Search

Intracranial Hemorrhage is a potentially lethal condition whose manifestation is vastly diverse and shifts across clinical centers worldwide. Deep-learning-based solutions are starting to model complex relations between brain structures, but still struggle to generalize. While gathering more diverse data is the most natural approach, privacy regulations often limit the sharing of medical data. We propose the first application of Federated Scene Graph Generation. We show that our models can leverage the increased training data diversity. For Scene Graph Generation, they can recall up to 20% more clinically relevant relations across datasets compared to models trained on a single centralized dataset. Learning structured data representation in a federated setting can open the way to the development of new methods that can leverage this finer information to regularize across clients more effectively.

Adaptive Mixed-Criticality (AMC) is a fixed-priority preemptive scheduling algorithm for mixed-criticality hard real-time systems. It dominates many other scheduling algorithms for mixed-criticality systems, but does so at the cost of occasionally dropping jobs of less important/critical tasks, when low-priority jobs overrun their time budgets. In this paper we enhance AMC with a deep reinforcement learning (DRL) approach based on a Deep-Q Network. The DRL agent is trained off-line, and at run-time adjusts the low-criticality budgets of tasks to avoid budget overruns, while ensuring that no job misses its deadline if it does not overrun its budget. We have implemented and evaluated this approach by simulating realistic workloads from the automotive domain. The results show that the agent is able to reduce budget overruns by at least up to 50%, even when the budget of each task is chosen based on sampling the distribution of its execution time. To the best of our knowledge, this is the first use of DRL in AMC reported in the literature., Comment: Version submitted to RTNS 2024, on 17/08/2024 (with some typos fixed)

When solving inverse problems, one has to deal with numerous potential sources of model inexactnesses, like object motion, calibration errors, or simplified data models. Regularized Sequential Subspace Optimization (ReSeSOp) allows to compensate for such inaccuracies within the reconstruction step by employing consecutive projections onto suitably defined subspaces. However, this approach relies on a priori estimates for the model inexactness levels which are typically unknown. In dynamic imaging applications, where inaccuracies arise from the unpredictable dynamics of the object, these estimates are particularly challenging to determine in advance. To overcome this limitation, we propose a learned version of ReSeSOp which allows to approximate inexactness levels on the fly. The proposed framework generalizes established unrolled iterative reconstruction schemes to inexact forward operators and is particularly tailored to the structure of dynamic problems. We also present a comprehensive mathematical analysis regarding the effect of dependencies within the forward problem, clarifying when and why dividing the overall problem into subproblems is essential. The proposed method is evaluated on various examples from dynamic imaging, including datasets from a rheological CT experiment, brain MRI, and real-time cardiac MRI. The respective results emphasize improvements in reconstruction quality while ensuring adequate data consistency., Comment: 21 pages, 7 figures, 4 tables

Studying protein mutations within amino acid sequences holds tremendous significance in life sciences. Protein language models (PLMs) have demonstrated strong capabilities in broad biological applications. However, due to architectural design and lack of supervision, PLMs model mutations implicitly with evolutionary plausibility, which is not satisfactory to serve as explainable and engineerable tools in real-world studies. To address these issues, we present MutaPLM, a unified framework for interpreting and navigating protein mutations with protein language models. MutaPLM introduces a protein delta network that captures explicit protein mutation representations within a unified feature space, and a transfer learning pipeline with a chain-of-thought (CoT) strategy to harvest protein mutation knowledge from biomedical texts. We also construct MutaDescribe, the first large-scale protein mutation dataset with rich textual annotations, which provides cross-modal supervision signals. Through comprehensive experiments, we demonstrate that MutaPLM excels at providing human-understandable explanations for mutational effects and prioritizing novel mutations with desirable properties. Our code, model, and data are open-sourced at https://github.com/PharMolix/MutaPLM., Comment: NeurIPS 2024 poster

The auto-regressive architecture, like GPTs, is widely used in modern Text-to-Speech (TTS) systems. However, it incurs substantial inference time, particularly due to the challenges in the next-token prediction posed by lengthy sequences of speech tokens. In this work, we introduce VADUSA, one of the first approaches to accelerate auto-regressive TTS through speculative decoding. Our results show that VADUSA not only significantly improves inference speed but also enhances performance by incorporating draft heads to predict future speech content auto-regressively. Furthermore, the inclusion of a tolerance mechanism during sampling accelerates inference without compromising quality. Our approach demonstrates strong generalization across large datasets and various types of speech tokens., Comment: 5 pages, 3 figures, 3 tables. Submitted to ICASSP 2025

Advanced image segmentation and processing tools present an opportunity to study cell processes and their dynamics. However, image analysis is often routine and time-consuming. Nowadays, alternative data-driven approaches using deep learning are potentially offering automatized, accurate, and fast image analysis. In this paper, we extend the applications of Cellpose, a state-of-the-art cell segmentation framework, with feature extraction capabilities to assess morphological characteristics. We also introduce a dataset of DAPI and FITC stained cells to which our new method is applied.

