We introduce a pioneering methodology for boosting large language models in the domain of protein representation learning. Our primary contribution lies in the refinement process for correlating the over-reliance on co-evolution knowledge, in a way that networks are trained to distill invaluable insights from negative samples, constituted by protein pairs sourced from disparate categories. By capitalizing on this novel approach, our technique steers the training of transformer-based models within the attention score space. This advanced strategy not only amplifies performance but also reflects the nuanced biological behaviors exhibited by proteins, offering aligned evidence with traditional biological mechanisms such as protein-protein interaction. We experimentally observed improved performance on various tasks over datasets, on top of several well-established large protein models. This innovative paradigm opens up promising horizons for further progress in the realms of protein research and computational biology.
Sleep is a complex physiological process evaluated through various modalities recording electrical brain, cardiac, and respiratory activities. We curate a large polysomnography dataset from over 14,000 participants comprising over 100,000 hours of multi-modal sleep recordings. Leveraging this extensive dataset, we developed SleepFM, the first multi-modal foundation model for sleep analysis. We show that a novel leave-one-out approach for contrastive learning significantly improves downstream task performance compared to representations from standard pairwise contrastive learning. A logistic regression model trained on SleepFM's learned embeddings outperforms an end-to-end trained convolutional neural network (CNN) on sleep stage classification (macro AUROC 0.88 vs 0.72 and macro AUPRC 0.72 vs 0.48) and sleep disordered breathing detection (AUROC 0.85 vs 0.69 and AUPRC 0.77 vs 0.61). Notably, the learned embeddings achieve 48% top-1 average accuracy in retrieving the corresponding recording clips of other modalities from 90,000 candidates. This work demonstrates the value of holistic multi-modal sleep modeling to fully capture the richness of sleep recordings. SleepFM is open source and available at https://github.com/rthapa84/sleepfm-codebase.
Recently, linear complexity sequence modeling networks have achieved modeling capabilities similar to Vision Transformers on a variety of computer vision tasks, while using fewer FLOPs and less memory. However, their advantage in terms of actual runtime speed is not significant. To address this issue, we introduce Gated Linear Attention (GLA) for vision, leveraging its superior hardware-awareness and efficiency. We propose direction-wise gating to capture 1D global context through bidirectional modeling and a 2D gating locality injection to adaptively inject 2D local details into 1D global context. Our hardware-aware implementation further merges forward and backward scanning into a single kernel, enhancing parallelism and reducing memory cost and latency. The proposed model, ViG, offers a favorable trade-off in accuracy, parameters, and FLOPs on ImageNet and downstream tasks, outperforming popular Transformer and CNN-based models. Notably, ViG-S matches DeiT-B's accuracy while using only 27% of the parameters and 20% of the FLOPs, running 2$\times$ faster on $224\times224$ images. At $1024\times1024$ resolution, ViG-T uses 5.2$\times$ fewer FLOPs, saves 90% GPU memory, runs 4.8$\times$ faster, and achieves 20.7% higher top-1 accuracy than DeiT-T. These results position ViG as an efficient and scalable solution for visual representation learning. Code is available at \url{https://github.com/hustvl/ViG}.
Time series analysis remains a major challenge due to its sparse characteristics, high dimensionality, and inconsistent data quality. Recent advancements in transformer-based techniques have enhanced capabilities in forecasting and imputation; however, these methods are still resource-heavy, lack adaptability, and face difficulties in integrating both local and global attributes of time series. To tackle these challenges, we propose a new architectural concept for time series analysis based on introspection. Central to this concept is the self-supervised pretraining of Time Series Representation Models (TSRMs), which once learned can be easily tailored and fine-tuned for specific tasks, such as forecasting and imputation, in an automated and resource-efficient manner. Our architecture is equipped with a flexible and hierarchical representation learning process, which is robust against missing data and outliers. It can capture and learn both local and global features of the structure, semantics, and crucial patterns of a given time series category, such as heart rate data. Our learned time series representation models can be efficiently adapted to a specific task, such as forecasting or imputation, without manual intervention. Furthermore, our architecture's design supports explainability by highlighting the significance of each input value for the task at hand. Our empirical study using four benchmark datasets shows that, compared to investigated state-of-the-art baseline methods, our architecture improves imputation and forecasting errors by up to 90.34% and 71.54%, respectively, while reducing the required trainable parameters by up to 92.43%. The source code is available at https://github.com/RobertLeppich/TSRM.
