Every month, we want to acknowledge some valuable TAILOR papers, selected among the papers published by scientists belonging to our network by TAILOR principal investigator Fredrik Heintz.
The list of the most valuable papers gathers contributions from different TAILOR partners, each providing valuable insights on different topics related to TrustworthyAI.
Stay tuned for other valuable insights and groundbreaking research from our diverse community!
AdaCL: Adaptive Continual Learning
E. C. G. Yildirim, M. O. Yildirim, M. Kilickaya, and J. Vanschoren
Proceedings of Machine Learning Research, 2023, pp. 15–24. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85203881017&partnerID=40&md5=ba3207dfe5144868fd5f9a155ae8afbe
Abstract: Class-Incremental Learning aims to update a deep classifier to learn new categories while maintaining or improving its accuracy on previously observed classes. Common methods to prevent forgetting previously learned classes include regularizing the neural network updates and storing exemplars in memory, which come with hyperparameters such as the learning rate, regularization strength, or the number of exemplars. However, these hyperparameters are usually only tuned at the start and then kept fixed throughout the learning sessions, ignoring the fact that newly encountered tasks may have varying levels of novelty or difficulty. This study investigates the necessity of hyperparameter ‘adaptivity’ in Class-Incremental Learning: the ability to dynamically adjust hyperparameters such as the learning rate, regularization strength, and memory size according to the properties of the new task at hand. We propose AdaCL, a Bayesian Optimization-based approach to automatically and efficiently determine the optimal values for those parameters with each learning task. We show that adapting hyperpararmeters on each new task leads to improvement in accuracy, forgetting and memory. Code is available at https://github.com/ElifCerenGokYildirim/AdaCL
Cutting the Black Box: Conceptual Interpretation of a Deep Neural Net with Multi-Modal Embeddings and Multi-Criteria Decision Aid
N. Atienza et al.
IJCAI International Joint Conference on Artificial Intelligence, 2024, pp. 3669–3678. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85204282712&partnerID=40&md5=ac85f7cac556fe5e2f34f3b4f2b007e6
Abstract: This paper tackles the concept-based explanation of neural models in computer vision, building upon the state of the art in Multi-Criteria Decision Aid (MCDA). The novelty of the approach is to leverage multi-modal embeddings from CLIP to bridge the gap between pixel-based and concept-based representations. The proposed Cut the Black Box (CB2) approach disentangles the latent representation of a trained pixel-based neural net, referred to as teacher model, along a 3-step process. Firstly, the pixel-based representation of the samples is mapped onto a conceptual representation using multi-modal embeddings. Secondly, an interpretable-by-design MCDA student model is trained by distillation from the teacher model using the conceptual sample representation. Thirdly, the alignment of the teacher and student latent representations spells out the concepts relevant to explaining the teacher model. The empirical validation of the approach on ResNet, VGG, and VisionTransformer on Cifar-10, Cifar-100, Tiny ImageNet, and Fashion-MNIST showcases the effectiveness of the interpretations provided for the teacher models. The analysis reveals that decision-making predominantly relies on few concepts, thereby exposing potential bias in the teacher’s decisions.
Do Llamas Work in English? On the Latent Language of Multilingual Transformers
C. Wendler, V. Veselovsky, G. Monea, and R. West
Proceedings of the Annual Meeting of the Association for Computational Linguistics, 2024, pp. 15366–15394. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85204439518&partnerID=40&md5=79544aca49d9ab545e6e471296dc8ca5
Abstract: We ask whether multilingual language models trained on unbalanced, English-dominated corpora use English as an internal pivot language-a question of key importance for understanding how language models function and the origins of linguistic bias. Focusing on the Llama-2 family of transformer models, our study uses carefully constructed non-English prompts with a unique correct single-token continuation. From layer to layer, transformers gradually map an input embedding of the final prompt token to an output embedding from which next-token probabilities are computed. Tracking intermediate embeddings through their high-dimensional space reveals three distinct phases, whereby intermediate embeddings (1) start far away from output token embeddings; (2) already allow for decoding a semantically correct next token in middle layers, but give higher probability to its version in English than in the input language; (3) finally move into an input-language-specific region of the embedding space. We cast these results into a conceptual model where the three phases operate in “input space”, “concept space”, and “output space”, respectively. Crucially, our evidence suggests that the abstract “concept space” lies closer to English than to other languages, which may have important consequences regarding the biases held by multilingual language models. Code and data is made available here: https://github.com/epfl-dlab/llm-latent-language
Graph2Tac: Online Representation Learning of Formal Math Concepts
L. Blaauwbroek, M. Olšák, J. Rute, F. I. S. Massolo, J. Piepenbrock, and V. Pestun
Proceedings of Machine Learning Research, 2024, pp. 4046–4076. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85203803136&partnerID=40&md5=a72a1de3a3bb7d75c5223a185fec6e70
Abstract:
Larger and more instructable language models become less reliable
L. Zhou, W. Schellaert, F. Martínez-Plumed, Y. Moros-Daval, C. Ferri, and J. Hernández-Orallo
Nature, vol. 634, no. 8032, pp. 61–68, 2024, doi: 10.1038/s41586-024-07930-y.