Traditionally, style has been primarily considered in terms of artistic elements such as colors, brushstrokes, and lighting. However, identical semantic subjects, like people, boats, and houses, can vary significantly across different artistic traditions, indicating that style also encompasses the underlying semantics. Therefore, in this study, we propose a zero-shot scheme for image variation with coordinated semantics. Specifically, our scheme transforms the image-to-image problem into an image-to-text-to-image problem. The image-to-text operation employs vision-language models e.g., BLIP) to generate text describing the content of the input image, including the objects and their positions. Subsequently, the input style keyword is elaborated into a detailed description of this style and then merged with the content text using the reasoning capabilities of ChatGPT. Finally, the text-to-image operation utilizes a Diffusion model to generate images based on the text prompt. To enable the Diffusion model to accommodate more styles, we propose a fine-tuning strategy that injects text and style constraints into cross-attention. This ensures that the output image exhibits similar semantics in the desired style. To validate the performance of the proposed scheme, we constructed a benchmark comprising images of various styles and scenes and introduced two novel metrics. Despite its simplicity, our scheme yields highly plausible results in a zero-shot manner, particularly for generating stylized images with high-fidelity semantics., Comment: 13 pages,6 figures

Precise energy load forecasting in residential households is crucial for mitigating carbon emissions and enhancing energy efficiency; indeed, accurate forecasting enables utility companies and policymakers, who advocate sustainable energy practices, to optimize resource utilization. Moreover, smart meters provide valuable information by allowing for granular insights into consumption patterns. Building upon available smart meter data, our study aims to cluster consumers into distinct groups according to their energy usage behaviours, effectively capturing a diverse spectrum of consumption patterns. Next, we design JITtrans (Just In Time transformer), a novel transformer deep learning model that significantly improves energy consumption forecasting accuracy, with respect to traditional forecasting methods. Extensive experimental results validate our claims using proprietary smart meter data. Our findings highlight the potential of advanced predictive technologies to revolutionize energy management and advance sustainable power systems: the development of efficient and eco-friendly energy solutions critically depends on such technologies.

Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs., Comment: 10 pages + appendix

This paper presents a systematic review of the infrastructure requirements for deploying Large Language Models (LLMs) on-device within the context of small and medium-sized enterprises (SMEs), focusing on both hardware and software perspectives. From the hardware viewpoint, we discuss the utilization of processing units like GPUs and TPUs, efficient memory and storage solutions, and strategies for effective deployment, addressing the challenges of limited computational resources typical in SME settings. From the software perspective, we explore framework compatibility, operating system optimization, and the use of specialized libraries tailored for resource-constrained environments. The review is structured to first identify the unique challenges faced by SMEs in deploying LLMs on-device, followed by an exploration of the opportunities that both hardware innovations and software adaptations offer to overcome these obstacles. Such a structured review provides practical insights, contributing significantly to the community by enhancing the technological resilience of SMEs in integrating LLMs., Comment: 9 pages, 1 figure. The work is supported by the SIT-NVIDIA Joint AI Centre

Simulation plays a crucial role in the rapid development and safe deployment of autonomous vehicles. Realistic traffic agent models are indispensable for bridging the gap between simulation and the real world. Many existing approaches for imitating human behavior are based on learning from demonstration. However, these approaches are often constrained by focusing on individual training strategies. Therefore, to foster a broader understanding of realistic traffic agent modeling, in this paper, we provide an extensive comparative analysis of different training principles, with a focus on closed-loop methods for highway driving simulation. We experimentally compare (i) open-loop vs. closed-loop multi-agent training, (ii) adversarial vs. deterministic supervised training, (iii) the impact of reinforcement losses, and (iv) the impact of training alongside log-replayed agents to identify suitable training techniques for realistic agent modeling. Furthermore, we identify promising combinations of different closed-loop training methods., Comment: 15 pages, 6 figures, 4 tables

This study showcases the digitalization of L\"ofstad Castle in Sweden to contribute to preserving its heritage values. The castle and its collections are deteriorating due to an inappropriate indoor climate. To address this, thirteen cloud-connected sensor boxes, equipped with 84 sensors, were installed throughout the main building, from the basement to the attic, to continuously monitor various indoor environmental parameters. The collected extensive multi-parametric data form the basis for creating a parametric digital twin of the building. The digital twin and detailed data analytics offer a deeper understanding of indoor climate and guide the adoption of appropriate heating and ventilation strategies. The results revealed the need to address high humidity problems in the basement and on the ground floor, such as installing vapor barriers. Opportunities for adopting energy-efficient heating and ventilation strategies on the upper floors were also highlighted. The digitalization solution and findings are not only applicable to L\"ofstad Castle but also provide valuable guidance for the conservation of other historic buildings facing similar challenges., Comment: This work has been submitted to the IEEE for possible publication

Recent LLM (Large Language Models) advancements benefit many fields such as education and finance, but HR has hundreds of repetitive processes, such as access requests, medical claim filing and time-off submissions, which are unaddressed. We relate these tasks to the LLM agent, which has addressed tasks such as writing assisting and customer support. We present HR-Agent, an efficient, confidential, and HR-specific LLM-based task-oriented dialogue system tailored for automating repetitive HR processes such as medical claims and access requests. Since conversation data is not sent to an LLM during inference, it preserves confidentiality required in HR-related tasks.

Knowledge distillation (KD) has been widely used to transfer knowledge from large, accurate models (teachers) to smaller, efficient ones (students). Recent methods have explored enforcing consistency by incorporating causal interpretations to distill invariant representations. In this work, we extend this line of research by introducing a dual augmentation strategy to promote invariant feature learning in both teacher and student models. Our approach leverages different augmentations applied to both models during distillation, pushing the student to capture robust, transferable features. This dual augmentation strategy complements invariant causal distillation by ensuring that the learned representations remain stable across a wider range of data variations and transformations. Extensive experiments on CIFAR-100 demonstrate the effectiveness of this approach, achieving competitive results in same-architecture KD., Comment: 8 pages, 1 figure, 3 tables. This paper presents preliminary results from a project that we have since discontinued, as our research focus has shifted to new directions

A new Kolmogorov-Arnold network (KAN) is proposed to approximate potentially irregular functions in high dimension. We show that it outperforms multilayer perceptrons in terms of accuracy and converges faster. We also compare it with several proposed KAN networks: the original spline-based KAN network appears to be more effective for smooth functions, while the P1-KAN network is more effective for irregular functions.