The goal of Fair Representation Learning (FRL) is to mitigate biases in machine learning models by learning data representations that enable high accuracy on downstream tasks while minimizing discrimination based on sensitive attributes. The evaluation of FRL methods in many recent works primarily focuses on the tradeoff between downstream fairness and accuracy with respect to a single task that was used to approximate the utility of representations during training (proxy task). This incentivizes retaining only features relevant to the proxy task while discarding all other information. In extreme cases, this can cause the learned representations to collapse to a trivial, binary value, rendering them unusable in transfer settings. In this work, we argue that this approach is fundamentally mismatched with the original motivation of FRL, which arises from settings with many downstream tasks unknown at training time (transfer tasks). To remedy this, we propose to refocus the evaluation protocol of FRL methods primarily around the performance on transfer tasks. A key challenge when conducting such an evaluation is the lack of adequate benchmarks. We address this by formulating four criteria that a suitable evaluation procedure should fulfill. Based on these, we propose TransFair, a benchmark that satisfies these criteria, consisting of novel variations of popular FRL datasets with carefully calibrated transfer tasks. In this setting, we reevaluate state-of-the-art FRL methods, observing that they often overfit to the proxy task, which causes them to underperform on certain transfer tasks. We further highlight the importance of task-agnostic learning signals for FRL methods, as they can lead to more transferrable representations.
Anomaly detection in continuous-time dynamic graphs is an emerging field yet under-explored in the context of learning algorithms. In this paper, we pioneer structured analyses of link-level anomalies and graph representation learning for identifying categorically anomalous graph links. First, we introduce a fine-grained taxonomy for edge-level anomalies leveraging structural, temporal, and contextual graph properties. Based on these properties, we introduce a method for generating and injecting typed anomalies into graphs. Next, we introduce a novel method to generate continuous-time dynamic graphs featuring consistencies across either or combinations of time, structure, and context. To enable temporal graph learning methods to detect specific types of anomalous links rather than the bare existence of a link, we extend the generic link prediction setting by: (1) conditioning link existence on contextual edge attributes; and (2) refining the training regime to accommodate diverse perturbations in the negative edge sampler. Comprehensive benchmarks on synthetic and real-world datasets -- featuring synthetic and labeled organic anomalies and employing six state-of-the-art link prediction methods -- validate our taxonomy and generation processes for anomalies and benign graphs, as well as our approach to adapting methods for anomaly detection. Our results reveal that different learning methods excel in capturing different aspects of graph normality and detecting different types of anomalies. We conclude with a comprehensive list of findings highlighting opportunities for future research.
We present RoboArm-NMP, a learning and evaluation environment that allows simple and thorough evaluations of Neural Motion Planning (NMP) algorithms, focused on robotic manipulators. Our Python-based environment provides baseline implementations for learning control policies (either supervised or reinforcement learning based), a simulator based on PyBullet, data of solved instances using a classical motion planning solver, various representation learning methods for encoding the obstacles, and a clean interface between the learning and planning frameworks. Using RoboArm-NMP, we compare several prominent NMP design points, and demonstrate that the best methods mostly succeed in generalizing to unseen goals in a scene with fixed obstacles, but have difficulty in generalizing to unseen obstacle configurations, suggesting focus points for future research.