Abstract: In proof assistants, the physical proximity between two formal mathematical concepts is a strong predictor of their mutual relevance. Furthermore, lemmas with close proximity regularly exhibit similar proof structures. We show that this locality property can be exploited through online learning techniques to obtain solving agents that far surpass offline learners when asked to prove theorems in an unseen mathematical setting. We extensively benchmark two such online solvers implemented in the Tactician platform for the Coq proof assistant: First, Tactician’s online k-nearest neighbor solver, which can learn from recent proofs, shows a 1.72× improvement in theorems proved over an offline equivalent. Second, we introduce a graph neural network, Graph2Tac, with a novel approach to build hierarchical representations for new definitions. Graph2Tac’s online definition task realizes a 1.5× improvement in theorems solved over an offline baseline. The k-NN and Graph2Tac solvers rely on orthogonal online data, making them highly complementary. Their combination improves 1.27× over their individual performances. Both solvers outperform all other general-purpose provers for Coq, including CoqHammer, Proverbot9001, and a transformer baseline by at least 1.48× and are available for practical use by end-users.
Learning Structural Causal Models through Deep Generative Models: Methods, Guarantees, and Challenges
A. Poinsot, A. Leite, N. Chesneau, M. Sébag, and M. Schoenauer
IJCAI International Joint Conference on Artificial Intelligence, 2024, pp. 8207–8215. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85204297358&partnerID=40&md5=a7eb498bf9422e5fddcf96b917819712
Abstract: This paper provides a comprehensive review of deep structural causal models (DSCMs), particularly focusing on their ability to answer counterfactual queries using observational data within known causal structures. It delves into the characteristics of DSCMs by analyzing the hypotheses, guarantees, and applications inherent to the underlying deep learning components and structural causal models, fostering a finer understanding of their capabilities and limitations in addressing different counterfactual queries. Furthermore, it highlights the challenges and open questions in the field of deep structural causal modeling. It sets the stages for researchers to identify future work directions and for practitioners to get an overview in order to find out the most appropriate methods for their needs.
Learning to Solve Abstract Reasoning Problems with Neurosymbolic Program Synthesis and Task Generation
J. Bednarek and K. Krawiec
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2024, pp. 386–402. doi: 10.1007/978-3-031-71167-1_21.
Abstract: The ability to think abstractly and reason by analogy is a prerequisite to rapidly adapt to new conditions, tackle newly encountered problems by decomposing them, and synthesize knowledge to solve problems comprehensively. We present TransCoder, a method for solving abstract problems based on neural program synthesis, and conduct a comprehensive analysis of decisions made by the generative module of the proposed architecture. At the core of TransCoder is a typed domain-specific language, designed to facilitate feature engineering and abstract reasoning. In training, we use the programs that failed to solve tasks to generate new tasks and gather them in a synthetic dataset. As each synthetic task created in this way has a known associated program (solution), the model is trained on them in supervised mode. Solutions are represented in a transparent programmatic form, which can be inspected and verified. We demonstrate TransCoder ’s performance using the Abstract Reasoning Corpus dataset, for which our framework generates tens of thousands of synthetic problems with corresponding solutions and facilitates systematic progress in learning.
On the Hardness of Probabilistic Neurosymbolic Learning
J. Maene, V. Derkinderen, and L. De Raedt
Proceedings of Machine Learning Research, 2024, pp. 34203–34218. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85203816760&partnerID=40&md5=205a9ab1044a333e6ef43cd9de6149d7
Abstract: The limitations of purely neural learning have sparked an interest in probabilistic neurosymbolic models, which combine neural networks with probabilistic logical reasoning. As these neurosymbolic models are trained with gradient descent, we study the complexity of differentiating probabilistic reasoning. We prove that although approximating these gradients is intractable in general, it becomes tractable during training. Furthermore, we introduce WeightME, an unbiased gradient estimator based on model sampling. Under mild assumptions, WeightME approximates the gradient with probabilistic guarantees using a logarithmic number of calls to a SAT solver. Lastly, we evaluate the necessity of these guarantees on the gradient. Our experiments indicate that the existing biased approximations indeed struggle to optimize even when exact solving is still feasible.