State of the art reinforcement learning methods sometimes encounter unsafe situations. Identifying when these situations occur is of interest both for post-hoc analysis and during deployment, where it might be advantageous to call out to a human overseer for help. Efforts to gauge the criticality of different points in time have been developed, but their accuracy is not well established due to a lack of ground truth, and they are not designed to be easily interpretable by end users. Therefore, we seek to define a criticality framework with both a quantifiable ground truth and a clear significance to users. We introduce true criticality as the expected drop in reward when an agent deviates from its policy for n consecutive random actions. We also introduce the concept of proxy criticality, a low-overhead metric that has a statistically monotonic relationship to true criticality. Safety margins make these interpretable, when defined as the number of random actions for which performance loss will not exceed some tolerance with high confidence. We demonstrate this approach in several environment-agent combinations; for an A3C agent in an Atari Beamrider environment, the lowest 5% of safety margins contain 47% of agent losses; i.e., supervising only 5% of decisions could potentially prevent roughly half of an agent's errors. This criticality framework measures the potential impacts of bad decisions, even before those decisions are made, allowing for more effective debugging and oversight of autonomous agents., Comment: 17 pages, 10 figures. This work has been submitted to the IEEE for possible publication

We propose a novel method for interpreting neural networks, focusing on convolutional neural network-based receiver model. The method identifies which unit or units of the model contain most (or least) information about the channel parameter(s) of the interest, providing insights at both global and local levels -- with global explanations aggregating local ones. Experiments on link-level simulations demonstrate the method's effectiveness in identifying units that contribute most (and least) to signal-to-noise ratio processing. Although we focus on a radio receiver model, the method generalizes to other neural network architectures and applications, offering robust estimation even in high-dimensional settings., Comment: 7+1 pages, 8 figures

Machine learning is attracting surging interest across nearly all scientific areas by enabling the analysis of large datasets and the extraction of scientific information from incomplete data. Data-driven science is rapidly growing, especially in X-ray methodologies, where advanced light sources and detection technologies accumulate vast amounts of data that exceed meticulous human inspection capabilities. Despite the increasing demands, the full application of machine learning has been hindered by the need for data-specific optimizations. In this study, we introduce a new deep-learning-based phase retrieval method for imperfect diffraction data. This method provides robust phase retrieval for simulated data and performs well on weak-signal single-pulse diffraction data from X-ray free-electron lasers. Moreover, the method significantly reduces data processing time, facilitating real-time image reconstructions that are crucial for high-repetition-rate data acquisition. Thus, this approach offers a reliable solution to the phase problem and is expected to be widely adopted across various research areas.

This study presents an uncertainty-aware stacked neural networks model for the reliable classification of COVID-19 from radiological images. The model addresses the critical gap in uncertainty-aware modeling by focusing on accurately identifying confidently correct predictions while alerting users to confidently incorrect and uncertain predictions, which can promote trust in automated systems. The architecture integrates uncertainty quantification methods, including Monte Carlo dropout and ensemble techniques, to enhance predictive reliability by assessing the certainty of diagnostic predictions. Within a two-tier model framework, the tier one model generates initial predictions and associated uncertainties, which the second tier model uses to produce a trust indicator alongside the diagnostic outcome. This dual-output model not only predicts COVID-19 cases but also provides a trust flag, indicating the reliability of each diagnosis and aiming to minimize the need for retesting and expert verification. The effectiveness of this approach is demonstrated through extensive experiments on the COVIDx CXR-4 dataset, showing a novel approach in identifying and handling confidently incorrect cases and uncertain cases, thus enhancing the trustworthiness of automated diagnostics in clinical settings., Comment: 15 pages, 7 figures, 6 tables

We developed a jitted compiler for training Artificial Neural Networks using C++, LLVM and Cuda. It features object-oriented characteristics, strong typing, parallel workers for data pre-processing, pythonic syntax for expressions, PyTorch like model declaration and Automatic Differentiation. We implement the mechanisms of cache and pooling in order to manage VRAM, cuBLAS for high performance matrix multiplication and cuDNN for convolutional layers. Our experiments with Residual Convolutional Neural Networks on ImageNet, we reach similar speed but degraded performance. Also, the GRU network experiments show similar accuracy, but our compiler have degraded speed in that task. However, our compiler demonstrates promising results at the CIFAR-10 benchmark, in which we reach the same performance and about the same speed as PyTorch. We make the code publicly available at: https://github.com/NoSavedDATA/NoSavedKaleidoscope, Comment: 12 pages, 3 figures and 3 tables

When employing deep neural networks (DNNs) for semantic segmentation in safety-critical applications like automotive perception or medical imaging, it is important to estimate their performance at runtime, e.g. via uncertainty estimates or prediction quality estimates. Previous works mostly performed uncertainty estimation on pixel-level. In a line of research, a connected-component-wise (segment-wise) perspective was taken, approaching uncertainty estimation on an object-level by performing so-called meta classification and regression to estimate uncertainty and prediction quality, respectively. In those works, each predicted segment is considered individually to estimate its uncertainty or prediction quality. However, the neighboring segments may provide additional hints on whether a given predicted segment is of high quality, which we study in the present work. On the basis of uncertainty indicating metrics on segment-level, we use graph neural networks (GNNs) to model the relationship of a given segment's quality as a function of the given segment's metrics as well as those of its neighboring segments. We compare different GNN architectures and achieve a notable performance improvement., Comment: 11 pages, 3 figures, submitted to BMVC "Workshop on Robust Recognition in the Open World" (https://rrow2024.github.io/call-for-papers)

