Search

Transformer models are increasingly prevalent in various applications, yet our understanding of their internal workings remains limited. This paper investigates the modularity and task specialization of neurons within transformer architectures, focusing on both vision (ViT) and language (Mistral 7B) models. Using a combination of selective pruning and MoEfication clustering techniques, we analyze the overlap and specialization of neurons across different tasks and data subsets. Our findings reveal evidence of task-specific neuron clusters, with varying degrees of overlap between related tasks. We observe that neuron importance patterns persist to some extent even in randomly initialized models, suggesting an inherent structure that training refines. Additionally, we find that neuron clusters identified through MoEfication correspond more strongly to task-specific neurons in earlier and later layers of the models. This work contributes to a more nuanced understanding of transformer internals and offers insights into potential avenues for improving model interpretability and efficiency., Comment: 11 pages, 6 figures

The use of transformer-based models is growing rapidly throughout society. With this growth, it is important to understand how they work, and in particular, how the attention mechanisms represent concepts. Though there are many interpretability methods, many look at models through their neuronal activations, which are poorly understood. We describe different lenses through which to view neuron activations, and investigate the effectiveness in language models and vision transformers through various methods of neural ablation: zero ablation, mean ablation, activation resampling, and a novel approach we term 'peak ablation'. Through experimental analysis, we find that in different regimes and models, each method can offer the lowest degradation of model performance compared to other methods, with resampling usually causing the most significant performance deterioration. We make our code available at https://github.com/nickypro/investigating-ablation., Comment: 9 pages, 2 figures, XAI World Conference 2024 Late-Breaking Work

In the era of Big Data, scalable and accurate clustering algorithms for high-dimensional data are essential. We present new Bayesian Distance Clustering (BDC) models and inference algorithms with improved scalability while maintaining the predictive accuracy of modern Bayesian non-parametric models. Unlike traditional methods, BDC models the distance between observations rather than the observations directly, offering a compromise between the scalability of distance-based methods and the enhanced predictive power and probabilistic interpretation of model-based methods. However, existing BDC models still rely on performing inference on the partition model to group observations into clusters. The support of this partition model grows exponentially with the dataset's size, complicating posterior space exploration and leading to many costly likelihood evaluations. Inspired by K-medoids, we propose using tessellations in discrete space to simplify inference by focusing the learning task on finding the best tessellation centers, or "medoids." Additionally, we extend our models to effectively handle multi-view data, such as data comprised of clusters that evolve across time, enhancing their applicability to complex datasets. The real data application in numismatics demonstrates the efficacy of our approach., Comment: 45 pages, 11 figures, 7 tables

With the expansion of cities over time, URT (Urban Rail Transit) networks have also grown significantly. Demand prediction plays an important role in supporting planning, scheduling, fleet management, and other operational decisions. In this study, we propose an Origin-Destination (OD) demand prediction model called Multi-Graph Inductive Representation Learning (mGraphSAGE) for large-scale URT networks under operational uncertainties. Our main contributions are twofold: we enhance prediction results while ensuring scalability for large networks by relying simultaneously on multiple graphs, where each OD pair is a node on a graph and distinct OD relationships, such as temporal and spatial correlations; we show the importance of including operational uncertainties such as train delays and cancellations as inputs in demand prediction for daily operations. The model is validated on three different scales of the URT network in Copenhagen, Denmark. Experimental results show that by leveraging information from neighboring ODs and learning node representations via sampling and aggregation, mGraphSAGE is particularly suitable for OD demand prediction in large-scale URT networks, outperforming reference machine learning methods. Furthermore, during periods with train cancellations and delays, the performance gap between mGraphSAGE and other methods improves compared to normal operating conditions, demonstrating its ability to leverage system reliability information for predicting OD demand under uncertainty., Comment: 18 pages, 3 figures

Consider a binary classification problem solved using a feed-forward artificial neural network (ANN). Let the ANN be composed of a ReLU layer and several linear layers (convolution, sum-pooling, or fully connected). We assume the network was trained with high accuracy. Despite numerous suggested approaches, interpreting an artificial neural network remains challenging for humans. For a new method of interpretation, we construct a bridge between a simple ANN and logic. As a result, we can analyze and manipulate the semantics of an ANN using the powerful tool set of logic. To achieve this, we decompose the input space of the ANN into several network partition cells. Each network partition cell represents a linear combination that maps input values to a classifying output value. For interpreting the linear map of a partition cell using logic expressions, we suggest minterm values as the input of a simple ANN. We derive logic expressions representing interaction patterns for separating objects classified as 1 from those classified as 0. To facilitate an interpretation of logic expressions, we present them as binary logic trees.

This paper presents a novel approach to epidemic surveillance, leveraging the power of Artificial Intelligence and Large Language Models (LLMs) for effective interpretation of unstructured big data sources, like the popular ProMED and WHO Disease Outbreak News. We explore several LLMs, evaluating their capabilities in extracting valuable epidemic information. We further enhance the capabilities of the LLMs using in-context learning, and test the performance of an ensemble model incorporating multiple open-source LLMs. The findings indicate that LLMs can significantly enhance the accuracy and timeliness of epidemic modelling and forecasting, offering a promising tool for managing future pandemic events., Comment: 11 pages, 4 figures, Ninth International Congress on Information and Communication Technology (ICICT 2024)

Developmental dysgraphia is a neurological disorder that hinders children's writing skills. In recent years, researchers have increasingly explored machine learning methods to support the diagnosis of dysgraphia based on offline and online handwriting. In most previous studies, the two types of handwriting have been analysed separately, which does not necessarily lead to promising results. In this way, the relationship between online and offline data cannot be explored. To address this limitation, we propose a novel multimodal machine learning approach utilizing both online and offline handwriting data. We created a new dataset by transforming an existing online handwritten dataset, generating corresponding offline handwriting images. We considered only different types of word data (simple word, pseudoword & difficult word) in our multimodal analysis. We trained SVM and XGBoost classifiers separately on online and offline features as well as implemented multimodal feature fusion and soft-voted ensemble. Furthermore, we proposed a novel ensemble with conditional feature fusion method which intelligently combines predictions from online and offline classifiers, selectively incorporating feature fusion when confidence scores fall below a threshold. Our novel approach achieves an accuracy of 88.8%, outperforming SVMs for single modalities by 12-14%, existing methods by 8-9%, and traditional multimodal approaches (soft-vote ensemble and feature fusion) by 3% and 5%, respectively. Our methodology contributes to the development of accurate and efficient dysgraphia diagnosis tools, requiring only a single instance of multimodal word/pseudoword data to determine the handwriting impairment. This work highlights the potential of multimodal learning in enhancing dysgraphia diagnosis, paving the way for accessible and practical diagnostic tools.

