Guillermo Sapiro

Antoine Wehenkel, Laura Manduchi, Jens Behrmann et al. · apple-ml

Over the past decades, hemodynamics simulators have steadily evolved and have become tools of choice for studying cardiovascular systems in-silico. While such tools are routinely used to simulate whole-body hemodynamics from physiological parameters, solving the corresponding inverse problem of mapping waveforms back to plausible physiological parameters remains both promising and challenging. Motivated by advances in simulation-based inference (SBI), we cast this inverse problem as statistical inference. In contrast to alternative approaches, SBI provides \textit{posterior distributions} for the parameters of interest, providing a \textit{multi-dimensional} representation of uncertainty for \textit{individual} measurements. We showcase this ability by performing an in-silico uncertainty analysis of five biomarkers of clinical interest comparing several measurement modalities. Beyond the corroboration of known facts, such as the feasibility of estimating heart rate, our study highlights the potential of estimating new biomarkers from standard-of-care measurements. SBI reveals practically relevant findings that cannot be captured by standard sensitivity analyses, such as the existence of sub-populations for which parameter estimation exhibits distinct uncertainty regimes. Finally, we study the gap between in-vivo and in-silico with the MIMIC-III waveform database and critically discuss how cardiovascular simulations can inform real-world data analysis.

Mariano Tepper, Guillermo Sapiro

Splines are a popular and attractive way of smoothing noisy data. Computing splines involves minimizing a functional which is a linear combination of a fitting term and a regularization term. The former is classically computed using a (weighted) L2 norm while the latter ensures smoothness. Thus, when dealing with grid data, the optimization can be solved very efficiently using the DCT. In this work we propose to replace the L2 norm in the fitting term with an L1 norm, leading to automatic robustness to outliers. To solve the resulting minimization problem we propose an extremely simple and efficient numerical scheme based on split-Bregman iteration combined with DCT. Experimental validation shows the high-quality results obtained in short processing times.

Ziye Xie, Oded Schlesinger, Raj Kundu et al.

Neuronavigation is widely used in biomedical research and interventions to guide the precise placement of instruments around the head to support procedures such as transcranial magnetic stimulation. Traditional systems, however, rely on subject-mounted markers that require manual registration, may shift during procedures, and can cause discomfort. We introduce and evaluate markerless approaches that replace expensive hardware and physical markers with low-cost visible and infrared light cameras incorporating stereo and depth sensing, combined with algorithmic modeling of the facial geometry. Validation with 50 human subjects yielded a median tracking discrepancy of only 2.32 mm and 2.01$^\circ$ for the best markerless algorithm compared to a conventional marker-based system, which indicates sufficient accuracy for transcranial magnetic stimulation and a substantial improvement over prior markerless results. The study also suggests that integration of the data from the various camera sensors can improve the overall accuracy further. The proposed markerless neuronavigation methods can reduce setup cost and complexity, improve patient comfort, and expand access to neuronavigation in clinical and research settings.

Oded Schlesinger, Amirhossein Farzam, J. Matias Di Martino et al.

While Vision Transformers (ViT) have demonstrated remarkable performance across diverse tasks, their computational demands are substantial, scaling quadratically with the number of processed tokens. Compact attention representations, reflecting token interaction distributions, can guide early detection and reduction of less salient tokens prior to attention computation. Motivated by this, we present SParsification with attentiOn dynamics via Token relevance (SPOT), a framework for early detection of redundant tokens within ViTs that leverages token embeddings, interactions, and attention dynamics across layers to infer token importance, resulting in a more context-aware and interpretable relevance detection process. SPOT informs token sparsification and facilitates the elimination of such tokens, improving computational efficiency without sacrificing performance. SPOT employs computationally lightweight predictors that can be plugged into various ViT architectures and learn to derive effective input-specific token prioritization across layers. Its versatile design supports a range of performance levels adaptable to varying resource constraints. Empirical evaluations demonstrate significant efficiency gains of up to 40% compared to standard ViTs, while maintaining or even improving accuracy. Code and models are available at https://github.com/odedsc/SPOT .

Jing Xiong, Gongye Liu, Lun Huang et al.

Autoregressive modeling has been a huge success in the field of natural language processing (NLP). Recently, autoregressive models have emerged as a significant area of focus in computer vision, where they excel in producing high-quality visual content. Autoregressive models in NLP typically operate on subword tokens. However, the representation strategy in computer vision can vary in different levels, i.e., pixel-level, token-level, or scale-level, reflecting the diverse and hierarchical nature of visual data compared to the sequential structure of language. This survey comprehensively examines the literature on autoregressive models applied to vision. To improve readability for researchers from diverse research backgrounds, we start with preliminary sequence representation and modeling in vision. Next, we divide the fundamental frameworks of visual autoregressive models into three general sub-categories, including pixel-based, token-based, and scale-based models based on the representation strategy. We then explore the interconnections between autoregressive models and other generative models. Furthermore, we present a multifaceted categorization of autoregressive models in computer vision, including image generation, video generation, 3D generation, and multimodal generation. We also elaborate on their applications in diverse domains, including emerging domains such as embodied AI and 3D medical AI, with about 250 related references. Finally, we highlight the current challenges to autoregressive models in vision with suggestions about potential research directions. We have also set up a Github repository to organize the papers included in this survey at: https://github.com/ChaofanTao/Autoregressive-Models-in-Vision-Survey.

Qiangqiang Wu, Tianyu Yang, Bo Fang et al.

Tracking Any Point (TAP) has emerged as a fundamental tool for video understanding. Current approaches adapt Vision Foundation Models (VFMs) like DINOv2 via offline finetuning or test-time optimization. However, these VFMs rely on static image pre-training, which is inherently sub-optimal for capturing dense temporal correspondence in videos. To address this, we propose Mask-to-Point (M2P) learning, which leverages rich video object segmentation (VOS) mask annotations to improve VFMs for dense point tracking. Our M2P introduces three new mask-based constraints for weakly-supervised representation learning. First, we propose a local structure consistency loss, which leverages Procrustes analysis to model the cohesive motion of points lying within a local structure, achieving more reliable point-to-point matching learning. Second, we propose a mask label consistency (MLC) loss, which enforces that sampled foreground points strictly match foreground regions across frames. The proposed MLC loss can be regarded as a regularization, which stabilizes training and prevents convergence to trivial solutions. Finally, mask boundary constrain is applied to explicitly supervise boundary points. We show that our weaklysupervised M2P models significantly outperform baseline VFMs with efficient training by using only 3.6K VOS training videos. Notably, M2P achieves 12.8% and 14.6% performance gains over DINOv2-B/14 and DINOv3-B/16 on the TAP-Vid-DAVIS benchmark, respectively. Moreover, the proposed M2P models are used as pre-trained backbones for both test-time optimized and offline fine-tuned TAP tasks, demonstrating its potential to serve as general pre-trained models for point tracking. Code will be made publicly available upon acceptance.