In this effort, we propose using the Mamba for handling tabular data in personalized recommendation systems. We present the \textit{FT-Mamba} (Feature Tokenizer\,$+$\,Mamba), a novel hybrid model that replaces Transformer layers with Mamba layers within the FT-Transformer architecture, for handling tabular data in personalized recommendation systems. The \textit{Mamba model} offers an efficient alternative to Transformers, reducing computational complexity from quadratic to linear by enhancing the capabilities of State Space Models (SSMs). FT-Mamba is designed to improve the scalability and efficiency of recommendation systems while maintaining performance. We evaluate FT-Mamba in comparison to a traditional Transformer-based model within a Two-Tower architecture on three datasets: Spotify music recommendation, H\&M fashion recommendation, and vaccine messaging recommendation. Each model is trained on 160,000 user-action pairs, and performance is measured using precision (P), recall (R), Mean Reciprocal Rank (MRR), and Hit Ratio (HR) at several truncation values. Our results demonstrate that FT-Mamba outperforms the Transformer-based model in terms of computational efficiency while maintaining or exceeding performance across key recommendation metrics. By leveraging Mamba layers, FT-Mamba provides a scalable and effective solution for large-scale personalized recommendation systems, showcasing the potential of the Mamba architecture to enhance both efficiency and accuracy., Comment: 8 pages, 6 figures, 2 tables

We propose TG-LMM (Text-Guided Large Multi-Modal Model), a novel approach that leverages textual descriptions of organs to enhance segmentation accuracy in medical images. Existing medical image segmentation methods face several challenges: current medical automatic segmentation models do not effectively utilize prior knowledge, such as descriptions of organ locations; previous text-visual models focus on identifying the target rather than improving the segmentation accuracy; prior models attempt to use prior knowledge to enhance accuracy but do not incorporate pre-trained models. To address these issues, TG-LMM integrates prior knowledge, specifically expert descriptions of the spatial locations of organs, into the segmentation process. Our model utilizes pre-trained image and text encoders to reduce the number of training parameters and accelerate the training process. Additionally, we designed a comprehensive image-text information fusion structure to ensure thorough integration of the two modalities of data. We evaluated TG-LMM on three authoritative medical image datasets, encompassing the segmentation of various parts of the human body. Our method demonstrated superior performance compared to existing approaches, such as MedSAM, SAM and nnUnet., Comment: 11 pages, 2 figures

In healthcare, medical image segmentation is crucial for accurate disease diagnosis and the development of effective treatment strategies. Early detection can significantly aid in managing diseases and potentially prevent their progression. Machine learning, particularly deep convolutional neural networks, has emerged as a promising approach to addressing segmentation challenges. Traditional methods like U-Net use encoding blocks for local representation modeling and decoding blocks to uncover semantic relationships. However, these models often struggle with multi-scale objects exhibiting significant variations in texture and shape, and they frequently fail to capture long-range dependencies in the input data. Transformers designed for sequence-to-sequence predictions have been proposed as alternatives, utilizing global self-attention mechanisms. Yet, they can sometimes lack precise localization due to insufficient granular details. To overcome these limitations, we introduce TransDAE: a novel approach that reimagines the self-attention mechanism to include both spatial and channel-wise associations across the entire feature space, while maintaining computational efficiency. Additionally, TransDAE enhances the skip connection pathway with an inter-scale interaction module, promoting feature reuse and improving localization accuracy. Remarkably, TransDAE outperforms existing state-of-the-art methods on the Synaps multi-organ dataset, even without relying on pre-trained weights.

Previous research has indicated that deep neural network based models for time series classification (TSC) tasks are prone to overfitting. This issue can be mitigated by employing strategies that prevent the model from becoming overly confident in its predictions, such as label smoothing and confidence penalty. Building upon the concept of label smoothing, we propose a novel approach to generate more reliable soft labels, which we refer to as representation soft label smoothing. We apply label smoothing, confidence penalty, and our method representation soft label smoothing to several TSC models and compare their performance with baseline method which only uses hard labels for training. Our results demonstrate that the use of these enhancement techniques yields competitive results compared to the baseline method. Importantly, our method demonstrates strong performance across models with varying structures and complexities., Comment: 14 pages,6 figures

In the field of medical imaging, the U-Net architecture, along with its variants, has established itself as a cornerstone for image segmentation tasks, particularly due to its strong performance when trained on limited datasets. Despite its impressive performance on identically distributed (in-domain) data, U-Nets exhibit a significant decline in performance when tested on data that deviates from the training distribution, out-of-distribution (out-of-domain) data. Current methodologies predominantly address this issue by employing generalization techniques that hinge on various forms of regularization, which have demonstrated moderate success in specific scenarios. This paper, however, ventures into uncharted territory by investigating the implications of utilizing models that are smaller by three orders of magnitude (i.e., x1000) compared to a conventional U-Net. A reduction of this size in U-net parameters typically adversely affects both in-domain and out-of-domain performance, possibly due to a significantly reduced receptive field. To circumvent this issue, we explore the concept of Neural Cellular Automata (NCA), which, despite its simpler model structure, can attain larger receptive fields through recursive processes. Experimental results on two distinct datasets reveal that NCA outperforms traditional methods in terms of generalization, while still maintaining a commendable IID performance.

