant set of attributes compared to the ground-truth. Our results show that the baseline outperformed concept-based explanations. Counterfactual explanations from an invertible neural network performed similarly as the baseline. Still, they allowed users to identify some attributes more accurately. Our results highlight the importance of measuring how well users can reason about biases of a model, rather than solely relying on technical evaluations or proxy tasks. We open-source our study and dataset so it can serve as a blue-print for future studies. For code see, https://github.com/berleon/do_users_benefit_from_interpretable_vision, Comment: Published at ICLR 2022

A variety of agent-based models has been proposed to account for the emergence of coordinated collective behavior of animal groups from simple interaction rules. A common, simplifying assumption of such collective movement models, is the consideration of individual agents moving with a constant speed. In this work we critically re-asses this assumption underlying a vast majority of collective movement models. First, we show the omnipresent speed variability observed in different species of live fish and artificial agents (RoboFish). Based on theoretical considerations accounting for inertia and rotational friction, we derive a functional dependence of the turning response of individuals on their instantaneous speed (confirmed by experimental data). We investigate how the interplay of variable speed and speed-dependent turning affects self-organized collective behavior by implementing an agent-based model which accounts for both effects. We show, that besides average speed, the individual speed variability may have a dramatic impact on the emergent collective dynamics, as two groups differing only in their speed variability, and being otherwise identical in all other behavioral parameters, can be in two fundamentally different stationary states. We find that the local coupling between group polarization and individual speed is strongest at the order-disorder transition. Furthermore, we demonstrate a decrease in polarization with group size for groups of individuals with variable speed, and a sudden decrease in mean individual speed at a critical group size (N=4 for Voronoi interactions) linked to a topological transition from an all-to-all to a distributed spatial interaction network. Overall, our work highlights the importance to account for fundamental kinematic constraints in general, and variable speed in particular, when modeling self-organized collective dynamics., Comment: 16 pages, 6 figures

Collective motion is commonly modeled with simple interaction rules between agents. Yet in nature, numerous observables vary within and between individuals and it remains largely unknown how animals respond to this variability, and how much of it may be the result of social responses. Here, we hypothesize that Guppies (\textit{Poecilia reticulata}) respond to avoidance behaviors of their shoal mates and that "socially competent" responses allow them to be more effective leaders. We test this hypothesis in an experimental setting in which a robotic Guppy, called RoboFish, is programmed to adapt to avoidance reactions of its live interaction partner. We compare the leadership performance between socially competent robots and two non-competent control behaviors and find that 1) behavioral variability itself appears attractive and that socially competent robots are better leaders that 2) require fewer approach attempts to 3) elicit longer average following behavior than non-competent agents. This work provides evidence that social responsiveness to avoidance reactions plays a role in the social dynamics of guppies. We showcase how social responsiveness can be modeled and tested directly embedded in a living animal model using adaptive, interactive robots.

Attribution methods provide insights into the decision-making of machine learning models like artificial neural networks. For a given input sample, they assign a relevance score to each individual input variable, such as the pixels of an image. In this work we adapt the information bottleneck concept for attribution. By adding noise to intermediate feature maps we restrict the flow of information and can quantify (in bits) how much information image regions provide. We compare our method against ten baselines using three different metrics on VGG-16 and ResNet-50, and find that our methods outperform all baselines in five out of six settings. The method's information-theoretic foundation provides an absolute frame of reference for attribution values (bits) and a guarantee that regions scored close to zero are not necessary for the network's decision. For reviews: https://openreview.net/forum?id=S1xWh1rYwB For code: https://github.com/BioroboticsLab/IBA, Comment: 18 pages, 12 figures, accepted at ICLR 2020 (Oral)

Attribution methods aim to explain a neural network's prediction by highlighting the most relevant image areas. A popular approach is to backpropagate (BP) a custom relevance score using modified rules, rather than the gradient. We analyze an extensive set of modified BP methods: Deep Taylor Decomposition, Layer-wise Relevance Propagation (LRP), Excitation BP, PatternAttribution, DeepLIFT, Deconv, RectGrad, and Guided BP. We find empirically that the explanations of all mentioned methods, except for DeepLIFT, are independent of the parameters of later layers. We provide theoretical insights for this surprising behavior and also analyze why DeepLIFT does not suffer from this limitation. Empirically, we measure how information of later layers is ignored by using our new metric, cosine similarity convergence (CSC). The paper provides a framework to assess the faithfulness of new and existing modified BP methods theoretically and empirically. For code see: https://github.com/berleon/when-explanations-lie, Comment: Published in ICML 2020 - Updated Proof

Over the last decades, honeybees have been a fascinating model to study insect navigation. While there is some controversy about the complexity of underlying neural correlates, the research of honeybee navigation makes progress through both the analysis of flight behavior and the synthesis of agent models. Since visual cues are believed to play a crucial role for the behavioral output of a navigating bee we have developed a realistic 3-dimensional virtual world, in which simulated agents can be tested, or in which the visual input of experimentally traced animals can be reconstructed. In this paper we present implementation details on how we reconstructed a large 3-dimensional world from aerial imagery of one of our field sites, how the distribution of ommatidia and their view geometry was modeled, and how the system samples from the scene to obtain realistic bee views. This system is made available as an open-source project to the community on \url{http://github.com/bioroboticslab/bee_view}.