Compositional generalization (the ability to respond correctly to novel combinations of familiar components) is thought to be a cornerstone of intelligent behavior. Compositionally structured (e.g. disentangled) representations support this ability; however, the conditions under which they are sufficient for the emergence of compositional generalization remain unclear. To address this gap, we present a theory of compositional generalization in kernel models with fixed, compositionally structured representations. This provides a tractable framework for characterizing the impact of training data statistics on generalization. We find that these models are limited to functions that assign values to each combination of components seen during training, and then sum up these values ("conjunction-wise additivity"). This imposes fundamental restrictions on the set of tasks compositionally structured kernel models can learn, in particular preventing them from transitively generalizing equivalence relations. Even for compositional tasks that they can learn in principle, we identify novel failure modes in compositional generalization (memorization leak and shortcut bias) that arise from biases in the training data. Finally, we empirically validate our theory, showing that it captures the behavior of deep neural networks (convolutional networks, residual networks, and Vision Transformers) trained on a set of compositional tasks with similarly structured data. Ultimately, this work examines how statistical structure in the training data can affect compositional generalization, with implications for how to identify and remedy failure modes in deep learning models.
In this paper, we elucidate how representations in deep neural networks (DNNs) evolve during training. Our focus is on overparameterized learning settings where the training continues much after the trained DNN starts to perfectly fit its training data. We examine the evolution of learned representations along the entire training process. We explore the representational similarity of DNN layers, each layer with respect to its own representations throughout the training process. For this, we use two similarity metrics: (1) The centered kernel alignment (CKA) similarity; (2) Similarity of decision regions of linear classifier probes that we train for the DNN layers. We visualize and analyze the decision regions of the DNN output and the layer probes during the DNN training to show how they geometrically evolve. Our extensive experiments discover training dynamics patterns that can emerge in layers depending on the relative layer-depth, architecture and optimizer. Among our findings: (i) The training phases, including those related to memorization, are more distinguishable in SGD training than in Adam training, and for Vision Transformer (ViT) than for ResNet; (ii) Unlike ResNet, the ViT layers have synchronized dynamics of representation learning.
The dynamics of information diffusion within graphs is a critical open issue that heavily influences graph representation learning, especially when considering long-range propagation. This calls for principled approaches that control and regulate the degree of propagation and dissipation of information throughout the neural flow. Motivated by this, we introduce (port-)Hamiltonian Deep Graph Networks, a novel framework that models neural information flow in graphs by building on the laws of conservation of Hamiltonian dynamical systems. We reconcile under a single theoretical and practical framework both non-dissipative long-range propagation and non-conservative behaviors, introducing tools from mechanical systems to gauge the equilibrium between the two components. Our approach can be applied to general message-passing architectures, and it provides theoretical guarantees on information conservation in time. Empirical results prove the effectiveness of our port-Hamiltonian scheme in pushing simple graph convolutional architectures to state-of-the-art performance in long-range benchmarks.
Machine Learning and AI have the potential to transform data-driven scientific discovery, enabling accurate predictions for several scientific phenomena. As many scientific questions are inherently causal, this paper looks at the causal inference task of treatment effect estimation, where the outcome of interest is recorded in high-dimensional observations in a Randomized Controlled Trial (RCT). Despite being the simplest possible causal setting and a perfect fit for deep learning, we theoretically find that many common choices in the literature may lead to biased estimates. To test the practical impact of these considerations, we recorded ISTAnt, the first real-world benchmark for causal inference downstream tasks on high-dimensional observations as an RCT studying how garden ants (Lasius neglectus) respond to microparticles applied onto their colony members by hygienic grooming. Comparing 6 480 models fine-tuned from state-of-the-art visual backbones, we find that the sampling and modeling choices significantly affect the accuracy of the causal estimate, and that classification accuracy is not a proxy thereof. We further validated the analysis, repeating it on a synthetically generated visual data set controlling the causal model. Our results suggest that future benchmarks should carefully consider real downstream scientific questions, especially causal ones. Further, we highlight guidelines for representation learning methods to help answer causal questions in the sciences.
Online service platforms offering a wide range of services through miniapps have become crucial for users who visit these platforms with clear intentions to find services they are interested in. Aiming at effective content delivery, cross-domain recommendation are introduced to learn high-quality representations by transferring behaviors from data-rich scenarios. However, these methods overlook the impact of the decision path that users take when conduct behaviors, that is, users ultimately exhibit different behaviors based on various intents. To this end, we propose HIER, a novel Hierarchical decIsion path Enhanced Representation learning for cross-domain recommendation. With the help of graph neural networks for high-order topological information of the knowledge graph between multi-source behaviors, we further adaptively learn decision paths through well-designed exemplar-level and information bottleneck based contrastive learning. Extensive experiments in online and offline environments show the superiority of HIER.