Amirhossein Farzam, Majid Behabahani, Mani Malek et al.

Large language models (LLMs) remain vulnerable to jailbreak prompts that are fluent and semantically coherent, and therefore difficult to detect with standard heuristics. A particularly challenging failure mode occurs when an attacker tries to hide the malicious goal of their request by manipulating its framing to induce compliance. Because these attacks maintain malicious intent through a flexible presentation, defenses that rely on structural artifacts or goal-specific signatures can fail. Motivated by this, we introduce a self-supervised framework for disentangling semantic factor pairs in LLM activations at inference. We instantiate the framework for goal and framing and construct GoalFrameBench, a corpus of prompts with controlled goal and framing variations, which we use to train Representation Disentanglement on Activations (ReDAct) module to extract disentangled representations in a frozen LLM. We then propose FrameShield, an anomaly detector operating on the framing representations, which improves model-agnostic detection across multiple LLM families with minimal computational overhead. Theoretical guarantees for ReDAct and extensive empirical validations show that its disentanglement effectively powers FrameShield. Finally, we use disentanglement as an interpretability probe, revealing distinct profiles for goal and framing signals and positioning semantic disentanglement as a building block for both LLM safety and mechanistic interpretability.

Martin Bertran, Natalia Martinez, Alex Oesterling et al.

Machine learning models are updated as new data is acquired or new architectures are developed. These updates usually increase model performance, but may introduce backward compatibility errors, where individual users or groups of users see their performance on the updated model adversely affected. This problem can also be present when training datasets do not accurately reflect overall population demographics, with some groups having overall lower participation in the data collection process, posing a significant fairness concern. We analyze how ideas from distributional robustness and minimax fairness can aid backward compatibility in this scenario, and propose two methods to directly address this issue. Our theoretical analysis is backed by experimental results on CIFAR-10, CelebA, and Waterbirds, three standard image classification datasets. Code available at github.com/natalialmg/GroupBC

Gilad Cohen, Guillermo Sapiro, Raja Giryes

Deep neural networks (DNNs) are notorious for their vulnerability to adversarial attacks, which are small perturbations added to their input images to mislead their prediction. Detection of adversarial examples is, therefore, a fundamental requirement for robust classification frameworks. In this work, we present a method for detecting such adversarial attacks, which is suitable for any pre-trained neural network classifier. We use influence functions to measure the impact of every training sample on the validation set data. From the influence scores, we find the most supportive training samples for any given validation example. A k-nearest neighbor (k-NN) model fitted on the DNN's activation layers is employed to search for the ranking of these supporting training samples. We observe that these samples are highly correlated with the nearest neighbors of the normal inputs, while this correlation is much weaker for adversarial inputs. We train an adversarial detector using the k-NN ranks and distances and show that it successfully distinguishes adversarial examples, getting state-of-the-art results on six attack methods with three datasets. Code is available at https://github.com/giladcohen/NNIF_adv_defense.

Natalia Martinez, Martin Bertran, Guillermo Sapiro et al.

Extracting the instantaneous heart rate (iHR) from face videos has been well studied in recent years. It is well known that changes in skin color due to blood flow can be captured using conventional cameras. One of the main limitations of methods that rely on this principle is the need of an illumination source. Moreover, they have to be able to operate under different light conditions. One way to avoid these constraints is using infrared cameras, allowing the monitoring of iHR under low light conditions. In this work, we present a simple, principled signal extraction method that recovers the iHR from infrared face videos. We tested the procedure on 7 participants, for whom we recorded an electrocardiogram simultaneously with their infrared face video. We checked that the recovered signal matched the ground truth iHR, showing that infrared is a promising alternative to conventional video imaging for heart rate monitoring, especially in low light conditions. Code is available at https://github.com/natalialmg/IR_iHR

Chacha Chen, Matthew Jörke, Adam Goliński et al.

Modern AI systems are being deployed in complex domains such as medicine, science, and law, where it is important that they not only produce correct answers, but also represent and update uncertain beliefs about the world as new evidence arrives. We introduce the novel technique of studying LLMs as information processing rules and utilize the information processing gap to study the internal (in)consistencies of how LLMs update their probabilistic beliefs from evidence. Our extensive experiments evaluate multiple approaches in which LLMs can incorporate evidence into their beliefs. Some of these approaches produce (nearly) Bayesian updates; others seem to use a learned heuristic. Surprisingly, the non-Bayesian heuristic updates often outperform exact Bayesian computation in terms of downstream task performance -- indicating the LLMs' probabilistic models of the world are misspecified. Lastly, we show how our measure can provide diagnostics to identify issues with LLM-powered inferential systems.

Young Kyung Kim, Oded Schlesinger, Yuzhou Zhao et al.

While state-of-the-art image generation models achieve remarkable visual quality, their internal generative processes remain a "black box." This opacity limits human observation and intervention, and poses a barrier to ensuring model reliability, safety, and control. Furthermore, their non-human-like workflows make them difficult for human observers to interpret. To address this, we introduce the Chain-of-Image Generation (CoIG) framework, which reframes image generation as a sequential, semantic process analogous to how humans create art. Similar to the advantages in monitorability and performance that Chain-of-Thought (CoT) brought to large language models (LLMs), CoIG can produce equivalent benefits in text-to-image generation. CoIG utilizes an LLM to decompose a complex prompt into a sequence of simple, step-by-step instructions. The image generation model then executes this plan by progressively generating and editing the image. Each step focuses on a single semantic entity, enabling direct monitoring. We formally assess this property using two novel metrics: CoIG Readability, which evaluates the clarity of each intermediate step via its corresponding output; and Causal Relevance, which quantifies the impact of each procedural step on the final generated image. We further show that our framework mitigates entity collapse by decomposing the complex generation task into simple subproblems, analogous to the procedural reasoning employed by CoT. Our experimental results indicate that CoIG substantially enhances quantitative monitorability while achieving competitive compositional robustness compared to established baseline models. The framework is model-agnostic and can be integrated with any image generation model.