Neural networks are trained to learn an approximate mapping from an input domain to a target domain. Incorporating prior knowledge about true mappings is critical to learning a useful approximation. With current architectures, it is challenging to enforce structure on the derivatives of the input-output mapping. We propose to use a neural network to directly learn the Jacobian of the input-output function, which allows easy control of the derivative. We focus on structuring the derivative to allow invertibility and also demonstrate that other useful priors, such as $k$-Lipschitz, can be enforced. Using this approach, we can learn approximations to simple functions that are guaranteed to be invertible and easily compute the inverse. We also show similar results for 1-Lipschitz functions., Comment: 6 pages, 3 Figures, ICML 2019 INNF Workshop

This paper introduces a novel supervised classification strategy that integrates functional data analysis (FDA) with tree-based methods, addressing the challenges of high-dimensional data and enhancing the classification performance of existing functional classifiers. Specifically, we propose augmented versions of functional classification trees and functional random forests, incorporating a new tool for assessing the importance of functional principal components. This tool provides an ad-hoc method for determining unbiased permutation feature importance in functional data, particularly when dealing with correlated features derived from successive derivatives. Our study demonstrates that these additional features can significantly enhance the predictive power of functional classifiers. Experimental evaluations on both real-world and simulated datasets showcase the effectiveness of the proposed methodology, yielding promising results compared to existing methods., Comment: 33 pages

The advent of big data has raised significant challenges in analysing high-dimensional datasets across various domains such as medicine, ecology, and economics. Functional Data Analysis (FDA) has proven to be a robust framework for addressing these challenges, enabling the transformation of high-dimensional data into functional forms that capture intricate temporal and spatial patterns. However, despite advancements in functional classification methods and very high performance demonstrated by combining FDA and ensemble methods, a critical gap persists in the literature concerning the transparency and interpretability of black-box models, e.g. Functional Random Forests (FRF). In response to this need, this paper introduces a novel suite of explainability tools to illuminate the inner mechanisms of FRF. We propose using Functional Partial Dependence Plots (FPDPs), Functional Principal Component (FPC) Probability Heatmaps, various model-specific and model-agnostic FPCs' importance metrics, and the FPC Internal-External Importance and Explained Variance Bubble Plot. These tools collectively enhance the transparency of FRF models by providing a detailed analysis of how individual FPCs contribute to model predictions. By applying these methods to an ECG dataset, we demonstrate the effectiveness of these tools in revealing critical patterns and improving the explainability of FRF., Comment: 33 pages

Evolving networks are complex data structures that emerge in a wide range of systems in science and engineering. Learning expressive representations for such networks that encode their structural connectivity and temporal evolution is essential for downstream data analytics and machine learning applications. In this study, we introduce a self-supervised method for learning representations of temporal networks and employ these representations in the dynamic link prediction task. While temporal networks are typically characterized as a sequence of interactions over the continuous time domain, our study focuses on their discrete-time versions. This enables us to balance the trade-off between computational complexity and precise modeling of the interactions. We propose a recurrent message-passing neural network architecture for modeling the information flow over time-respecting paths of temporal networks. The key feature of our method is the contrastive training objective of the model, which is a combination of three loss functions: link prediction, graph reconstruction, and contrastive predictive coding losses. The contrastive predictive coding objective is implemented using infoNCE losses at both local and global scales of the input graphs. We empirically show that the additional self-supervised losses enhance the training and improve the model's performance in the dynamic link prediction task. The proposed method is tested on Enron, COLAB, and Facebook datasets and exhibits superior results compared to existing models., Comment: 16 pages, 6 figures

Predictive coding (PC) is an energy-based learning algorithm that performs iterative inference over network activities before weight updates. Recent work suggests that PC can converge in fewer learning steps than backpropagation thanks to its inference procedure. However, these advantages are not always observed, and the impact of PC inference on learning is theoretically not well understood. Here, we study the geometry of the PC energy landscape at the (inference) equilibrium of the network activities. For deep linear networks, we first show that the equilibrated energy is simply a rescaled mean squared error loss with a weight-dependent rescaling. We then prove that many highly degenerate (non-strict) saddles of the loss including the origin become much easier to escape (strict) in the equilibrated energy. Our theory is validated by experiments on both linear and non-linear networks. Based on these results, we conjecture that all the saddles of the equilibrated energy are strict. Overall, this work suggests that PC inference makes the loss landscape more benign and robust to vanishing gradients, while also highlighting the challenge of speeding up PC inference on large-scale models., Comment: 26 pages, 12 figures

Reinforcement Learning (RL) has experienced significant advancement over the past decade, prompting a growing interest in applications within finance. This survey critically evaluates 167 publications, exploring diverse RL applications and frameworks in finance. Financial markets, marked by their complexity, multi-agent nature, information asymmetry, and inherent randomness, serve as an intriguing test-bed for RL. Traditional finance offers certain solutions, and RL advances these with a more dynamic approach, incorporating machine learning methods, including transfer learning, meta-learning, and multi-agent solutions. This survey dissects key RL components through the lens of Quantitative Finance. We uncover emerging themes, propose areas for future research, and critique the strengths and weaknesses of existing methods., Comment: This work is currently submitted to and under-review for ACM Computing Surveys. This copy is an unedited, pre-print version and it is the author's version of the work. I