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

The loss functions of many learning problems contain multiple additive terms that can disagree and yield conflicting update directions. For Physics-Informed Neural Networks (PINNs), loss terms on initial/boundary conditions and physics equations are particularly interesting as they are well-established as highly difficult tasks. To improve learning the challenging multi-objective task posed by PINNs, we propose the ConFIG method, which provides conflict-free updates by ensuring a positive dot product between the final update and each loss-specific gradient. It also maintains consistent optimization rates for all loss terms and dynamically adjusts gradient magnitudes based on conflict levels. We additionally leverage momentum to accelerate optimizations by alternating the back-propagation of different loss terms. The proposed method is evaluated across a range of challenging PINN scenarios, consistently showing superior performance and runtime compared to baseline methods. We also test the proposed method in a classic multi-task benchmark, where the ConFIG method likewise exhibits a highly promising performance. Source code is available at \url{https://tum-pbs.github.io/ConFIG}., Comment: Project homepage: https://tum-pbs.github.io/ConFIG/

Whole-body CT is used for multi-trauma patients in the search of any and all injuries. Since an initial assessment needs to be rapid and the search for lesions is done for the whole body, very little time can be allocated for the inspection of a specific anatomy. In particular, intracranial hemorrhages are still missed, especially by clinical students. In this work, we present a Deep Learning approach for highlighting such lesions to improve the diagnostic accuracy. While most works on intracranial hemorrhages perform segmentation, detection only requires bounding boxes for the localization of the bleeding. In this paper, we propose a novel Voxel-Complete IoU (VC-IoU) loss that encourages the network to learn the 3D aspect ratios of bounding boxes and leads to more precise detections. We extensively experiment on brain bleeding detection using a publicly available dataset, and validate it on a private cohort, where we achieve 0.877 AR30, 0.728 AP30, and 0.653 AR30, 0.514 AP30 respectively. These results constitute a relative +5% improvement in Average Recall for both datasets compared to other loss functions. Finally, as there is little data currently publicly available for 3D object detection and as annotation resources are limited in the clinical setting, we evaluate the cost of different annotation methods, as well as the impact of imprecise bounding boxes in the training data on the detection performance.

This paper compared physics-informed neural network (PINN), conventional neural network (NN) and traditional numerical discretization methods on solving differential equations (DEs) through literature investigation and experimental validation. We focused on the soft-constrained PINN approach and formalized its mathematical framework and computational flow for solving Ordinary DEs and Partial DEs (ODEs/PDEs). The working mechanism and its accuracy and efficiency were experimentally verified by solving typical linear and non-linear oscillator ODEs. We demonstrate that the DeepXDE-based implementation of PINN is not only light code and efficient in training, but also flexible across CPU/GPU platforms. PINN greatly reduces the need for labeled data: when the nonlinearity of the ODE is weak, a very small amount of supervised training data plus a few unsupervised collocation points are sufficient to predict the solution; in the minimalist case, only one or two training points (with initial values) are needed for first- or second-order ODEs, respectively. We also find that, with the aid of collocation points and the use of physical information, PINN has the ability to extrapolate data outside the time domain of the training set, and especially is robust to noisy data, thus with enhanced generalization capabilities. Training is accelerated when the gains obtained along with the reduction in the amount of data outweigh the delay caused by the increase in the loss function terms. The soft-constrained PINN can easily impose a physical law (e.g., conservation of energy) constraint by adding a regularization term to the total loss function, thus improving the solution performance to ODEs that obey this physical law. Furthermore, PINN can also be used for stiff ODEs, PDEs, and other types of DEs, and is becoming a favorable catalyst for the era of Digital Twins., Comment: 24 pages, 7 figures, 2 tables, etc. Ready for submission

In this paper we construct and theoretically analyse group equivariant convolutional kernel networks (CKNs) which are useful in understanding the geometry of (equivariant) CNNs through the lens of reproducing kernel Hilbert spaces (RKHSs). We then proceed to study the stability analysis of such equiv-CKNs under the action of diffeomorphism and draw a connection with equiv-CNNs, where the goal is to analyse the geometry of inductive biases of equiv-CNNs through the lens of reproducing kernel Hilbert spaces (RKHSs). Traditional deep learning architectures, including CNNs, trained with sophisticated optimization algorithms is vulnerable to perturbations, including `adversarial examples'. Understanding the RKHS norm of such models through CKNs is useful in designing the appropriate architecture and can be useful in designing robust equivariant representation learning models.

In a typical e-commerce setting, Content Ranking Optimization (CRO) mechanisms are employed to surface content on the search page to fulfill customers' shopping missions. CRO commonly utilizes models such as contextual deep bandits model to independently rank content at different positions, e.g., one optimizer dedicated to organic search results and another to sponsored results. However, this regional optimization approach does not necessarily translate to whole page optimization, e.g., maximizing revenue at the top of the page may inadvertently diminish the revenue of lower positions. In this paper, we propose a reinforcement learning based method for whole page ranking to jointly optimize across all positions by: 1) shifting from position level optimization to whole page level optimization to achieve an overall optimized ranking; 2) applying reinforcement learning to optimize for the cumulative rewards instead of the instant reward. We formulate page level CRO as a cooperative Multi-agent Markov Decision Process , and address it with the novel Multi-Agent Deep Deterministic Policy Gradient (MADDPG) model. MADDPG supports a flexible and scalable joint optimization framework by adopting a "centralized training and decentralized execution" approach. Extensive experiments demonstrate that MADDPG scales to a 2.5 billion action space in the public Mujoco environment, and outperforms the deep bandits modeling by 25.7% on the offline CRO data set from a leading e-commerce company. We foresee that this novel multi-agent optimization is applicable to similar joint optimization problems in the field of information retrieval., Comment: 14 pages

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

Large-scale self-supervised pre-training has paved the way for one foundation model to handle many different vision tasks. Most pre-training methodologies train a single model of a certain size at one time. Nevertheless, various computation or storage constraints in real-world scenarios require substantial efforts to develop a series of models with different sizes to deploy. Thus, in this study, we propose a novel tri-branch self-supervised training framework, termed as POA (Pre-training Once for All), to tackle this aforementioned issue. Our approach introduces an innovative elastic student branch into a modern self-distillation paradigm. At each pre-training step, we randomly sample a sub-network from the original student to form the elastic student and train all branches in a self-distilling fashion. Once pre-trained, POA allows the extraction of pre-trained models of diverse sizes for downstream tasks. Remarkably, the elastic student facilitates the simultaneous pre-training of multiple models with different sizes, which also acts as an additional ensemble of models of various sizes to enhance representation learning. Extensive experiments, including k-nearest neighbors, linear probing evaluation and assessments on multiple downstream tasks demonstrate the effectiveness and advantages of our POA. It achieves state-of-the-art performance using ViT, Swin Transformer and ResNet backbones, producing around a hundred models with different sizes through a single pre-training session. The code is available at: https://github.com/Qichuzyy/POA., Comment: Accepted by ECCV2024