Antoine Wehenkel, Juan L. Gamella, Ozan Sener et al. · eth-zurich

Driven by steady progress in deep generative modeling, simulation-based inference (SBI) has emerged as the workhorse for inferring the parameters of stochastic simulators. However, recent work has demonstrated that model misspecification can compromise the reliability of SBI, preventing its adoption in important applications where only misspecified simulators are available. This work introduces robust posterior estimation~(RoPE), a framework that overcomes model misspecification with a small real-world calibration set of ground-truth parameter measurements. We formalize the misspecification gap as the solution of an optimal transport~(OT) problem between learned representations of real-world and simulated observations, allowing RoPE to learn a model of the misspecification without placing additional assumptions on its nature. RoPE demonstrates how OT and a calibration set provide a controllable balance between calibrated uncertainty and informative inference, even under severely misspecified simulators. Results on four synthetic tasks and two real-world problems with ground-truth labels demonstrate that RoPE outperforms baselines and consistently returns informative and calibrated credible intervals.

Young Kyung Kim, J. Matías Di Martino, Guillermo Sapiro

Tokens or patches within Vision Transformers (ViT) lack essential semantic information, unlike their counterparts in natural language processing (NLP). Typically, ViT tokens are associated with rectangular image patches that lack specific semantic context, making interpretation difficult and failing to effectively encapsulate information. We introduce a novel transformer model, Semantic Vision Transformers (sViT), which leverages recent progress on segmentation models to design novel tokenizer strategies. sViT effectively harnesses semantic information, creating an inductive bias reminiscent of convolutional neural networks while capturing global dependencies and contextual information within images that are characteristic of transformers. Through validation using real datasets, sViT demonstrates superiority over ViT, requiring less training data while maintaining similar or superior performance. Furthermore, sViT demonstrates significant superiority in out-of-distribution generalization and robustness to natural distribution shifts, attributed to its scale invariance semantic characteristic. Notably, the use of semantic tokens significantly enhances the model's interpretability. Lastly, the proposed paradigm facilitates the introduction of new and powerful augmentation techniques at the token (or segment) level, increasing training data diversity and generalization capabilities. Just as sentences are made of words, images are formed by semantic objects; our proposed methodology leverages recent progress in object segmentation and takes an important and natural step toward interpretable and robust vision transformers.

Laura Manduchi, Antoine Wehenkel, Jens Behrmann et al.

Whole-body hemodynamics simulators, which model blood flow and pressure waveforms as functions of physiological parameters, are now essential tools for studying cardiovascular systems. However, solving the corresponding inverse problem of mapping observations (e.g., arterial pressure waveforms at specific locations in the arterial network) back to plausible physiological parameters remains challenging. Leveraging recent advances in simulation-based inference, we cast this problem as statistical inference by training an amortized neural posterior estimator on a newly built large dataset of cardiac simulations that we publicly release. To better align simulated data with real-world measurements, we incorporate stochastic elements modeling exogenous effects. The proposed framework can further integrate in-vivo data sources to refine its predictive capabilities on real-world data. In silico, we demonstrate that the proposed framework enables finely quantifying uncertainty associated with individual measurements, allowing trustworthy prediction of four biomarkers of clinical interest--namely Heart Rate, Cardiac Output, Systemic Vascular Resistance, and Left Ventricular Ejection Time--from arterial pressure waveforms and photoplethysmograms. Furthermore, we validate the framework in vivo, where our method accurately captures temporal trends in CO and SVR monitoring on the VitalDB dataset. Finally, the predictive error made by the model monotonically increases with the predicted uncertainty, thereby directly supporting the automatic rejection of unusable measurements.

Afroditi Papadaki, Natalia Martinez, Martin Bertran et al.

Current approaches to group fairness in federated learning assume the existence of predefined and labeled sensitive groups during training. However, due to factors ranging from emerging regulations to dynamics and location-dependency of protected groups, this assumption may be unsuitable in many real-world scenarios. In this work, we propose a new approach to guarantee group fairness that does not rely on any predefined definition of sensitive groups or additional labels. Our objective allows the federation to learn a Pareto efficient global model ensuring worst-case group fairness and it enables, via a single hyper-parameter, trade-offs between fairness and utility, subject only to a group size constraint. This implies that any sufficiently large subset of the population is guaranteed to receive at least a minimum level of utility performance from the model. The proposed objective encompasses existing approaches as special cases, such as empirical risk minimization and subgroup robustness objectives from centralized machine learning. We provide an algorithm to solve this problem in federation that enjoys convergence and excess risk guarantees. Our empirical results indicate that the proposed approach can effectively improve the worst-performing group that may be present without unnecessarily hurting the average performance, exhibits superior or comparable performance to relevant baselines, and achieves a large set of solutions with different fairness-utility trade-offs.

Young Kyung Kim, Oded Schlesinger, Qiangqiang Wu et al.

Test-Time Adaptation (TTA) enables pre-trained models to adjust to distribution shift by learning from unlabeled test-time streams. However, existing methods typically treat these streams as independent samples, overlooking the supervisory signal inherent in temporal dynamics. To address this, we introduce Order-Aware Test-Time Adaptation (OATTA). We formulate test-time adaptation as a gradient-free recursive Bayesian estimation task, using a learned dynamic transition matrix as a temporal prior to refine the base model's predictions. To ensure safety in weakly structured streams, we introduce a likelihood-ratio gate (LLR) that reverts to the base predictor when temporal evidence is absent. OATTA is a lightweight, model-agnostic module that incurs negligible computational overhead. Extensive experiments across image classification, wearable and physiological signal analysis, and language sentiment analysis demonstrate its universality; OATTA consistently boosts established baselines, improving accuracy by up to 6.35%. Our findings establish that modeling temporal dynamics provides a critical, orthogonal signal beyond standard order-agnostic TTA approaches.

Emanuele Palumbo, Sorawit Saengkyongam, Maria R. Cervera et al.