The study of animal behavior increasingly relies on (semi-) automatic methods for the extraction of relevant behavioral features from video or picture data. To date, several specialized software products exist to detect and track animals' positions in simple (laboratory) environments. Tracking animals in their natural environments, however, often requires substantial customization of the image processing algorithms to the problem-specific image characteristics. Here we introduce BioTracker, an open-source computer vision framework, that provides programmers with core functionalities that are essential parts of a tracking software, such as video I/O, graphics overlays and mouse and keyboard interfaces. BioTracker additionally provides a number of different tracking algorithms suitable for a variety of image recording conditions. The main feature of BioTracker is however the straightforward implementation of new problem-specific tracking modules and vision algorithms that can build upon BioTracker's core functionalities. With this open-source framework the scientific community can accelerate their research and focus on the development of new vision algorithms., Comment: 4 pages, 2 figures

The honey bee dance communication system is one of the most popular examples of animal communication. Forager bees communicate the flight vector towards food, water, or resin sources to nestmates by performing a stereotypical motion pattern on the comb surface in the darkness of the hive. Bees that actively follow the circles of the dancer, so called dance-followers, may decode the message and fly according to the indicated vector that refers to the sun compass and their visual odometer. We investigated the dance communication system with a honeybee robot that reproduced the waggle dance pattern for a flight vector chosen by the experimenter. The dancing robot, called RoboBee, generated multiple cues contained in the biological dance pattern and elicited natural dance-following behavior in live bees. By tracking the flight trajectory of departing bees after following the dancing robot via harmonic radar we confirmed that bees used information obtained from the robotic dance to adjust their flight path. This is the first report on successful dance following and subsequent flight performance of bees recruited by a biomimetic robot.

The honeybee is a fascinating model animal to investigate how collective behavior emerges from (inter-)actions of thousands of individuals. Bees may acquire unique memories throughout their lives. These experiences affect social interactions even over large time frames. Tracking and identifying all bees in the colony over their lifetimes therefore may likely shed light on the interplay of individual differences and colony behavior. This paper proposes a software pipeline based on two deep convolutional neural networks for the localization and decoding of custom binary markers that honeybees carry from their first to the last day in their life. We show that this approach outperforms similar systems proposed in recent literature. By opening this software for the public, we hope that the resulting datasets will help advancing the understanding of honeybee collective intelligence.

Computational approaches to the analysis of collective behavior in social insects increasingly rely on motion paths as an intermediate data layer from which one can infer individual behaviors or social interactions. Honey bees are a popular model for learning and memory. Previous experience has been shown to affect and modulate future social interactions. So far, no lifetime history observations have been reported for all bees of a colony. In a previous work we introduced a tracking system customized to track up to $4000$ bees over several weeks. In this contribution we present an in-depth description of the underlying multi-step algorithm which both produces the motion paths, and also improves the marker decoding accuracy significantly. We automatically tracked ${\sim}2000$ marked honey bees over 10 weeks with inexpensive recording hardware using markers without any error correction bits. We found that the proposed two-step tracking reduced incorrect ID decodings from initially ${\sim}13\%$ to around $2\%$ post-tracking. Alongside this paper, we publish the first trajectory dataset for all bees in a colony, extracted from ${\sim} 4$ million images. We invite researchers to join the collective scientific effort to investigate this intriguing animal system. All components of our system are open-source.

The waggle dance is one of the most popular examples of animal communication. Forager bees direct their nestmates to profitable resources via a complex motor display. Essentially, the dance encodes the polar coordinates to the resource in the field. Unemployed foragers follow the dancer's movements and then search for the advertised spots in the field. Throughout the last decades, biologists have employed different techniques to measure key characteristics of the waggle dance and decode the information it conveys. Early techniques involved the use of protractors and stopwatches to measure the dance orientation and duration directly from the observation hive. Recent approaches employ digital video recordings and manual measurements on screen. However, manual approaches are very time-consuming. Most studies, therefore, regard only small numbers of animals in short periods of time. We have developed a system capable of automatically detecting, decoding and mapping communication dances in real-time. In this paper, we describe our recording setup, the image processing steps performed for dance detection and decoding and an algorithm to map dances to the field. The proposed system performs with a detection accuracy of 90.07\%. The decoded waggle orientation has an average error of -2.92{\deg} ($\pm$ 7.37{\deg} ), well within the range of human error. To evaluate and exemplify the system's performance, a group of bees was trained to an artificial feeder, and all dances in the colony were automatically detected, decoded and mapped. The system presented here is the first of this kind made publicly available, including source code and hardware specifications. We hope this will foster quantitative analyses of the honey bee waggle dance., Comment: 16 pages, LaTeX; a new value for the ratio distance-waggle run duration was computed. Figure 2 has been updated and discussion section was improved

Deep Convolutional Neuronal Networks (DCNNs) are showing remarkable performance on many computer vision tasks. Due to their large parameter space, they require many labeled samples when trained in a supervised setting. The costs of annotating data manually can render the use of DCNNs infeasible. We present a novel framework called RenderGAN that can generate large amounts of realistic, labeled images by combining a 3D model and the Generative Adversarial Network framework. In our approach, image augmentations (e.g. lighting, background, and detail) are learned from unlabeled data such that the generated images are strikingly realistic while preserving the labels known from the 3D model. We apply the RenderGAN framework to generate images of barcode-like markers that are attached to honeybees. Training a DCNN on data generated by the RenderGAN yields considerably better performance than training it on various baselines.