Arnab Sharma

Yasir Mahmood, Arnab Sharma, Axel-Cyrille Ngonga Ngomo et al.

There has been a growing interest in explaining entailments over description logic (DL) knowledge bases. The existing explanation formalisms focus on justifications to explain true axioms, and abductive reasoning to explain missing axioms in a knowledge base. However, these formalisms only point out the reasoning steps behind a (missing) entailment and lack a user-centered approach as they do not consider an inquirer's needs, level of understanding, or prior knowledge. We propose contrastive explanations, aiming at answering "why an axiom P (fact) is true instead of another axiom Q (foil)" over description logic knowledge bases. The motivation arises from the observation that when a user discovers that P has occurred, they are often surprised because they anticipated the occurrence of another similar event Q. Furthermore, individual explanations for "why P" and "why not Q" are unsatisfactory since a user expects to see the difference between P and Q. In this work, we first present formal foundations of contrasting questions and then define contrastive explanations within description logics. To this end, facts include ABox assertions of the form C(x) for a concept C and individual x. Possible foils for such facts are assertions C(y) (contrasting against an individual y), or D(x) (contrasting against a concept D). Additionally, we explore the properties of contrastive explanations in the DL EL and ALC. We also provide an implementation of our definition and an experimental evaluation on KBs of varying sizes.

Sourabh Kapoor, Arnab Sharma, Michael Röder et al.

Knowledge Graph Embedding (KGE) transforms a discrete Knowledge Graph (KG) into a continuous vector space facilitating its use in various AI-driven applications like Semantic Search, Question Answering, or Recommenders. While KGE approaches are effective in these applications, most existing approaches assume that all information in the given KG is correct. This enables attackers to influence the output of these approaches, e.g., by perturbing the input. Consequently, the robustness of such KGE approaches has to be addressed. Recent work focused on adversarial attacks. However, non-adversarial attacks on all attack surfaces of these approaches have not been thoroughly examined. We close this gap by evaluating the impact of non-adversarial attacks on the performance of 5 state-of-the-art KGE algorithms on 5 datasets with respect to attacks on 3 attack surfaces-graph, parameter, and label perturbation. Our evaluation results suggest that label perturbation has a strong effect on the KGE performance, followed by parameter perturbation with a moderate and graph with a low effect.

Fadel Mamar Seydou, Arnab Sharma

In recent years, machine learning (ML) based software systems are increasingly deployed in several critical applications, yet systematic testing of their behavior remains challenging due to complex model architectures, large input spaces, and evolving deployment environments. Existing testing approaches often rely on generating test cases based on given requirements, which often fail to reveal critical bugs of modern ML models due to their complex nature. Most importantly, such approaches, although they can be used to detect the presence of specific failures in the ML software, they hardly provide any message as to how to fix such errors. To tackle this, in this paper, we present DeepFix, a tool for automated testing of the entire ML pipeline using an agentic AI framework. Our testing approach first leverages Deepchecks to test the ML software for any potential bugs, and thereafter, uses an agentic AI-based approach to generate a detailed bug report. This includes a ranking, based on the severity of the found bugs, along with their explanations, which can be interpreted easily by any non-data science experts and most importantly, also provides possible ways to fix these bugs. Additionally, DeepFix supports several types of ML software systems and can be integrated easily to any ML workflow, enabling continuous testing throughout the development lifecycle. We discuss our already validated cases as well as some planned validations designed to demonstrate how the agentic testing process can reveal hidden failure modes that remain undetected by conventional testing methods. A 5-minute screencast demonstrating the tool's core functionality is available at https://youtu.be/WfwZmFcQgBQ.

Daniel Leite, Alisson Silva, Gabriella Casalino et al.