Diffusion-based representation learning has achieved substantial attention due to its promising capabilities in latent representation and sample generation. Recent studies have employed an auxiliary encoder to identify a corresponding representation from a sample and to adjust the dimensionality of a latent variable z. Meanwhile, this auxiliary structure invokes information split problem because the diffusion and the auxiliary encoder would divide the information from the sample into two representations for each model. Particularly, the information modeled by the diffusion becomes over-regularized because of the static prior distribution on xT. To address this problem, we introduce Diffusion Bridge AuteEncoders (DBAE), which enable z-dependent endpoint xT inference through a feed-forward architecture. This structure creates an information bottleneck at z, so xT becomes dependent on z in its generation. This results in two consequences: 1) z holds the full information of samples, and 2) xT becomes a learnable distribution, not static any further. We propose an objective function for DBAE to enable both reconstruction and generative modeling, with their theoretical justification. Empirical evidence supports the effectiveness of the intended design in DBAE, which notably enhances downstream inference quality, reconstruction, and disentanglement. Additionally, DBAE generates high-fidelity samples in the unconditional generation.
Vision transformers have established a precedent of patchifying images into uniformly-sized chunks before processing. We hypothesize that this design choice may limit models in learning comprehensive and compositional representations from visual data. This paper explores the notion of providing semantically-meaningful visual tokens to transformer encoders within a vision-language pre-training framework. Leveraging off-the-shelf segmentation and scene-graph models, we extract representations of instance segmentation masks (referred to as tangible tokens) and relationships and actions (referred to as intangible tokens). Subsequently, we pre-train a vision-side transformer by incorporating these newly extracted tokens and aligning the resultant embeddings with caption embeddings from a text-side encoder. To capture the structural and semantic relationships among visual tokens, we introduce additive attention weights, which are used to compute self-attention scores. Our experiments on COCO demonstrate notable improvements over ViTs in learned representation quality across text-to-image (+47%) and image-to-text retrieval (+44%) tasks. Furthermore, we showcase the advantages on compositionality benchmarks such as ARO (+18%) and Winoground (+10%).
Daily monitoring of intra-personal facial changes associated with health and emotional conditions has great potential to be useful for medical, healthcare, and emotion recognition fields. However, the approach for capturing intra-personal facial changes is relatively unexplored due to the difficulty of collecting temporally changing face images. In this paper, we propose a facial representation learning method using synthetic images for comparing faces, called ComFace, which is designed to capture intra-personal facial changes. For effective representation learning, ComFace aims to acquire two feature representations, i.e., inter-personal facial differences and intra-personal facial changes. The key point of our method is the use of synthetic face images to overcome the limitations of collecting real intra-personal face images. Facial representations learned by ComFace are transferred to three extensive downstream tasks for comparing faces: estimating facial expression changes, weight changes, and age changes from two face images of the same individual. Our ComFace, trained using only synthetic data, achieves comparable to or better transfer performance than general pre-training and state-of-the-art representation learning methods trained using real images.
The vulnerability of neural network classifiers to adversarial attacks is a major obstacle to their deployment in safety-critical applications. Regularization of network parameters during training can be used to improve adversarial robustness and generalization performance. Usually, the network is regularized end-to-end, with parameters at all layers affected by regularization. However, in settings where learning representations is key, such as self-supervised learning (SSL), layers after the feature representation will be discarded when performing inference. For these models, regularizing up to the feature space is more suitable. To this end, we propose a new spectral regularizer for representation learning that encourages black-box adversarial robustness in downstream classification tasks. In supervised classification settings, we show empirically that this method is more effective in boosting test accuracy and robustness than previously-proposed methods that regularize all layers of the network. We then show that this method improves the adversarial robustness of classifiers using representations learned with self-supervised training or transferred from another classification task. In all, our work begins to unveil how representational structure affects adversarial robustness.