We lay the foundations for a database-inspired approach to interpreting and understanding neural network models by querying them using declarative languages. Towards this end we study different query languages, based on first-order logic, that mainly differ in their access to the neural network model. First-order logic over the reals naturally yields a language which views the network as a black box; only the input--output function defined by the network can be queried. This is essentially the approach of constraint query languages. On the other hand, a white-box language can be obtained by viewing the network as a weighted graph, and extending first-order logic with summation over weight terms. The latter approach is essentially an abstraction of SQL. In general, the two approaches are incomparable in expressive power, as we will show. Under natural circumstances, however, the white-box approach can subsume the black-box approach; this is our main result. We prove the result concretely for linear constraint queries over real functions definable by feedforward neural networks with a fixed number of hidden layers and piecewise linear activation functions., Comment: To appear at ICDT 2025

This study investigates the meta-issues surrounding social media, which, while theoretically designed to enhance social interactions and improve our social lives by facilitating the sharing of personal experiences and life events, often results in adverse psychological impacts. Our investigation reveals a paradoxical outcome: rather than fostering closer relationships and improving social lives, the algorithms and structures that underlie social media platforms inadvertently contribute to a profound psychological impact on individuals, influencing them in unforeseen ways. This phenomenon is particularly pronounced among teenagers, who are disproportionately affected by curated online personas, peer pressure to present a perfect digital image, and the constant bombardment of notifications and updates that characterize their social media experience. As such, we issue a call to action for policymakers, platform developers, and educators to prioritize the well-being of teenagers in the digital age and work towards creating secure and safe social media platforms that protect the young from harm, online harassment, and exploitation., Comment: 7 pages, 0 figures, 2 tables, 2024 IEEE Conference on Digital Platforms and Societal Harms

Continual learning enables AI models to learn new data sequentially without retraining in real-world scenarios. Most existing methods assume the training data are balanced, aiming to reduce the catastrophic forgetting problem that models tend to forget previously generated data. However, data imbalance and the mixture of new and old data in real-world scenarios lead the model to ignore categories with fewer training samples. To solve this problem, we propose an analytic imbalance rectifier algorithm (AIR), a novel online exemplar-free continual learning method with an analytic (i.e., closed-form) solution for data-imbalanced class-incremental learning (CIL) and generalized CIL scenarios in real-world continual learning. AIR introduces an analytic re-weighting module (ARM) that calculates a re-weighting factor for each class for the loss function to balance the contribution of each category to the overall loss and solve the problem of imbalanced training data. AIR uses the least squares technique to give a non-discriminatory optimal classifier and its iterative update method in continual learning. Experimental results on multiple datasets show that AIR significantly outperforms existing methods in long-tailed and generalized CIL scenarios. The source code is available at https://github.com/fang-d/AIR.

Poverty is a serious issue that harms humanity progression. The simplest solution is to use one-shirt-size policy to alleviate it. Nevertheless, each region has its unique issues, which require a unique solution to solve them. In the aspect of spatial analysis, neighbor regions can provide useful information to analyze issues of a given region. In this work, we proposed inferred boundaries of regions of Thailand that can explain better the poverty dynamics, instead of the usual government administrative regions. The proposed regions maximize a trade-off between poverty-related features and geographical coherence. We use a spatial analysis together with Moran's cluster algorithms and Bayesian hierarchical regression models, with the potential of assist the implementation of the right policy to alleviate the poverty phenomenon. We found that all variables considered show a positive spatial autocorrelation. The results of analysis illustrate that 1) Northern, Northeastern Thailand, and in less extend Northcentral Thailand are the regions that require more attention in the aspect of poverty issues, 2) Northcentral, Northeastern, Northern and Southern Thailand present dramatically low levels of education, income and amount of savings contrasted with large cities such as Bangkok-Pattaya and Central Thailand, and 3) Bangkok-Pattaya is the only region whose average years of education is above 12 years, which corresponds (approx.) with a complete senior high school., Comment: Codes to reproduce our results are available in https://github.com/IrvingGomez/SpatialPovertyFactors

Concept-based interpretability methods are a popular form of explanation for deep learning models which provide explanations in the form of high-level human interpretable concepts. These methods typically find concept activation vectors (CAVs) using a probe dataset of concept examples. This requires labelled data for these concepts -- an expensive task in the medical domain. We introduce TextCAVs: a novel method which creates CAVs using vision-language models such as CLIP, allowing for explanations to be created solely using text descriptions of the concept, as opposed to image exemplars. This reduced cost in testing concepts allows for many concepts to be tested and for users to interact with the model, testing new ideas as they are thought of, rather than a delay caused by image collection and annotation. In early experimental results, we demonstrate that TextCAVs produces reasonable explanations for a chest x-ray dataset (MIMIC-CXR) and natural images (ImageNet), and that these explanations can be used to debug deep learning-based models., Comment: 11 pages, 2 figures. Accepted at iMIMIC Workshop at MICCAI 2024

The human brain undergoes rapid development during the third trimester of pregnancy. In this work, we model the neonatal development of the infant brain in this age range. As a basis, we use MR images of preterm- and term-birth neonates from the developing human connectome project (dHCP). We propose a neural network, specifically an implicit neural representation (INR), to predict 2D- and 3D images of varying time points. In order to model a subject-specific development process, it is necessary to disentangle the age from the subjects' identity in the latent space of the INR. We propose two methods, Subject Specific Latent Vectors (SSL) and Stochastic Global Latent Augmentation (SGLA), enabling this disentanglement. We perform an analysis of the results and compare our proposed model to an age-conditioned denoising diffusion model as a baseline. We also show that our method can be applied in a memory-efficient way, which is especially important for 3D data., Comment: Preprint, Accepted for PRIME MICCAI 2024

In this paper, we propose a novel learning-based framework for non-rigid point cloud matching, which can be trained purely on point clouds without any correspondence annotation but also be extended naturally to partial-to-full matching. Our key insight is to incorporate semantic features derived from large vision models (LVMs) to geometry-based shape feature learning. Our framework effectively leverages the structural information contained in the semantic features to address ambiguities arise from self-similarities among local geometries. Furthermore, our framework also enjoys the strong generalizability and robustness regarding partial observations of LVMs, leading to improvements in the regarding point cloud matching tasks. In order to achieve the above, we propose a pixel-to-point feature aggregation module, a local and global attention network as well as a geometrical similarity loss function. Experimental results show that our method achieves state-of-the-art results in matching non-rigid point clouds in both near-isometric and heterogeneous shape collection as well as more realistic partial and noisy data., Comment: 12 pages, 4 figures