Continuous cardiovascular monitoring can play a key role in precision health. However, some fundamental cardiac biomarkers of interest, including stroke volume and cardiac output, require invasive measurements, e.g., arterial pressure waveforms (APW). As a non-invasive alternative, photoplethysmography (PPG) measurements are routinely collected in hospital settings. Unfortunately, the prediction of key cardiac biomarkers from PPG instead of APW remains an open challenge, further complicated by the scarcity of annotated PPG measurements. As a solution, we propose a hybrid approach that uses hemodynamic simulations and unlabeled clinical data to estimate cardiovascular biomarkers directly from PPG signals. Our hybrid model combines a conditional variational autoencoder trained on paired PPG-APW data with a conditional density estimator of cardiac biomarkers trained on labeled simulated APW segments. As a key result, our experiments demonstrate that the proposed approach can detect fluctuations of cardiac output and stroke volume and outperform a supervised baseline in monitoring temporal changes in these biomarkers.

Kyle Verrier, Achille Nazaret, Joseph Futoma et al.

Whether wearable photoplethysmography (PPG) contains dietary information remains unknown. We trained a language model on 1.1M meals to predict meal descriptions from PPG, aligning PPG to text. PPG nontrivially predicts meal content; predictability decreases for PPGs farther from meals. This transfers to dietary tasks: PPG increases AUC by 11% for intake and satiety across held-out and independent cohorts, with gains robust to text degradation. Wearable PPG may enable passive dietary monitoring.

Lun Huang, You Xie, Hongyi Xu et al.

Diffusion Transformers have demonstrated remarkable capabilities in visual synthesis, yet they often struggle with high-level semantic reasoning and long-horizon planning. This limitation frequently leads to visual hallucinations and mis-alignments with user instructions, especially in scenarios involving complex scene understanding, human-object interactions, multi-stage actions, and in-context motion reasoning. To address these challenges, we propose Plan-X, a framework that explicitly enforces high-level semantic planning to instruct video generation process. At its core lies a Semantic Planner, a learnable multimodal language model that reasons over the user's intent from both text prompts and visual context, and autoregressively generates a sequence of text-grounded spatio-temporal semantic tokens. These semantic tokens, complementary to high-level text prompt guidance, serve as structured "semantic sketches" over time for the video diffusion model, which has its strength at synthesizing high-fidelity visual details. Plan-X effectively integrates the strength of language models in multimodal in-context reasoning and planning, together with the strength of diffusion models in photorealistic video synthesis. Extensive experiments demonstrate that our framework substantially reduces visual hallucinations and enables fine-grained, instruction-aligned video generation consistent with multimodal context.

Jens Behrmann, Maria R. Cervera, Antoine Wehenkel et al.

Smart wearables enable continuous tracking of established biomarkers such as heart rate, heart rate variability, and blood oxygen saturation via photoplethysmography (PPG). Beyond these metrics, PPG waveforms contain richer physiological information, as recent deep learning (DL) studies demonstrate. However, DL models often rely on features with unclear physiological meaning, creating a tension between predictive power, clinical interpretability, and sensor design. We address this gap by introducing PPGen, a biophysical model that relates PPG signals to interpretable physiological and optical parameters. Building on PPGen, we propose hybrid amortized inference (HAI), enabling fast, robust, and scalable estimation of relevant physiological parameters from PPG signals while correcting for model misspecification. In extensive in-silico experiments, we show that HAI can accurately infer physiological parameters under diverse noise and sensor conditions. Our results illustrate a path toward PPG models that retain the fidelity needed for DL-based features while supporting clinical interpretation and informed hardware design.

Antoine Wehenkel, Jens Behrmann, Hsiang Hsu et al.

Hybrid modelling reduces the misspecification of expert models by combining them with machine learning (ML) components learned from data. Similarly to many ML algorithms, hybrid model performance guarantees are limited to the training distribution. Leveraging the insight that the expert model is usually valid even outside the training domain, we overcome this limitation by introducing a hybrid data augmentation strategy termed \textit{expert augmentation}. Based on a probabilistic formalization of hybrid modelling, we demonstrate that expert augmentation, which can be incorporated into existing hybrid systems, improves generalization. We empirically validate the expert augmentation on three controlled experiments modelling dynamical systems with ordinary and partial differential equations. Finally, we assess the potential real-world applicability of expert augmentation on a dataset of a real double pendulum.

Afroditi Papadaki, Natalia Martinez, Martin Bertran et al.

Federated learning is an increasingly popular paradigm that enables a large number of entities to collaboratively learn better models. In this work, we study minimax group fairness in federated learning scenarios where different participating entities may only have access to a subset of the population groups during the training phase. We formally analyze how our proposed group fairness objective differs from existing federated learning fairness criteria that impose similar performance across participants instead of demographic groups. We provide an optimization algorithm -- FedMinMax -- for solving the proposed problem that provably enjoys the performance guarantees of centralized learning algorithms. We experimentally compare the proposed approach against other state-of-the-art methods in terms of group fairness in various federated learning setups, showing that our approach exhibits competitive or superior performance.

Afroditi Papadaki, Natalia Martinez, Martin Bertran et al.

Federated learning is an increasingly popular paradigm that enables a large number of entities to collaboratively learn better models. In this work, we study minmax group fairness in paradigms where different participating entities may only have access to a subset of the population groups during the training phase. We formally analyze how this fairness objective differs from existing federated learning fairness criteria that impose similar performance across participants instead of demographic groups. We provide an optimization algorithm -- FedMinMax -- for solving the proposed problem that provably enjoys the performance guarantees of centralized learning algorithms. We experimentally compare the proposed approach against other methods in terms of group fairness in various federated learning setups.

William E. Carson, Dmitry Isaev, Samatha Major et al.

Discovering reliable measures that inform on autism spectrum disorder (ASD) diagnosis is critical for providing appropriate and timely treatment for this neurodevelopmental disorder. In this work we present applications of adversarial linear factor models in the creation of improved biomarkers for ASD diagnosis. First, we demonstrate that an adversarial linear factor model can be used to remove confounding information from our biomarkers, ensuring that they contain only pertinent information on ASD. Second, we show this same model can be used to learn a disentangled representation of multimodal biomarkers that results in an increase in predictive performance. These results demonstrate that adversarial methods can address both biomarker confounds and improve biomarker predictive performance.

Chenfei Wu, Lun Huang, Qianxi Zhang et al.