Patients with Intracranial Hemorrhage (ICH) face a potentially life-threatening condition, and patient-centered individualized treatment remains challenging due to possible clinical complications. Deep-Learning-based methods can efficiently analyze the routinely acquired head CTs to support the clinical decision-making. The majority of early work focuses on the detection and segmentation of ICH, but do not model the complex relations between ICH and adjacent brain structures. In this work, we design a tailored object detection method for ICH, which we unite with segmentation-grounded Scene Graph Generation (SGG) methods to learn a holistic representation of the clinical cerebral scene. To the best of our knowledge, this is the first application of SGG for 3D voxel images. We evaluate our method on two head-CT datasets and demonstrate that our model can recall up to 74% of clinically relevant relations. This work lays the foundation towards SGG for 3D voxel data. The generated Scene Graphs can already provide insights for the clinician, but are also valuable for all downstream tasks as a compact and interpretable representation.

Vision Transformer (ViT) acceleration with field programmable gate array (FPGA) is promising but challenging. Existing FPGA-based ViT accelerators mainly rely on temporal architectures, which process different operators by reusing the same hardware blocks and suffer from extensive memory access overhead. Pipelined architectures, either coarse-grained or fine-grained, unroll the ViT computation spatially for memory access efficiency. However, they usually suffer from significant hardware resource constraints and pipeline bubbles induced by the global computation dependency of ViT. In this paper, we introduce HG-PIPE, a pipelined FPGA accelerator for high-throughput and low-latency ViT processing. HG-PIPE features a hybrid-grained pipeline architecture to reduce on-chip buffer cost and couples the computation dataflow and parallelism design to eliminate the pipeline bubbles. HG-PIPE further introduces careful approximations to implement both linear and non-linear operators with abundant Lookup Tables (LUTs), thus alleviating resource constraints. On a ZCU102 FPGA, HG-PIPE achieves 2.78 times better throughput and 2.52 times better resource efficiency than the prior-art accelerators, e.g., AutoViTAcc. With a VCK190 FPGA, HG-PIPE realizes end-to-end ViT acceleration on a single device and achieves 7118 images/s, which is 2.81 times faster than a V100 GPU., Comment: Accepted by ICCAD 2024

Low-precision formats such as float8 have been introduced in machine learning accelerated hardware to improve computational efficiency for large language models training and inference. Nevertheless, adoption by the ML community has been slowed down by the complex, and sometimes brittle, techniques required to match higher precision training accuracy. In this work, we present Scalify, a end-to-end scale propagation paradigm for computational graphs, generalizing and formalizing existing tensor scaling methods. Experiment results show that Scalify supports out-of-the-box float8 matrix multiplication and gradients representation, as well as float16 optimizer state storage. Our JAX implementation of Scalify is open-sourced at https://github.com/graphcore-research/jax-scalify, Comment: 11 pages, 5 figures, ICML 2024 WANT workshop

It has been observed that even a small amount of noise introduced into the dataset can significantly degrade the performance of KAN. In this brief note, we aim to quantitatively evaluate the performance when noise is added to the dataset. We propose an oversampling technique combined with denoising to alleviate the impact of noise. Specifically, we employ kernel filtering based on diffusion maps for pre-filtering the noisy data for training KAN network. Our experiments show that while adding i.i.d. noise with any fixed SNR, when we increase the amount of training data by a factor of $r$, the test-loss (RMSE) of KANs will exhibit a performance trend like $\text{test-loss} \sim \mathcal{O}(r^{-\frac{1}{2}})$ as $r\to +\infty$. We conclude that applying both oversampling and filtering strategies can reduce the detrimental effects of noise. Nevertheless, determining the optimal variance for the kernel filtering process is challenging, and enhancing the volume of training data substantially increases the associated costs, because the training dataset needs to be expanded multiple times in comparison to the initial clean data. As a result, the noise present in the data ultimately diminishes the effectiveness of Kolmogorov-Arnold networks.

Knowledge distillation (KD) is an effective method for transferring knowledge from a large, well-trained teacher model to a smaller, more efficient student model. Despite its success, one of the main challenges in KD is ensuring the efficient transfer of complex knowledge while maintaining the student's computational efficiency. Unlike previous works that applied contrastive objectives promoting explicit negative instances with little attention to the relationships between them, we introduce Relational Representation Distillation (RRD). Our approach leverages pairwise similarities to explore and reinforce the relationships between the teacher and student models. Inspired by self-supervised learning principles, it uses a relaxed contrastive loss that focuses on similarity rather than exact replication. This method aligns the output distributions of teacher samples in a large memory buffer, improving the robustness and performance of the student model without the need for strict negative instance differentiation. Our approach demonstrates superior performance on CIFAR-100 and ImageNet ILSVRC-2012, outperforming traditional KD and sometimes even outperforms the teacher network when combined with KD. It also transfers successfully to other datasets like Tiny ImageNet and STL-10. Code is available at https://github.com/giakoumoglou/distillers., Comment: 8 pages, 4 figures, 4 tables. Updated experiments on ImageNet ILSVRC-2012. arXiv admin note: text overlap with arXiv:2407.11802