In the ever-evolving field of artificial neural networks and learning systems, complex-valued neural networks (CVNNs) have become a cornerstone, achieving exceptional performance in image processing and telecommunications. More precisely, in digital communication systems, CVNNs have been delivering significant results in tasks like equalization, channel estimation, beamforming, and decoding. Among the CVNN architectures, the complex-valued radial basis function neural network (C-RBF) stands out, especially when operating in noisy environments such as 5G multiple-input multiple-output (MIMO) systems. In such a context, this paper extends the classical shallow C-RBF to deep architectures, increasing its flexibility for a wider range of applications. Also, based on the parameter selection of the phase transmittance radial basis function (PT-RBF) neural network, we propose an initialization scheme for the deep C-RBF. Via rigorous simulations conforming to 3GPP TS 38 standards for digital communications, our method not only outperforms conventional initialization strategies like random, $K$-means, and constellation-based methods but it also seems to be the only approach to achieve successful convergence for deep C-RBF architectures. These findings pave the way to more robust and efficient neural network deployments in complex-valued digital communication systems., Comment: 19th International Symposium on Wireless Communication Systems (ISWCS'24). arXiv admin note: substantial text overlap with arXiv:2408.07692

Background/introduction: Pre-trained transformer models shine in many natural language processing tasks and therefore are expected to bear the representation of the input sentence or text meaning. These sentence-level embeddings are also important in retrieval-augmented generation. But do commonly used plain averaging or prompt templates surface it enough? Methods: Given 110M parameters BERT's hidden representations from multiple layers and multiple tokens we tried various ways to extract optimal sentence representations. We tested various token aggregation and representation post-processing techniques. We also tested multiple ways of using a general Wikitext dataset to complement BERTs sentence representations. All methods were tested on 8 Semantic Textual Similarity (STS), 6 short text clustering, and 12 classification tasks. We also evaluated our representation-shaping techniques on other static models, including random token representations. Results: Proposed representation extraction methods improved the performance on STS and clustering tasks for all models considered. Very high improvements for static token-based models, especially random embeddings for STS tasks almost reach the performance of BERT-derived representations. Conclusions: Our work shows that for multiple tasks simple baselines with representation shaping techniques reach or even outperform more complex BERT-based models or are able to contribute to their performance.

Graph neural networks (GNNs) are frequently used to predict missing facts in knowledge graphs (KGs). Motivated by the lack of explainability for the outputs of these models, recent work has aimed to explain their predictions using Datalog, a widely used logic-based formalism. However, such work has been restricted to certain subclasses of GNNs. In this paper, we consider one of the most popular GNN architectures for KGs, R-GCN, and we provide two methods to extract rules that explain its predictions and are sound, in the sense that each fact derived by the rules is also predicted by the GNN, for any input dataset. Furthermore, we provide a method that can verify that certain classes of Datalog rules are not sound for the R-GCN. In our experiments, we train R-GCNs on KG completion benchmarks, and we are able to verify that no Datalog rule is sound for these models, even though the models often obtain high to near-perfect accuracy. This raises some concerns about the ability of R-GCN models to generalise and about the explainability of their predictions. We further provide two variations to the training paradigm of R-GCN that encourage it to learn sound rules and find a trade-off between model accuracy and the number of learned sound rules., Comment: Full version (with appendices) of paper accepted to KR 2024 (21st International Conference on Principles of Knowledge Representation and Reasoning)

In the ever-evolving field of digital communication systems, complex-valued neural networks (CVNNs) have become a cornerstone, delivering exceptional performance in tasks like equalization, channel estimation, beamforming, and decoding. Among the myriad of CVNN architectures, the phase-transmittance radial basis function neural network (PT-RBF) stands out, especially when operating in noisy environments such as 5G MIMO systems. Despite its capabilities, achieving convergence in multi-layered, multi-input, and multi-output PT-RBFs remains a daunting challenge. Addressing this gap, this paper presents a novel Deep PT-RBF parameter initialization technique. Through rigorous simulations conforming to 3GPP TS 38 standards, our method not only outperforms conventional initialization strategies like random, $K$-means, and constellation-based methods but is also the only approach to achieve successful convergence in deep PT-RBF architectures. These findings pave the way to more robust and efficient neural network deployments in complex digital communication systems., Comment: IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN 2024)

Multi-agent Reinforcement Learning (MARL) is emerging as a key framework for various sequential decision-making and control tasks. Unlike their single-agent counterparts, multi-agent systems necessitate successful cooperation among the agents. The deployment of these systems in real-world scenarios often requires decentralized training, a diverse set of agents, and learning from infrequent environmental reward signals. These challenges become more pronounced under partial observability and the lack of prior knowledge about agent heterogeneity. While notable studies use intrinsic motivation (IM) to address reward sparsity or cooperation in decentralized settings, those dealing with heterogeneity typically assume centralized training, parameter sharing, and agent indexing. To overcome these limitations, we propose the CoHet algorithm, which utilizes a novel Graph Neural Network (GNN) based intrinsic motivation to facilitate the learning of heterogeneous agent policies in decentralized settings, under the challenges of partial observability and reward sparsity. Evaluation of CoHet in the Multi-agent Particle Environment (MPE) and Vectorized Multi-Agent Simulator (VMAS) benchmarks demonstrates superior performance compared to the state-of-the-art in a range of cooperative multi-agent scenarios. Our research is supplemented by an analysis of the impact of the agent dynamics model on the intrinsic motivation module, insights into the performance of different CoHet variants, and its robustness to an increasing number of heterogeneous agents., Comment: 8 pages, 4 figures

Predictive process monitoring aims to support the execution of a process during runtime with various predictions about the further evolution of a process instance. In the last years a plethora of deep learning architectures have been established as state-of-the-art for different prediction targets, among others the transformer architecture. The transformer architecture is equipped with a powerful attention mechanism, assigning attention scores to each input part that allows to prioritize most relevant information leading to more accurate and contextual output. However, deep learning models largely represent a black box, i.e., their reasoning or decision-making process cannot be understood in detail. This paper examines whether the attention scores of a transformer based next-activity prediction model can serve as an explanation for its decision-making. We find that attention scores in next-activity prediction models can serve as explainers and exploit this fact in two proposed graph-based explanation approaches. The gained insights could inspire future work on the improvement of predictive business process models as well as enabling a neural network based mining of process models from event logs.