Generating videos from text is a challenging task due to its high computational requirements for training and infinite possible answers for evaluation. Existing works typically experiment on simple or small datasets, where the generalization ability is quite limited. In this work, we propose GODIVA, an open-domain text-to-video pretrained model that can generate videos from text in an auto-regressive manner using a three-dimensional sparse attention mechanism. We pretrain our model on Howto100M, a large-scale text-video dataset that contains more than 136 million text-video pairs. Experiments show that GODIVA not only can be fine-tuned on downstream video generation tasks, but also has a good zero-shot capability on unseen texts. We also propose a new metric called Relative Matching (RM) to automatically evaluate the video generation quality. Several challenges are listed and discussed as future work.

Ze Wang, Sihao Ding, Ying Li et al.

In this paper, we introduce Cirrus, a new long-range bi-pattern LiDAR public dataset for autonomous driving tasks such as 3D object detection, critical to highway driving and timely decision making. Our platform is equipped with a high-resolution video camera and a pair of LiDAR sensors with a 250-meter effective range, which is significantly longer than existing public datasets. We record paired point clouds simultaneously using both Gaussian and uniform scanning patterns. Point density varies significantly across such a long range, and different scanning patterns further diversify object representation in LiDAR. In Cirrus, eight categories of objects are exhaustively annotated in the LiDAR point clouds for the entire effective range. To illustrate the kind of studies supported by this new dataset, we introduce LiDAR model adaptation across different ranges, scanning patterns, and sensor devices. Promising results show the great potential of this new dataset to the robotics and computer vision communities.

Haoyu Dong, Ze Wang, Qiang Qiu et al.

Image retrieval relies heavily on the quality of the data modeling and the distance measurement in the feature space. Building on the concept of image manifold, we first propose to represent the feature space of images, learned via neural networks, as a graph. Neighborhoods in the feature space are now defined by the geodesic distance between images, represented as graph vertices or manifold samples. When limited images are available, this manifold is sparsely sampled, making the geodesic computation and the corresponding retrieval harder. To address this, we augment the manifold samples with geometrically aligned text, thereby using a plethora of sentences to teach us about images. In addition to extensive results on standard datasets illustrating the power of text to help in image retrieval, a new public dataset based on CLEVR is introduced to quantify the semantic similarity between visual data and text data. The experimental results show that the joint embedding manifold is a robust representation, allowing it to be a better basis to perform image retrieval given only an image and a textual instruction on the desired modifications over the image

Natalia Martinez, Martin Bertran, Guillermo Sapiro

In this work we formulate and formally characterize group fairness as a multi-objective optimization problem, where each sensitive group risk is a separate objective. We propose a fairness criterion where a classifier achieves minimax risk and is Pareto-efficient w.r.t. all groups, avoiding unnecessary harm, and can lead to the best zero-gap model if policy dictates so. We provide a simple optimization algorithm compatible with deep neural networks to satisfy these constraints. Since our method does not require test-time access to sensitive attributes, it can be applied to reduce worst-case classification errors between outcomes in unbalanced classification problems. We test the proposed methodology on real case-studies of predicting income, ICU patient mortality, skin lesions classification, and assessing credit risk, demonstrating how our framework compares favorably to other approaches.

Martin Bertran, Natalia Martinez, Mariano Phielipp et al.

Agents trained via deep reinforcement learning (RL) routinely fail to generalize to unseen environments, even when these share the same underlying dynamics as the training levels. Understanding the generalization properties of RL is one of the challenges of modern machine learning. Towards this goal, we analyze policy learning in the context of Partially Observable Markov Decision Processes (POMDPs) and formalize the dynamics of training levels as instances. We prove that, independently of the exploration strategy, reusing instances introduces significant changes on the effective Markov dynamics the agent observes during training. Maximizing expected rewards impacts the learned belief state of the agent by inducing undesired instance specific speedrunning policies instead of generalizeable ones, which are suboptimal on the training set. We provide generalization bounds to the value gap in train and test environments based on the number of training instances, and use insights based on these to improve performance on unseen levels. We propose training a shared belief representation over an ensemble of specialized policies, from which we compute a consensus policy that is used for data collection, disallowing instance specific exploitation. We experimentally validate our theory, observations, and the proposed computational solution over the CoinRun benchmark.

Ze Wang, Xiuyuan Cheng, Guillermo Sapiro et al.

Convolutional Neural Networks (CNNs) are known to be significantly over-parametrized, and difficult to interpret, train and adapt. In this paper, we introduce a structural regularization across convolutional kernels in a CNN. In our approach, each convolution kernel is first decomposed as 2D dictionary atoms linearly combined by coefficients. The widely observed correlation and redundancy in a CNN hint a common low-rank structure among the decomposed coefficients, which is here further supported by our empirical observations. We then explicitly regularize CNN kernels by enforcing decomposed coefficients to be shared across sub-structures, while leaving each sub-structure only its own dictionary atoms, a few hundreds of parameters typically, which leads to dramatic model reductions. We explore models with sharing across different sub-structures to cover a wide range of trade-offs between parameter reduction and expressiveness. Our proposed regularized network structures open the door to better interpreting, training and adapting deep models. We validate the flexibility and compatibility of our method by image classification experiments on multiple datasets and underlying network structures, and show that CNNs now maintain performance with dramatic reduction in parameters and computations, e.g., only 5\% parameters are used in a ResNet-18 to achieve comparable performance. Further experiments on few-shot classification show that faster and more robust task adaptation is obtained in comparison with models with standard convolutions.

Raphaël Achddou, J. Matias di Martino, Guillermo Sapiro

Standard deep neural networks (DNNs) are commonly trained in an end-to-end fashion for specific tasks such as object recognition, face identification, or character recognition, among many examples. This specificity often leads to overconfident models that generalize poorly to samples that are not from the original training distribution. Moreover, such standard DNNs do not allow to leverage information from heterogeneously annotated training data, where for example, labels may be provided with different levels of granularity. Furthermore, DNNs do not produce results with simultaneous different levels of confidence for different levels of detail, they are most commonly an all or nothing approach. To address these challenges, we introduce the concept of nested learning: how to obtain a hierarchical representation of the input such that a coarse label can be extracted first, and sequentially refine this representation, if the sample permits, to obtain successively refined predictions, all of them with the corresponding confidence. We explicitly enforce this behavior by creating a sequence of nested information bottlenecks. Looking at the problem of nested learning from an information theory perspective, we design a network topology with two important properties. First, a sequence of low dimensional (nested) feature embeddings are enforced. Then we show how the explicit combination of nested outputs can improve both the robustness and the accuracy of finer predictions. Experimental results on Cifar-10, Cifar-100, MNIST, Fashion-MNIST, Dbpedia, and Plantvillage demonstrate that nested learning outperforms the same network trained in the standard end-to-end fashion.