We introduce a modified incremental learning algorithm for evolving Granular Neural Network Classifiers (eGNN-C+). We use double-boundary hyper-boxes to represent granules, and customize the adaptation procedures to enhance the robustness of outer boxes for data coverage and noise suppression, while ensuring that inner boxes remain flexible to capture drifts. The classifier evolves from scratch, incorporates new classes on the fly, and performs local incremental feature weighting. As an application, we focus on the classification of emotion-related patterns within electroencephalogram (EEG) signals. Emotion recognition is crucial for enhancing the realism and interactivity of computer systems. We extract features from the Fourier spectrum of EEG signals obtained from 28 individuals engaged in playing computer games -- a public dataset. Each game elicits a different predominant emotion: boredom, calmness, horror, or joy. We analyze individual electrodes, time window lengths, and frequency bands to assess the accuracy and interpretability of resulting user-independent neural models. The findings indicate that both brain hemispheres assist classification, especially electrodes on the temporal (T8) and parietal (P7) areas, alongside contributions from frontal and occipital electrodes. While patterns may manifest in any band, the Alpha (8-13Hz), Delta (1-4Hz), and Theta (4-8Hz) bands, in this order, exhibited higher correspondence with the emotion classes. The eGNN-C+ demonstrates effectiveness in learning EEG data. It achieves an accuracy of 81.7% and a 0.0029 II interpretability using 10-second time windows, even in face of a highly-stochastic time-varying 4-class classification problem.

Arnab Sharma

Neural Encoders are frequently used in the NLP domain to perform dense retrieval tasks, for instance, to generate the candidate documents for a given query in question-answering tasks. However, sparse annotation and label noise in the training data make it challenging to train or fine-tune such retrieval models. Although existing works have attempted to mitigate these problems by incorporating modified loss functions or data cleaning, these approaches either require some hyperparameters to tune during training or add substantial complexity to the training setup. In this work, we consider a label weakening approach to generate robust retrieval models in the presence of label noise. Instead of enforcing a single, potentially erroneous label for each query document pair, we allow for a set of plausible labels derived from both the observed supervision and the model's confidence scores. We perform an extensive evaluation considering two retrieval models, one re-ranking model, considering four diverse ranking datasets. To this end, we also consider a realistic noisy setting by using a semantic-aware noise generation technique to generate different ratios of noise. Our initial results show that label weakening can improve the performance of the retrieval tasks in comparison to 10 different state-of-the-art loss functions.

Rupesh Sapkota, Caglar Demir, Arnab Sharma et al.

Ensemble methods are widely employed to improve generalization in machine learning. This has also prompted the adoption of ensemble learning for the knowledge graph embedding (KGE) models in performing link prediction. Typical approaches to this end train multiple models as part of the ensemble, and the diverse predictions are then averaged. However, this approach has some significant drawbacks. For instance, the computational overhead of training multiple models increases latency and memory overhead. In contrast, model merging approaches offer a promising alternative that does not require training multiple models. In this work, we introduce model merging, specifically weighted averaging, in KGE models. Herein, a running average of model parameters from a training epoch onward is maintained and used for predictions. To address this, we additionally propose an approach that selectively updates the running average of the ensemble model parameters only when the generalization performance improves on a validation dataset. We evaluate these two different weighted averaging approaches on link prediction tasks, comparing the state-of-the-art benchmark ensemble approach. Additionally, we evaluate the weighted averaging approach considering literal-augmented KGE models and multi-hop query answering tasks as well. The results demonstrate that the proposed weighted averaging approach consistently improves performance across diverse evaluation settings.

Ding Ning, Varvara Vetrova, Karin R. Bryan et al.

This overview paper details the findings from the Diving Deep: Forecasting Sea Surface Temperatures and Anomalies Challenge at the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) 2024. The challenge focused on the data-driven predictability of global sea surface temperatures (SSTs), a key factor in climate forecasting, ecosystem management, fisheries management, and climate change monitoring. The challenge involved forecasting SST anomalies (SSTAs) three months in advance using historical data and included a special task of predicting SSTAs nine months ahead for the Baltic Sea. Participants utilized various machine learning approaches to tackle the task, leveraging data from ERA5. This paper discusses the methodologies employed, the results obtained, and the lessons learned, offering insights into the future of climate-related predictive modeling.