In multi-agent cooperative tasks, the presence of heterogeneous agents is familiar. Compared to cooperation among homogeneous agents, collaboration requires considering the best-suited sub-tasks for each agent. However, the operation of multi-agent systems often involves a large amount of complex interaction information, making it more challenging to learn heterogeneous strategies. Related multi-agent reinforcement learning methods sometimes use grouping mechanisms to form smaller cooperative groups or leverage prior domain knowledge to learn strategies for different roles. In contrast, agents should learn deeper role features without relying on additional information. Therefore, we propose QTypeMix, which divides the value decomposition process into homogeneous and heterogeneous stages. QTypeMix learns to extract type features from local historical observations through the TE loss. In addition, we introduce advanced network structures containing attention mechanisms and hypernets to enhance the representation capability and achieve the value decomposition process. The results of testing the proposed method on 14 maps from SMAC and SMACv2 show that QTypeMix achieves state-of-the-art performance in tasks of varying difficulty., Comment: 16 pages, 8 figures

Discrimination mitigation with machine learning (ML) models could be complicated because multiple factors may interweave with each other including hierarchically and historically. Yet few existing fairness measures are able to capture the discrimination level within ML models in the face of multiple sensitive attributes. To bridge this gap, we propose a fairness measure based on distances between sets from a manifold perspective, named as 'harmonic fairness measure via manifolds (HFM)' with two optional versions, which can deal with a fine-grained discrimination evaluation for several sensitive attributes of multiple values. To accelerate the computation of distances of sets, we further propose two approximation algorithms named 'Approximation of distance between sets for one sensitive attribute with multiple values (ApproxDist)' and 'Approximation of extended distance between sets for several sensitive attributes with multiple values (ExtendDist)' to respectively resolve bias evaluation of one single sensitive attribute with multiple values and that of several sensitive attributes with multiple values. Moreover, we provide an algorithmic effectiveness analysis for ApproxDist under certain assumptions to explain how well it could work. The empirical results demonstrate that our proposed fairness measure HFM is valid and approximation algorithms (i.e., ApproxDist and ExtendDist) are effective and efficient., Comment: The first two authors contributed equally, listed in alphabetical order. arXiv admin note: substantial text overlap with arXiv:2405.09251

The Variational Bayesian method (VB) is used to solve the probability distributions of latent variables with the minimum free energy criterion. This criterion is not easy to understand, and the computation is complex. For these reasons, this paper proposes the Semantic Variational Bayes' method (SVB). The Semantic Information Theory the author previously proposed extends the rate-distortion function R(D) to the rate-fidelity function R(G), where R is the minimum mutual information for given semantic mutual information G. SVB came from the parameter solution of R(G), where the variational and iterative methods originated from Shannon et al.'s research on the rate-distortion function. The constraint functions SVB uses include likelihood, truth, membership, similarity, and distortion functions. SVB uses the maximum information efficiency (G/R) criterion, including the maximum semantic information criterion for optimizing model parameters and the minimum mutual information criterion for optimizing the Shannon channel. For the same tasks, SVB is computationally simpler than VB. The computational experiments in the paper include 1) using a mixture model as an example to show that the mixture model converges as G/R increases; 2) demonstrating the application of SVB in data compression with a group of error ranges as the constraint; 3) illustrating how the semantic information measure and SVB can be used for maximum entropy control and reinforcement learning in control tasks with given range constraints, providing numerical evidence for balancing control's purposiveness and efficiency. Further research is needed to apply SVB to neural networks and deep learning., Comment: 21 pages, 7 figures, 39 references

Thin layers and reservoirs may be concealed in areas of low seismic reflection amplitude, making them difficult to recognize. Deep learning (DL) techniques provide new opportunities for accurate impedance prediction by establishing a nonlinear mapping between seismic data and impedance. However, existing methods primarily use time domain seismic data, which limits the capture of frequency bands, thus leading to insufficient resolution of the inversion results. To address these problems, we introduce a new time-frequency-phase (TFP) mixed-domain closed-loop seismic inversion network (TFP-CSIN) to improve the identification of thin layers and reservoirs. First, the inversion network and closed-loop network are constructed by using bidirectional gated recurrent units (Bi-GRU) and convolutional neural network (CNN) architectures, enabling bidirectional mapping between seismic data and impedance data. Next, to comprehensive learning across the entire frequency spectrum, the Fourier transform is used to capture frequency information and establish frequency domain constraints. At the same time, the phase domain constraint is introduced through Hilbert transformation, which improves the method's ability to recognize the weak reflection region features. Both experiments on the synthetic data show that TFP-CSIN outperforms the traditional supervised learning method and time domain semi-supervised learning methods in seismic inversion. The field data further verify that the proposed method improves the identification ability of weak reflection areas and thin layers., Comment: This manuscript has been submitted to Geophysics for reviewing

Explaining decisions of black-box classifiers is both important and computationally challenging. In this paper, we scrutinize explainers that generate feature-based explanations from samples or datasets. We start by presenting a set of desirable properties that explainers would ideally satisfy, delve into their relationships, and highlight incompatibilities of some of them. We identify the entire family of explainers that satisfy two key properties which are compatible with all the others. Its instances provide sufficient reasons, called weak abductive explanations.We then unravel its various subfamilies that satisfy subsets of compatible properties. Indeed, we fully characterize all the explainers that satisfy any subset of compatible properties. In particular, we introduce the first (broad family of) explainers that guarantee the existence of explanations and their global consistency.We discuss some of its instances including the irrefutable explainer and the surrogate explainer whose explanations can be found in polynomial time.

In this paper we numerically analyse problems that arise in numerical integration used in various problems that are nowadays being solved by physics-informed neural networks. In particular we show how inaccurately evaluated integrands can lead to severe precision problems. The tricky and serious part of the problem is especially in its detection, when for example a simple macroscopic visualization of the integrand does not indicate any potential problems and obtained results are often wrongly treated as correct. To solve the problem, it is crucial to decide if it is sufficient to evaluate the integrand in the standard double arithmetic or if a higher precision arithmetic has to be used. We compare and recommend numerical quadratures for selected examples and different parameter values, especially for problematic cases.