J. Matias Di Martino, Fernando Suzacq, Mauricio Delbracio et al.

Active illumination is a prominent complement to enhance 2D face recognition and make it more robust, e.g., to spoofing attacks and low-light conditions. In the present work we show that it is possible to adopt active illumination to enhance state-of-the-art 2D face recognition approaches with 3D features, while bypassing the complicated task of 3D reconstruction. The key idea is to project over the test face a high spatial frequency pattern, which allows us to simultaneously recover real 3D information plus a standard 2D facial image. Therefore, state-of-the-art 2D face recognition solution can be transparently applied, while from the high frequency component of the input image, complementary 3D facial features are extracted. Experimental results on ND-2006 dataset show that the proposed ideas can significantly boost face recognition performance and dramatically improve the robustness to spoofing attacks.

Natalia Martinez, Martin Bertran, Guillermo Sapiro

Common fairness definitions in machine learning focus on balancing notions of disparity and utility. In this work, we study fairness in the context of risk disparity among sub-populations. We are interested in learning models that minimize performance discrepancies across sensitive groups without causing unnecessary harm. This is relevant to high-stakes domains such as healthcare, where non-maleficence is a core principle. We formalize this objective using Pareto frontiers, and provide analysis, based on recent works in fairness, to exemplify scenarios were perfect fairness might not be feasible without doing unnecessary harm. We present a methodology for training neural networks that achieve our goal by dynamically re-balancing subgroups risks. We argue that even in domains where fairness at cost is required, finding a non-unnecessary-harm fairness model is the optimal initial step. We demonstrate this methodology on real case-studies of predicting ICU patient mortality, and classifying skin lesions from dermatoscopic images.

Zhuoqing Chang, Matias Di Martino, Qiang Qiu et al.

Traditional gaze estimation methods typically require explicit user calibration to achieve high accuracy. This process is cumbersome and recalibration is often required when there are changes in factors such as illumination and pose. To address this challenge, we introduce SalGaze, a framework that utilizes saliency information in the visual content to transparently adapt the gaze estimation algorithm to the user without explicit user calibration. We design an algorithm to transform a saliency map into a differentiable loss map that can be used for the optimization of CNN-based models. SalGaze is also able to greatly augment standard point calibration data with implicit video saliency calibration data using a unified framework. We show accuracy improvements over 24% using our technique on existing methods.

Ze Wang, Sihao Ding, Ying Li et al.

LiDAR-based 3D object detection plays a crucial role in modern autonomous driving systems. LiDAR data often exhibit severe changes in properties across different observation ranges. In this paper, we explore cross-range adaptation for 3D object detection using LiDAR, i.e., far-range observations are adapted to near-range. This way, far-range detection is optimized for similar performance to near-range one. We adopt a bird-eyes view (BEV) detection framework to perform the proposed model adaptation. Our model adaptation consists of an adversarial global adaptation, and a fine-grained local adaptation. The proposed cross range adaptation framework is validated on three state-of-the-art LiDAR based object detection networks, and we consistently observe performance improvement on the far-range objects, without adding any auxiliary parameters to the model. To the best of our knowledge, this paper is the first attempt to study cross-range LiDAR adaptation for object detection in point clouds. To demonstrate the generality of the proposed adaptation framework, experiments on more challenging cross-device adaptation are further conducted, and a new LiDAR dataset with high-quality annotated point clouds is released to promote future research.

Ze Wang, Xiuyuan Cheng, Guillermo Sapiro et al.

While generative adversarial networks (GANs) have revolutionized machine learning, a number of open questions remain to fully understand them and exploit their power. One of these questions is how to efficiently achieve proper diversity and sampling of the multi-mode data space. To address this, we introduce BasisGAN, a stochastic conditional multi-mode image generator. By exploiting the observation that a convolutional filter can be well approximated as a linear combination of a small set of basis elements, we learn a plug-and-played basis generator to stochastically generate basis elements, with just a few hundred of parameters, to fully embed stochasticity into convolutional filters. By sampling basis elements instead of filters, we dramatically reduce the cost of modeling the parameter space with no sacrifice on either image diversity or fidelity. To illustrate this proposed plug-and-play framework, we construct variants of BasisGAN based on state-of-the-art conditional image generation networks, and train the networks by simply plugging in a basis generator, without additional auxiliary components, hyperparameters, or training objectives. The experimental success is complemented with theoretical results indicating how the perturbations introduced by the proposed sampling of basis elements can propagate to the appearance of generated images.

Ze Wang, Xiuyuan Cheng, Guillermo Sapiro et al.

In this paper, we consider domain-invariant deep learning by explicitly modeling domain shifts with only a small amount of domain-specific parameters in a Convolutional Neural Network (CNN). By exploiting the observation that a convolutional filter can be well approximated as a linear combination of a small set of dictionary atoms, we show for the first time, both empirically and theoretically, that domain shifts can be effectively handled by decomposing a convolutional layer into a domain-specific atom layer and a domain-shared coefficient layer, while both remain convolutional. An input channel will now first convolve spatially only with each respective domain-specific dictionary atom to "absorb" domain variations, and then output channels are linearly combined using common decomposition coefficients trained to promote shared semantics across domains. We use toy examples, rigorous analysis, and real-world examples with diverse datasets and architectures, to show the proposed plug-in framework's effectiveness in cross and joint domain performance and domain adaptation. With the proposed architecture, we need only a small set of dictionary atoms to model each additional domain, which brings a negligible amount of additional parameters, typically a few hundred.

Wei Zhu, Qiang Qiu, Robert Calderbank et al.

Encoding the scale information explicitly into the representation learned by a convolutional neural network (CNN) is beneficial for many computer vision tasks especially when dealing with multiscale inputs. We study, in this paper, a scaling-translation-equivariant (ST-equivariant) CNN with joint convolutions across the space and the scaling group, which is shown to be both sufficient and necessary to achieve equivariance for the regular representation of the scaling-translation group ST . To reduce the model complexity and computational burden, we decompose the convolutional filters under two pre-fixed separable bases and truncate the expansion to low-frequency components. A further benefit of the truncated filter expansion is the improved deformation robustness of the equivariant representation, a property which is theoretically analyzed and empirically verified. Numerical experiments demonstrate that the proposed scaling-translation-equivariant network with decomposed convolutional filters (ScDCFNet) achieves significantly improved performance in multiscale image classification and better interpretability than regular CNNs at a reduced model size.