Remote sensing data provide a low-cost solution for large-scale monitoring of air pollution via the retrieval of aerosol optical depth (AOD), but is often limited by cloud contamination. Existing methods for AOD reconstruction rely on temporal information. However, for remote sensing data at high spatial resolution, multi-temporal observations are often unavailable. In this letter, we take advantage of deep representation learning from convolutional neural networks and propose Deep Feature Gaussian Processes (DFGP) for single-scene AOD reconstruction. By using deep learning, we transform the variables to a feature space with better explainable power. By using Gaussian processes, we explicitly consider the correlation between observed AOD and missing AOD in spatial and feature domains. Experiments on two AOD datasets with real-world cloud patterns showed that the proposed method outperformed deep CNN and random forest, achieving R$^2$ of 0.7431 on MODIS AOD and R$^2$ of 0.9211 on EMIT AOD, compared to deep CNN's R$^2$ of 0.6507 and R$^2$ of 0.8619. The proposed methods increased R$^2$ by over 0.35 compared to the popular random forest in AOD reconstruction. The data and code used in this study are available at \url{https://skrisliu.com/dfgp}.
Learning high-quality multi-modal entity representations is an important goal of multi-modal knowledge graph (MMKG) representation learning, which can enhance reasoning tasks within the MMKGs, such as MMKG completion (MMKGC). The main challenge is to collaboratively model the structural information concealed in massive triples and the multi-modal features of the entities. Existing methods focus on crafting elegant entity-wise multi-modal fusion strategies, yet they overlook the utilization of multi-perspective features concealed within the modalities under diverse relational contexts. To address this issue, we introduce a novel framework with Mixture of Modality Knowledge experts (MoMoK for short) to learn adaptive multi-modal entity representations for better MMKGC. We design relation-guided modality knowledge experts to acquire relation-aware modality embeddings and integrate the predictions from multi-modalities to achieve joint decisions. Additionally, we disentangle the experts by minimizing their mutual information. Experiments on four public MMKG benchmarks demonstrate the outstanding performance of MoMoK under complex scenarios.
Text-Attributed Graphs (TAGs) enhance graph structures with natural language descriptions, enabling detailed representation of data and their relationships across a broad spectrum of real-world scenarios. Despite the potential for deeper insights, existing TAG representation learning primarily relies on supervised methods, necessitating extensive labeled data and limiting applicability across diverse contexts. This paper introduces a new self-supervised learning framework, Text-And-Graph Multi-View Alignment (TAGA), which overcomes these constraints by integrating TAGs' structural and semantic dimensions. TAGA constructs two complementary views: Text-of-Graph view, which organizes node texts into structured documents based on graph topology, and the Graph-of-Text view, which converts textual nodes and connections into graph data. By aligning representations from both views, TAGA captures joint textual and structural information. In addition, a novel structure-preserving random walk algorithm is proposed for efficient training on large-sized TAGs. Our framework demonstrates strong performance in zero-shot and few-shot scenarios across eight real-world datasets.
Multivariate time series (MTS) data, when sampled irregularly and asynchronously, often present extensive missing values. Conventional methodologies for MTS analysis tend to rely on temporal embeddings based on timestamps that necessitate subsequent imputations, yet these imputed values frequently deviate substantially from their actual counterparts, thereby compromising prediction accuracy. Furthermore, these methods typically fail to provide robust initial embeddings for values infrequently observed or even absent within the training set, posing significant challenges to model generalizability. In response to these challenges, we propose SCAlable Numerical Embedding (SCANE), a novel framework that treats each feature value as an independent token, effectively bypassing the need for imputation. SCANE regularizes the traits of distinct feature embeddings and enhances representational learning through a scalable embedding mechanism. Coupling SCANE with the Transformer Encoder architecture, we develop the Scalable nUMerical eMbeddIng Transformer (SUMMIT), which is engineered to deliver precise predictive outputs for MTS characterized by prevalent missing entries. Our experimental validation, conducted across three disparate electronic health record (EHR) datasets marked by elevated missing value frequencies, confirms the superior performance of SUMMIT over contemporary state-of-the-art approaches addressing similar challenges. These results substantiate the efficacy of SCANE and SUMMIT, underscoring their potential applicability across a broad spectrum of MTS data analytical tasks.