The training phase of deep neural networks requires substantial resources and as such is often performed on cloud servers. However, this raises privacy concerns when the training dataset contains sensitive content, e.g., face images. In this work, we propose a method to perform the training phase of a deep learning model on both an edge device and a cloud server that prevents sensitive content being transmitted to the cloud while retaining the desired information. The proposed privacy-preserving method uses adversarial early exits to suppress the sensitive content at the edge and transmits the task-relevant information to the cloud. This approach incorporates noise addition during the training phase to provide a differential privacy guarantee. We extensively test our method on different facial datasets with diverse face attributes using various deep learning architectures, showcasing its outstanding performance. We also demonstrate the effectiveness of privacy preservation through successful defenses against different white-box and deep reconstruction attacks., Comment: 16 pages, 16 figures, 6 tables

Deep Learning (DL) in neuroimaging has become increasingly relevant for detecting neurological conditions and neurodegenerative disorders. One of the most predominant biomarkers in neuroimaging is represented by brain age, which has been shown to be a good indicator for different conditions, such as Alzheimer's Disease. Using brain age for pretraining DL models in transfer learning settings has also recently shown promising results, especially when dealing with data scarcity of different conditions. On the other hand, anatomical information of brain MRIs (e.g. cortical thickness) can provide important information for learning good representations that can be transferred to many downstream tasks. In this work, we propose AnatCL, an anatomical foundation model for brain MRIs that i.) leverages anatomical information with a weakly contrastive learning approach and ii.) achieves state-of-the-art performances in many different downstream tasks. To validate our approach we consider 12 different downstream tasks for diagnosis classification, and prediction of 10 different clinical assessment scores., Comment: 12 pages

Time series forecasting is an important task in many fields ranging from supply chain management to weather forecasting. Recently, Transformer neural network architectures have shown promising results in forecasting on common time series benchmark datasets. However, application to supply chain demand forecasting, which can have challenging characteristics such as sparsity and cross-series effects, has been limited. In this work, we explore the application of Transformer-based models to supply chain demand forecasting. In particular, we develop a new Transformer-based forecasting approach using a shared, multi-task per-time series network with an initial component applying attention across time series, to capture interactions and help address sparsity. We provide a case study applying our approach to successfully improve demand prediction for a medical device manufacturing company. To further validate our approach, we also apply it to public demand forecasting datasets as well and demonstrate competitive to superior performance compared to a variety of baseline and state-of-the-art forecast methods across the private and public datasets.

Catheter ablation has limited therapeutic efficacy against non-paroxysmal atrial fibrillation (AF), and electrophysiological studies using mapping catheters have been applied to evaluate the AF substrate. However, many of these approaches rely on detecting excitation timing from electrograms (ECGs), potentially compromising their effectiveness in complex AF scenarios. Herein, we introduce Deep-learning-based Electrode Action Potential Mapping (DEAP Mapping), a deep learning model designed to reconstruct membrane potential images from annotation-free unipolar ECG signals. We conducted ex vivo experiments using porcine hearts (N = 6) to evaluate the accuracy of DEAP Mapping by simultaneously performing fluorescence measurement of membrane potentials and measurements of epicardial unipolar ECGs. Membrane potentials estimated via DEAP Mapping were compared with those measured via optical mapping. We assessed the clinical applicability of DEAP Mapping by comparing the DEAP Mapping's estimations from clinically measured catheter electrode signals with those from established electrode-mapping techniques. DEAP Mapping accurately estimated conduction delays and blocks in ex vivo experiments. Phase variance analysis, an AF substrate evaluation method, revealed that the substrate identified from optical mapping closely resembled that identified from DEAP Mapping estimations (structural similarity index of >0.8). In clinical evaluations, DEAP Mapping estimation observed several conduction delays and blocks that were not observed with existing methods, indicating that DEAP Mapping can estimate excitation patterns with higher spatiotemporal resolution. DEAP Mapping has a potential to derive detailed changes in membrane potential from intra-operative catheter electrode signals, offering enhanced visualisation of the AF substrate from the estimated membrane potentials., Comment: 17 pages, 7 figures, 6 supplemental movies

Detecting vulnerabilities is a crucial task for maintaining the integrity, availability, and security of software systems. Utilizing DL-based models for vulnerability detection has become commonplace in recent years. However, such deep learning-based vulnerability detectors (DLVD) suffer from a shortage of sizable datasets to train effectively. Data augmentation can potentially alleviate the shortage of data, but augmenting vulnerable code is challenging and requires designing a generative solution that maintains vulnerability. Hence, the work on generating vulnerable code samples has been limited and previous works have only focused on generating samples that contain single statements or specific types of vulnerabilities. Lately, large language models (LLMs) are being used for solving various code generation and comprehension tasks and have shown inspiring results, especially when fused with retrieval augmented generation (RAG). In this study, we explore three different strategies to augment vulnerabilities both single and multi-statement vulnerabilities, with LLMs, namely Mutation, Injection, and Extension. We conducted an extensive evaluation of our proposed approach on three vulnerability datasets and three DLVD models, using two LLMs. Our results show that our injection-based clustering-enhanced RAG method beats the baseline setting (NoAug), Vulgen, and VGX (two SOTA methods), and Random Oversampling (ROS) by 30.80\%, 27.48\%, 27.93\%, and 15.41\% in f1-score with 5K generated vulnerable samples on average, and 53.84\%, 54.10\%, 69.90\%, and 40.93\% with 15K generated vulnerable samples. Our approach demonstrates its feasibility for large-scale data augmentation by generating 1K samples at as cheap as US$ 1.88., Comment: 13 pages, 6 figures, 5 tables, 3 prompt templates, 1 algorithm