Reuben R Shamir, Yuval Duchin, Jinyoung Kim et al.

Objective: Overlapping measures are often utilized to quantify the similarity between two binary regions. However, modern segmentation algorithms output a probability or confidence map with continuous values in the zero-to-one interval. Moreover, these binary overlapping measures are biased to structure size. Addressing these challenges is the objective of this work. Methods: We extend the definition of the classical Dice coefficient (DC) overlap to facilitate the direct comparison of a ground truth binary image with a probabilistic map. We call the extended method continuous Dice coefficient (cDC) and show that 1) cDC is less or equal to 1 and cDC = 1 if-and-only-if the structures overlap is complete, and, 2) cDC is monotonically decreasing with the amount of overlap. We compare the classical DC and the cDC in a simulation of partial volume effects that incorporates segmentations of common targets for deep-brainstimulation. Lastly, we investigate the cDC for an automatic segmentation of the subthalamic-nucleus. Results: Partial volume effect simulation on thalamus (large structure) resulted with DC and cDC averages (SD) of 0.98 (0.006) and 0.99 (0.001), respectively. For subthalamic-nucleus (small structure) DC and cDC were 0.86 (0.025) and 0.97 (0.006), respectively. The DC and cDC for automatic STN segmentation were 0.66 and 0.80, respectively. Conclusion: The cDC is well defined for probabilistic segmentation, less biased to structure size and more robust to partial volume effects in comparison to DC. Significance: The proposed method facilitates a better evaluation of segmentation algorithms. As a better measurement tool, it opens the door for the development of better segmentation methods.

Martin Bertran, Natalia Martinez, Afroditi Papadaki et al.

It is becoming increasingly clear that users should own and control their data. Utility providers are also becoming more interested in guaranteeing data privacy. As such, users and utility providers should collaborate in data privacy, a paradigm that has not yet been developed in the privacy research community. We introduce this concept and present explicit architectures where the user controls what characteristics of the data she/he wants to share and what she/he wants to keep private. This is achieved by collaborative learning a sensitization function, either a deterministic or a stochastic one, that retains valuable information for the utility tasks but it also eliminates necessary information for the privacy ones. As illustration examples, we implement them using a plug-and-play approach, where no algorithm is changed at the system provider end, and an adversarial approach, where minor re-training of the privacy inferring engine is allowed. In both cases the learned sanitization function keeps the data in the original domain, thereby allowing the system to use the same algorithms it was using before for both original and privatized data. We show how we can maintain utility while fully protecting private information if the user chooses to do so, even when the first is harder than the second, as in the case here illustrated of identity detection while hiding gender.

Wei Zhu, Qiang Qiu, Bao Wang et al.

Deep neural networks (DNNs) typically have enough capacity to fit random data by brute force even when conventional data-dependent regularizations focusing on the geometry of the features are imposed. We find out that the reason for this is the inconsistency between the enforced geometry and the standard softmax cross entropy loss. To resolve this, we propose a new framework for data-dependent DNN regularization, the Geometrically-Regularized-Self-Validating neural Networks (GRSVNet). During training, the geometry enforced on one batch of features is simultaneously validated on a separate batch using a validation loss consistent with the geometry. We study a particular case of GRSVNet, the Orthogonal-Low-rank Embedding (OLE)-GRSVNet, which is capable of producing highly discriminative features residing in orthogonal low-rank subspaces. Numerical experiments show that OLE-GRSVNet outperforms DNNs with conventional regularization when trained on real data. More importantly, unlike conventional DNNs, OLE-GRSVNet refuses to memorize random data or random labels, suggesting it only learns intrinsic patterns by reducing the memorizing capacity of the baseline DNN.

Xiuyuan Cheng, Qiang Qiu, Robert Calderbank et al.

Explicit encoding of group actions in deep features makes it possible for convolutional neural networks (CNNs) to handle global deformations of images, which is critical to success in many vision tasks. This paper proposes to decompose the convolutional filters over joint steerable bases across the space and the group geometry simultaneously, namely a rotation-equivariant CNN with decomposed convolutional filters (RotDCF). This decomposition facilitates computing the joint convolution, which is proved to be necessary for the group equivariance. It significantly reduces the model size and computational complexity while preserving performance, and truncation of the bases expansion serves implicitly to regularize the filters. On datasets involving in-plane and out-of-plane object rotations, RotDCF deep features demonstrate greater robustness and interpretability than regular CNNs. The stability of the equivariant representation to input variations is also proved theoretically under generic assumptions on the filters in the decomposed form. The RotDCF framework can be extended to groups other than rotations, providing a general approach which achieves both group equivariance and representation stability at a reduced model size.

Gilad Cohen, Guillermo Sapiro, Raja Giryes

This paper studies the relationship between the classification performed by deep neural networks (DNNs) and the decision of various classical classifiers, namely k-nearest neighbours (k-NN), support vector machines (SVM) and logistic regression (LR), at various layers of the network. This comparison provides us with new insights as to the ability of neural networks to both memorize the training data and generalize to new data at the same time, where k-NN serves as the ideal estimator that perfectly memorizes the data. We show that memorization of non-generalizing networks happens only at the last layers. Moreover, the behavior of DNNs compared to the linear classifiers SVM and LR is quite the same on the training and test data regardless of whether the network generalizes. On the other hand, the similarity to k-NN holds only at the absence of overfitting. Our results suggests that k-NN behavior of the network on new data is a sign of generalization. Moreover, it shows that memorization and generalization, which are traditionally considered to be contradicting to each other, are compatible and complementary.