Knowledge distillation (KD) involves transferring the knowledge from one neural network to another, often from a larger, well-trained model (teacher) to a smaller, more efficient model (student). Traditional KD methods minimize the Kullback-Leibler (KL) divergence between the probabilistic outputs of the teacher and student networks. However, this approach often overlooks crucial structural knowledge embedded within the teacher's network. In this paper, we introduce Invariant Consistency Distillation (ICD), a novel methodology designed to enhance KD by ensuring that the student model's representations are both discriminative and invariant with respect to the teacher's outputs. Our approach is based on causal inference principles and combines contrastive learning with an explicit invariance penalty, capturing significantly more information from the teacher's representation. ICD uses an efficient, parameter-free approach for flexible teacher-student alignment. We provide a theoretical foundation for ICD and demonstrate its effectiveness through extensive experiments. Our results on CIFAR-100 and ImageNet ILSVRC-2012 show that ICD outperforms traditional KD techniques and surpasses state-of-the-art methods. In some cases, the student model even exceeds the teacher model in terms of accuracy. Furthermore, we successfully apply our method to other datasets, such as Tiny ImageNet and STL-10, demonstrating superior cross-dataset generalization. Code is available at https://github.com/giakoumoglou/distillers., Comment: 8 pages, 2 figures, 4 tables. The paper's title has been changed to better emphasize its theoretical foundation

Fine-tuning advanced diffusion models for high-quality image stylization usually requires large training datasets and substantial computational resources, hindering their practical applicability. We propose Ada-Adapter, a novel framework for few-shot style personalization of diffusion models. Ada-Adapter leverages off-the-shelf diffusion models and pre-trained image feature encoders to learn a compact style representation from a limited set of source images. Our method enables efficient zero-shot style transfer utilizing a single reference image. Furthermore, with a small number of source images (three to five are sufficient) and a few minutes of fine-tuning, our method can capture intricate style details and conceptual characteristics, generating high-fidelity stylized images that align well with the provided text prompts. We demonstrate the effectiveness of our approach on various artistic styles, including flat art, 3D rendering, and logo design. Our experimental results show that Ada-Adapter outperforms existing zero-shot and few-shot stylization methods in terms of output quality, diversity, and training efficiency., Comment: 16 pages, 11 figures

AI-enabled capabilities are reaching the requisite level of maturity to be deployed in the real world, yet do not always make correct or safe decisions. One way of addressing these concerns is to leverage AI control systems alongside and in support of human decisions, relying on the AI control system in safe situations while calling on a human co-decider for critical situations. We extend a methodology for adversarial explanations (AE) to state-of-the-art reinforcement learning frameworks, including MuZero. Multiple improvements to the base agent architecture are proposed. We demonstrate how this technology has two applications: for intelligent decision tools and to enhance training / learning frameworks. In a decision support context, adversarial explanations help a user make the correct decision by highlighting those contextual factors that would need to change for a different AI-recommended decision. As another benefit of adversarial explanations, we show that the learned AI control system demonstrates robustness against adversarial tampering. Additionally, we supplement AE by introducing strategically similar autoencoders (SSAs) to help users identify and understand all salient factors being considered by the AI system. In a training / learning framework, this technology can improve both the AI's decisions and explanations through human interaction. Finally, to identify when AI decisions would most benefit from human oversight, we tie this combined system to our prior art on statistically verified analyses of the criticality of decisions at any point in time., Comment: 19 pages, 12 figures

With the rapid advances in deep learning and smart manufacturing in Industry 4.0, there is an imperative for high-throughput, high-performance, and fully integrated visual inspection systems. Most anomaly detection approaches using defect detection datasets, such as MVTec AD, employ one-class models that require fitting separate models for each class. On the contrary, unified models eliminate the need for fitting separate models for each class and significantly reduce cost and memory requirements. Thus, in this work, we experiment with considering a unified multi-class setup. Our experimental study shows that multi-class models perform at par with one-class models for the standard MVTec AD dataset. Hence, this indicates that there may not be a need to learn separate object/class-wise models when the object classes are significantly different from each other, as is the case of the dataset considered. Furthermore, we have deployed three different unified lightweight architectures on the CPU and an edge device (NVIDIA Jetson Xavier NX). We analyze the quantized multi-class anomaly detection models in terms of latency and memory requirements for deployment on the edge device while comparing quantization-aware training (QAT) and post-training quantization (PTQ) for performance at different precision widths. In addition, we explored two different methods of calibration required in post-training scenarios and show that one of them performs notably better, highlighting its importance for unsupervised tasks. Due to quantization, the performance drop in PTQ is further compensated by QAT, which yields at par performance with the original 32-bit Floating point in two of the models considered., Comment: 20 pages

We introduce the concept of deceptive diffusion -- training a generative AI model to produce adversarial images. Whereas a traditional adversarial attack algorithm aims to perturb an existing image to induce a misclassificaton, the deceptive diffusion model can create an arbitrary number of new, misclassified images that are not directly associated with training or test images. Deceptive diffusion offers the possibility of strengthening defence algorithms by providing adversarial training data at scale, including types of misclassification that are otherwise difficult to find. In our experiments, we also investigate the effect of training on a partially attacked data set. This highlights a new type of vulnerability for generative diffusion models: if an attacker is able to stealthily poison a portion of the training data, then the resulting diffusion model will generate a similar proportion of misleading outputs.