We introduce the Robustness of Hierarchically Organized Time Series (RHiOTS) framework, designed to assess the robustness of hierarchical time series forecasting models and algorithms on real-world datasets. Hierarchical time series, where lower-level forecasts must sum to upper-level ones, are prevalent in various contexts, such as retail sales across countries. Current empirical evaluations of forecasting methods are often limited to a small set of benchmark datasets, offering a narrow view of algorithm behavior. RHiOTS addresses this gap by systematically altering existing datasets and modifying the characteristics of individual series and their interrelations. It uses a set of parameterizable transformations to simulate those changes in the data distribution. Additionally, RHiOTS incorporates an innovative visualization component, turning complex, multidimensional robustness evaluation results into intuitive, easily interpretable visuals. This approach allows an in-depth analysis of algorithm and model behavior under diverse conditions. We illustrate the use of RHiOTS by analyzing the predictive performance of several algorithms. Our findings show that traditional statistical methods are more robust than state-of-the-art deep learning algorithms, except when the transformation effect is highly disruptive. Furthermore, we found no significant differences in the robustness of the algorithms when applying specific reconciliation methods, such as MinT. RHiOTS provides researchers with a comprehensive tool for understanding the nuanced behavior of forecasting algorithms, offering a more reliable basis for selecting the most appropriate method for a given problem., Comment: Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '24), August 25--29, 2024, Barcelona, Spain

History eXplanation based on Predicates (HXP), studies the behavior of a Reinforcement Learning (RL) agent in a sequence of agent's interactions with the environment (a history), through the prism of an arbitrary predicate. To this end, an action importance score is computed for each action in the history. The explanation consists in displaying the most important actions to the user. As the calculation of an action's importance is #W[1]-hard, it is necessary for long histories to approximate the scores, at the expense of their quality. We therefore propose a new HXP method, called Backward-HXP, to provide explanations for these histories without having to approximate scores. Experiments show the ability of B-HXP to summarise long histories.

Safe operation of machine learning models requires architectures that explicitly delimit their operational ranges. We evaluate the ability of anomaly detection algorithms to provide indicators correlated with degraded model performance. By placing acceptance thresholds over such indicators, hard boundaries are formed that define the model's coverage. As a use case, we consider the extraction of exoplanetary spectra from transit light curves, specifically within the context of ESA's upcoming Ariel mission. Isolation Forests are shown to effectively identify contexts where prediction models are likely to fail. Coverage/error trade-offs are evaluated under conditions of data and concept drift. The best performance is seen when Isolation Forests model projections of the prediction model's explainability SHAP values., Comment: To appear in "Proceedings of SPAICE 2024: 1st ESA/IAA conference on AI in and for Space". Conference page at https://spaice.esa.int/

Long-term time series forecasting (LTSF) has been widely applied in finance, traffic prediction, and other domains. Recently, patch-based transformers have emerged as a promising approach, segmenting data into sub-level patches that serve as input tokens. However, existing methods mostly rely on predetermined patch lengths, necessitating expert knowledge and posing challenges in capturing diverse characteristics across various scales. Moreover, time series data exhibit diverse variations and fluctuations across different temporal scales, which traditional approaches struggle to model effectively. In this paper, we propose a dynamic tokenizer with a dynamic sparse learning algorithm to capture diverse receptive fields and sparse patterns of time series data. In order to build hierarchical receptive fields, we develop a multi-scale Transformer model, coupled with multi-scale sequence extraction, capable of capturing multi-resolution features. Additionally, we introduce a group-aware rotary position encoding technique to enhance intra- and inter-group position awareness among representations across different temporal scales. Our proposed model, named DRFormer, is evaluated on various real-world datasets, and experimental results demonstrate its superiority compared to existing methods. Our code is available at: https://github.com/ruixindingECNU/DRFormer.

In this paper, we propose an off-policy deep reinforcement learning (DRL) method utilizing the average reward criterion. While most existing DRL methods employ the discounted reward criterion, this can potentially lead to a discrepancy between the training objective and performance metrics in continuing tasks, making the average reward criterion a recommended alternative. We introduce RVI-SAC, an extension of the state-of-the-art off-policy DRL method, Soft Actor-Critic (SAC), to the average reward criterion. Our proposal consists of (1) Critic updates based on RVI Q-learning, (2) Actor updates introduced by the average reward soft policy improvement theorem, and (3) automatic adjustment of Reset Cost enabling the average reward reinforcement learning to be applied to tasks with termination. We apply our method to the Gymnasium's Mujoco tasks, a subset of locomotion tasks, and demonstrate that RVI-SAC shows competitive performance compared to existing methods., Comment: Accepted at ICML 2024; Code: https://github.com/yhisaki/average-reward-drl

This study explores the use of several LLMs for providing quantitative zero-shot sentiment analysis of implicit software desirability expressed by users. The study provides scaled numerical sentiment analysis unlike other methods that simply classify sentiment as positive, neutral, or negative. Numerical analysis provides deeper insights into the magnitude of sentiment, to drive better decisions regarding product desirability. Data is collected through the use of the Microsoft Product Desirability Toolkit (PDT), a well-known qualitative user experience analysis tool. For initial exploration, the PDT metric was given to users of ZORQ, a gamification system used in undergraduate computer science education. The PDT data collected was fed through several LLMs (Claude Sonnet 3 and 3.5, GPT4, and GPT4o) and through a leading transfer learning technique, Twitter-Roberta-Base-Sentiment (TRBS), and through Vader, a leading sentiment analysis tool, for quantitative sentiment analysis. Each system was asked to evaluate the data in two ways, first by looking at the sentiment expressed in the PDT word/explanation pairs; and by looking at the sentiment expressed by the users in their grouped selection of five words and explanations, as a whole. Each LLM was also asked to provide its confidence (low, medium, high) in its sentiment score, along with an explanation of why it selected the sentiment value. All LLMs tested were able to statistically detect user sentiment from the users' grouped data, whereas TRBS and Vader were not. The confidence and explanation of confidence provided by the LLMs assisted in understanding the user sentiment. This study adds to a deeper understanding of evaluating user experiences, toward the goal of creating a universal tool that quantifies implicit sentiment expressed., Comment: 6 pages, 2 figures, 2 tables