Arnab Sharma, N'Dah Jean Kouagou, Axel-Cyrille Ngonga Ngomo

In recent years, knowledge graphs have gained interest and witnessed widespread applications in various domains, such as information retrieval, question-answering, recommendation systems, amongst others. Large-scale knowledge graphs to this end have demonstrated their utility in effectively representing structured knowledge. To further facilitate the application of machine learning techniques, knowledge graph embedding (KGE) models have been developed. Such models can transform entities and relationships within knowledge graphs into vectors. However, these embedding models often face challenges related to noise, missing information, distribution shift, adversarial attacks, etc. This can lead to sub-optimal embeddings and incorrect inferences, thereby negatively impacting downstream applications. While the existing literature has focused so far on adversarial attacks on KGE models, the challenges related to the other critical aspects remain unexplored. In this paper, we, first of all, give a unified definition of resilience, encompassing several factors such as generalisation, performance consistency, distribution adaption, and robustness. After formalizing these concepts for machine learning in general, we define them in the context of knowledge graphs. To find the gap in the existing works on resilience in the context of knowledge graphs, we perform a systematic survey, taking into account all these aspects mentioned previously. Our survey results show that most of the existing works focus on a specific aspect of resilience, namely robustness. After categorizing such works based on their respective aspects of resilience, we discuss the challenges and future research directions.

Caglar Demir, Arnab Sharma, Axel-Cyrille Ngonga Ngomo

Ensemble models often improve generalization performances in challenging tasks. Yet, traditional techniques based on prediction averaging incur three well-known disadvantages: the computational overhead of training multiple models, increased latency, and memory requirements at test time. To address these issues, the Stochastic Weight Averaging (SWA) technique maintains a running average of model parameters from a specific epoch onward. Despite its potential benefits, maintaining a running average of parameters can hinder generalization, as an underlying running model begins to overfit. Conversely, an inadequately chosen starting point can render SWA more susceptible to underfitting compared to an underlying running model. In this work, we propose Adaptive Stochastic Weight Averaging (ASWA) technique that updates a running average of model parameters, only when generalization performance is improved on the validation dataset. Hence, ASWA can be seen as a combination of SWA with the early stopping technique, where the former accepts all updates on a parameter ensemble model and the latter rejects any update on an underlying running model. We conducted extensive experiments ranging from image classification to multi-hop reasoning over knowledge graphs. Our experiments over 11 benchmark datasets with 7 baseline models suggest that ASWA leads to a statistically better generalization across models and datasets

Arnab Sharma, Caglar Demir, Axel-Cyrille Ngonga Ngomo et al.

In recent years, we observe an increasing amount of software with machine learning components being deployed. This poses the question of quality assurance for such components: how can we validate whether specified requirements are fulfilled by a machine learned software? Current testing and verification approaches either focus on a single requirement (e.g., fairness) or specialize on a single type of machine learning model (e.g., neural networks). In this paper, we propose property-driven testing of machine learning models. Our approach MLCheck encompasses (1) a language for property specification, and (2) a technique for systematic test case generation. The specification language is comparable to property-based testing languages. Test case generation employs advanced verification technology for a systematic, property-dependent construction of test suites, without additional user-supplied generator functions. We evaluate MLCheck using requirements and data sets from three different application areas (software discrimination, learning on knowledge graphs and security). Our evaluation shows that despite its generality MLCheck can even outperform specialised testing approaches while having a comparable runtime.

Arnab Sharma, Heike Wehrheim

Today, machine learning (ML) models are increasingly applied in decision making. This induces an urgent need for quality assurance of ML models with respect to (often domain-dependent) requirements. Monotonicity is one such requirement. It specifies a software as 'learned' by an ML algorithm to give an increasing prediction with the increase of some attribute values. While there exist multiple ML algorithms for ensuring monotonicity of the generated model, approaches for checking monotonicity, in particular of black-box models, are largely lacking. In this work, we propose verification-based testing of monotonicity, i.e., the formal computation of test inputs on a white-box model via verification technology, and the automatic inference of this approximating white-box model from the black-box model under test. On the white-box model, the space of test inputs can be systematically explored by a directed computation of test cases. The empirical evaluation on 90 black-box models shows verification-based testing can outperform adaptive random testing as well as property-based techniques with respect to effectiveness and efficiency.