The most recent advances in medical imaging that have transformed diagnosis, especially in the case of interpreting X-ray images, are actively involved in the healthcare sector. The advent of digital image processing technology and the implementation of deep learning models such as Convolutional Neural Networks (CNNs) have made the analysis of X-rays much more accurate and efficient. In this article, some essential techniques such as edge detection, region-growing technique, and thresholding approach, and the deep learning models such as variants of YOLOv8-which is the best object detection and segmentation framework-are reviewed. We further investigate that the traditional image processing techniques like segmentation are very much simple and provides the alternative to the advanced methods as well. Our review gives useful knowledge on the practical usage of the innovative and traditional approaches of manual X-ray interpretation. The discovered information will help professionals and researchers to gain more profound knowledge in digital interpretation techniques in medical imaging., Comment: 29 pages, 20 figures

Glaucoma is a chronic eye disease characterized by optic neuropathy, leading to irreversible vision loss. It progresses gradually, often remaining undiagnosed until advanced stages. Early detection is crucial to monitor atrophy and develop treatment strategies to prevent further vision impairment. Data-centric methods have enabled computer-aided algorithms for precise glaucoma diagnosis. In this study, we use deep learning models to identify complex disease traits and progression criteria, detecting subtle changes indicative of glaucoma. We explore the structure-function relationship in glaucoma progression and predict functional impairment from structural eye deterioration. We analyze statistical and machine learning methods, including deep learning techniques with optical coherence tomography (OCT) scans for accurate progression prediction. Addressing challenges like age variability, data imbalances, and noisy labels, we develop novel semi-supervised time-series algorithms: 1. Weakly-Supervised Time-Series Learning: We create a CNN-LSTM model to encode spatiotemporal features from OCT scans. This approach uses age-related progression and positive-unlabeled data to establish robust pseudo-progression criteria, bypassing gold-standard labels. 2. Semi-Supervised Time-Series Learning: Using labels from Guided Progression Analysis (GPA) in a contrastive learning scheme, the CNN-LSTM architecture learns from potentially mislabeled data to improve prediction accuracy. Our methods outperform conventional and state-of-the-art techniques., Comment: Dissertation

Converting input symbols to output audio in TTS requires modelling the durations of speech sounds. Leading non-autoregressive (NAR) TTS models treat duration modelling as a regression problem. The same utterance is then spoken with identical timings every time, unlike when a human speaks. Probabilistic models of duration have been proposed, but there is mixed evidence of their benefits. However, prior studies generally only consider speech read aloud, and ignore spontaneous speech, despite the latter being both a more common and a more variable mode of speaking. We compare the effect of conventional deterministic duration modelling to durations sampled from a powerful probabilistic model based on conditional flow matching (OT-CFM), in three different NAR TTS approaches: regression-based, deep generative, and end-to-end. Across four different corpora, stochastic duration modelling improves probabilistic NAR TTS approaches, especially for spontaneous speech. Please see https://shivammehta25.github.io/prob_dur/ for audio and resources., Comment: 5 pages, 2 figures. Final version, accepted to Interspeech 2024

The accurate prediction of stock movements is crucial for investment strategies. Stock prices are subject to the influence of various forms of information, including financial indicators, sentiment analysis, news documents, and relational structures. Predominant analytical approaches, however, tend to address only unimodal or bimodal sources, neglecting the complexity of multimodal data. Further complicating the landscape are the issues of data sparsity and semantic conflicts between these modalities, which are frequently overlooked by current models, leading to unstable performance and limiting practical applicability. To address these shortcomings, this study introduces a novel architecture, named Multimodal Stable Fusion with Gated Cross-Attention (MSGCA), designed to robustly integrate multimodal input for stock movement prediction. The MSGCA framework consists of three integral components: (1) a trimodal encoding module, responsible for processing indicator sequences, dynamic documents, and a relational graph, and standardizing their feature representations; (2) a cross-feature fusion module, where primary and consistent features guide the multimodal fusion of the three modalities via a pair of gated cross-attention networks; and (3) a prediction module, which refines the fused features through temporal and dimensional reduction to execute precise movement forecasting. Empirical evaluations demonstrate that the MSGCA framework exceeds current leading methods, achieving performance gains of 8.1%, 6.1%, 21.7% and 31.6% on four multimodal datasets, respectively, attributed to its enhanced multimodal fusion stability., Comment: 29 pages, 10 figures

Contrastive language-audio pre-training (CLAP) enables zero-shot (ZS) inference of audio and exhibits promising performance in several classification tasks. However, conventional audio representations are still crucial for many tasks where ZS is not applicable (e.g., regression problems). Here, we explore a new representation, a general-purpose audio-language representation, that performs well in both ZS and transfer learning. To do so, we propose a new method, M2D-CLAP, which combines self-supervised learning Masked Modeling Duo (M2D) and CLAP. M2D learns an effective representation to model audio signals, and CLAP aligns the representation with text embedding. As a result, M2D-CLAP learns a versatile representation that allows for both ZS and transfer learning. Experiments show that M2D-CLAP performs well on linear evaluation, fine-tuning, and ZS classification with a GTZAN state-of-the-art of 75.17%, thus achieving a general-purpose audio-language representation., Comment: 5 pages, 1 figure, 5 tables. Accepted by Interspeech 2024

In the realm of advanced steganography, the scale of the model typically correlates directly with the resolution of the fundamental grid, necessitating the training of a distinct neural network for message extraction. This paper proposes an image steganography based on generative implicit neural representation. This approach transcends the constraints of image resolution by portraying data as continuous functional expressions. Notably, this method permits the utilization of a diverse array of multimedia data as cover images, thereby broadening the spectrum of potential carriers. Additionally, by fixing a neural network as the message extractor, we effectively redirect the training burden to the image itself, resulting in both a reduction in computational overhead and an enhancement in steganographic speed. This approach also circumvents potential transmission challenges associated with the message extractor. Experimental findings reveal that this methodology achieves a commendable optimization efficiency, achieving a completion time of just 3 seconds for 64x64 dimensional images, while concealing only 1 bpp of information. Furthermore, the accuracy of message extraction attains an impressive mark of 100%., Comment: 33 pages, 15 figures and 5 tables