Autoencoders have found widespread application, in both their original deterministic form and in their variational formulation (VAEs). In scientific applications it is often of interest to consider data that are comprised of functions; the same perspective is useful in image processing. In practice, discretisation (of differential equations arising in the sciences) or pixellation (of images) renders problems finite dimensional, but conceiving first of algorithms that operate on functions, and only then discretising or pixellating, leads to better algorithms that smoothly operate between different levels of discretisation or pixellation. In this paper function-space versions of the autoencoder (FAE) and variational autoencoder (FVAE) are introduced, analysed, and deployed. Well-definedness of the objective function governing VAEs is a subtle issue, even in finite dimension, and more so on function space. The FVAE objective is well defined whenever the data distribution is compatible with the chosen generative model; this happens, for example, when the data arise from a stochastic differential equation. The FAE objective is valid much more broadly, and can be straightforwardly applied to data governed by differential equations. Pairing these objectives with neural operator architectures, which can thus be evaluated on any mesh, enables new applications of autoencoders to inpainting, superresolution, and generative modelling of scientific data., Comment: 56 pages, 25 figures

In the Western music tradition, chords are the main constituent components of harmony, a fundamental dimension of music. Despite its relevance for several Music Information Retrieval (MIR) tasks, chord-annotated audio datasets are limited and need more diversity. One way to improve those resources is to leverage the large number of chord annotations available online, but this requires aligning them with music audio. However, existing audio-to-score alignment techniques, which typically rely on Dynamic Time Warping (DTW), fail to address this challenge, as they require weakly aligned data for precise synchronisation. In this paper, we introduce ChordSync, a novel conformer-based model designed to seamlessly align chord annotations with audio, eliminating the need for weak alignment. We also provide a pre-trained model and a user-friendly library, enabling users to synchronise chord annotations with audio tracks effortlessly. In this way, ChordSync creates opportunities for harnessing crowd-sourced chord data for MIR, especially in audio chord estimation, thereby facilitating the generation of novel datasets. Additionally, our system extends its utility to music education, enhancing music learning experiences by providing accurately aligned annotations, thus enabling learners to engage in synchronised musical practices., Comment: 8 pages, 3 figures, 3 tables

We introduce tulip agent, an architecture for autonomous LLM-based agents with Create, Read, Update, and Delete access to a tool library containing a potentially large number of tools. In contrast to state-of-the-art implementations, tulip agent does not encode the descriptions of all available tools in the system prompt, which counts against the model's context window, or embed the entire prompt for retrieving suitable tools. Instead, the tulip agent can recursively search for suitable tools in its extensible tool library, implemented exemplarily as a vector store. The tulip agent architecture significantly reduces inference costs, allows using even large tool libraries, and enables the agent to adapt and extend its set of tools. We evaluate the architecture with several ablation studies in a mathematics context and demonstrate its generalizability with an application to robotics. A reference implementation and the benchmark are available at github.com/HRI-EU/tulip_agent., Comment: 19 pages, 4 figures

The segmentation of ultra-high resolution images poses challenges such as loss of spatial information or computational inefficiency. In this work, a novel approach that combines encoder-decoder architectures with domain decomposition strategies to address these challenges is proposed. Specifically, a domain decomposition-based U-Net (DDU-Net) architecture is introduced, which partitions input images into non-overlapping patches that can be processed independently on separate devices. A communication network is added to facilitate inter-patch information exchange to enhance the understanding of spatial context. Experimental validation is performed on a synthetic dataset that is designed to measure the effectiveness of the communication network. Then, the performance is tested on the DeepGlobe land cover classification dataset as a real-world benchmark data set. The results demonstrate that the approach, which includes inter-patch communication for images divided into $16\times16$ non-overlapping subimages, achieves a $2-3\,\%$ higher intersection over union (IoU) score compared to the same network without inter-patch communication. The performance of the network which includes communication is equivalent to that of a baseline U-Net trained on the full image, showing that our model provides an effective solution for segmenting ultra-high-resolution images while preserving spatial context. The code is available at https://github.com/corne00/HiRes-Seg-CNN.

The field of reinforcement learning offers a large variety of concepts and methods to tackle sequential decision-making problems. This variety has become so large that choosing an algorithm for a task at hand can be challenging. In this work, we streamline the process of choosing reinforcement-learning algorithms and action-distribution families. We provide a structured overview of existing methods and their properties, as well as guidelines for when to choose which methods. An interactive version of these guidelines is available online at https://rl-picker.github.io/., Comment: 40 pages

This article describes the methods that achieved 4th and 6th place in the forecasting and investment challenges, respectively, of the M6 competition, ultimately securing the 1st place in the overall duathlon ranking. In the forecasting challenge, we tested a novel meta-learning model that utilizes hypernetworks to design a parametric model tailored to a specific family of forecasting tasks. This approach allowed us to leverage similarities observed across individual forecasting tasks while also acknowledging potential heterogeneity in their data generating processes. The model's training can be directly performed with backpropagation, eliminating the need for reliance on higher-order derivatives and is equivalent to a simultaneous search over the space of parametric functions and their optimal parameter values. The proposed model's capabilities extend beyond M6, demonstrating superiority over state-of-the-art meta-learning methods in the sinusoidal regression task and outperforming conventional parametric models on time-series from the M4 competition. In the investment challenge, we adjusted portfolio weights to induce greater or smaller correlation between our submission and that of other participants, depending on the current ranking, aiming to maximize the probability of achieving a good rank.

As artificial intelligence (AI) systems become increasingly integrated into critical decision-making processes, the need for transparent and interpretable models has become paramount. In this article we present a new ruleset creation method based on surrogate decision trees (SRules), designed to improve the interpretability of black-box machine learning models. SRules balances the accuracy, coverage, and interpretability of machine learning models by recursively creating surrogate interpretable decision tree models that approximate the decision boundaries of a complex model. We propose a systematic framework for generating concise and meaningful rules from these surrogate models, allowing stakeholders to understand and trust the AI system's decision-making process. Our approach not only provides interpretable rules, but also quantifies the confidence and coverage of these rules. The proposed model allows to adjust its parameters to counteract the lack of interpretability by precision and coverage by allowing a near perfect fit and high interpretability of some parts of the model . The results show that SRules improves on other state-of-the-art techniques and introduces the possibility of creating highly interpretable specific rules for specific sub-parts of the model.