J. Matias Di Martino, Qiang Qiu, Trishul Nagenalli et al.

Spoofing attacks are a threat to modern face recognition systems. In this work we present a simple yet effective liveness detection approach to enhance 2D face recognition methods and make them robust against spoofing attacks. We show that the risk to spoofing attacks can be re- duced through the use of an additional source of light, for example a flash. From a pair of input images taken under different illumination, we define discriminative features that implicitly contain facial three-dimensional in- formation. Furthermore, we show that when multiple sources of light are considered, we are able to validate which one has been activated. This makes possible the design of a highly secure active-light authentication framework. Finally, further investigating the use of 3D features without 3D reconstruction, we introduce an approximated disparity-based implicit 3D feature obtained from an uncalibrated stereo-pair of cameras. Valida- tion experiments show that the proposed methods produce state-of-the-art results in challenging scenarios with nearly no feature extraction latency.

Albert Gong, Qiang Qiu, Guillermo Sapiro

In this paper we introduce an ensemble method for convolutional neural network (CNN), called "virtual branching," which can be implemented with nearly no additional parameters and computation on top of standard CNNs. We propose our method in the context of person re-identification (re-ID). Our CNN model consists of shared bottom layers, followed by "virtual" branches, where neurons from a block of regular convolutional and fully-connected layers are partitioned into multiple sets. Each virtual branch is trained with different data to specialize in different aspects, e.g., a specific body region or pose orientation. In this way, robust ensemble representations are obtained against human body misalignment, deformations, or variations in viewing angles, at nearly no any additional cost. The proposed method achieves competitive performance on multiple person re-ID benchmark datasets, including Market-1501, CUHK03, and DukeMTMC-reID.

Qiang Qiu, Xiuyuan Cheng, Robert Calderbank et al.

Filters in a Convolutional Neural Network (CNN) contain model parameters learned from enormous amounts of data. In this paper, we suggest to decompose convolutional filters in CNN as a truncated expansion with pre-fixed bases, namely the Decomposed Convolutional Filters network (DCFNet), where the expansion coefficients remain learned from data. Such a structure not only reduces the number of trainable parameters and computation, but also imposes filter regularity by bases truncation. Through extensive experiments, we consistently observe that DCFNet maintains accuracy for image classification tasks with a significant reduction of model parameters, particularly with Fourier-Bessel (FB) bases, and even with random bases. Theoretically, we analyze the representation stability of DCFNet with respect to input variations, and prove representation stability under generic assumptions on the expansion coefficients. The analysis is consistent with the empirical observations.

José Lezama, Qiang Qiu, Pablo Musé et al.

Deep neural networks trained using a softmax layer at the top and the cross-entropy loss are ubiquitous tools for image classification. Yet, this does not naturally enforce intra-class similarity nor inter-class margin of the learned deep representations. To simultaneously achieve these two goals, different solutions have been proposed in the literature, such as the pairwise or triplet losses. However, such solutions carry the extra task of selecting pairs or triplets, and the extra computational burden of computing and learning for many combinations of them. In this paper, we propose a plug-and-play loss term for deep networks that explicitly reduces intra-class variance and enforces inter-class margin simultaneously, in a simple and elegant geometric manner. For each class, the deep features are collapsed into a learned linear subspace, or union of them, and inter-class subspaces are pushed to be as orthogonal as possible. Our proposed Orthogonal Low-rank Embedding (OLÉ) does not require carefully crafting pairs or triplets of samples for training, and works standalone as a classification loss, being the first reported deep metric learning framework of its kind. Because of the improved margin between features of different classes, the resulting deep networks generalize better, are more discriminative, and more robust. We demonstrate improved classification performance in general object recognition, plugging the proposed loss term into existing off-the-shelf architectures. In particular, we show the advantage of the proposed loss in the small data/model scenario, and we significantly advance the state-of-the-art on the Stanford STL-10 benchmark.

Qiang Qiu, Jose Lezama, Alex Bronstein et al.

Hash codes are efficient data representations for coping with the ever growing amounts of data. In this paper, we introduce a random forest semantic hashing scheme that embeds tiny convolutional neural networks (CNN) into shallow random forests, with near-optimal information-theoretic code aggregation among trees. We start with a simple hashing scheme, where random trees in a forest act as hashing functions by setting `1' for the visited tree leaf, and `0' for the rest. We show that traditional random forests fail to generate hashes that preserve the underlying similarity between the trees, rendering the random forests approach to hashing challenging. To address this, we propose to first randomly group arriving classes at each tree split node into two groups, obtaining a significantly simplified two-class classification problem, which can be handled using a light-weight CNN weak learner. Such random class grouping scheme enables code uniqueness by enforcing each class to share its code with different classes in different trees. A non-conventional low-rank loss is further adopted for the CNN weak learners to encourage code consistency by minimizing intra-class variations and maximizing inter-class distance for the two random class groups. Finally, we introduce an information-theoretic approach for aggregating codes of individual trees into a single hash code, producing a near-optimal unique hash for each class. The proposed approach significantly outperforms state-of-the-art hashing methods for image retrieval tasks on large-scale public datasets, while performing at the level of other state-of-the-art image classification techniques while utilizing a more compact and efficient scalable representation. This work proposes a principled and robust procedure to train and deploy in parallel an ensemble of light-weight CNNs, instead of simply going deeper.

Wei Zhu, Qiang Qiu, Jiaji Huang et al.

Deep neural networks have proved very successful on archetypal tasks for which large training sets are available, but when the training data are scarce, their performance suffers from overfitting. Many existing methods of reducing overfitting are data-independent, and their efficacy is often limited when the training set is very small. Data-dependent regularizations are mostly motivated by the observation that data of interest lie close to a manifold, which is typically hard to parametrize explicitly and often requires human input of tangent vectors. These methods typically only focus on the geometry of the input data, and do not necessarily encourage the networks to produce geometrically meaningful features. To resolve this, we propose a new framework, the Low-Dimensional-Manifold-regularized neural Network (LDMNet), which incorporates a feature regularization method that focuses on the geometry of both the input data and the output features. In LDMNet, we regularize the network by encouraging the combination of the input data and the output features to sample a collection of low dimensional manifolds, which are searched efficiently without explicit parametrization. To achieve this, we directly use the manifold dimension as a regularization term in a variational functional. The resulting Euler-Lagrange equation is a Laplace-Beltrami equation over a point cloud, which is solved by the point integral method without increasing the computational complexity. We demonstrate two benefits of LDMNet in the experiments. First, we show that LDMNet significantly outperforms widely-used network regularizers such as weight decay and DropOut. Second, we show that LDMNet can be designed to extract common features of an object imaged via different modalities, which proves to be very useful in real-world applications such as cross-spectral face recognition.