Designing novel functional proteins crucially depends on accurately modeling their fitness landscape. Given the limited availability of functional annotations from wet-lab experiments, previous methods have primarily relied on self-supervised models trained on vast, unlabeled protein sequence or structure datasets. While initial protein representation learning studies solely focused on either sequence or structural features, recent hybrid architectures have sought to merge these modalities to harness their respective strengths. However, these sequence-structure models have so far achieved only incremental improvements when compared to the leading sequence-only approaches, highlighting unresolved challenges effectively leveraging these modalities together. Moreover, the function of certain proteins is highly dependent on the granular aspects of their surface topology, which have been overlooked by prior models. To address these limitations, we introduce the Sequence-Structure-Surface Fitness (S3F) model - a novel multimodal representation learning framework that integrates protein features across several scales. Our approach combines sequence representations from a protein language model with Geometric Vector Perceptron networks encoding protein backbone and detailed surface topology. The proposed method achieves state-of-the-art fitness prediction on the ProteinGym benchmark encompassing 217 substitution deep mutational scanning assays, and provides insights into the determinants of protein function. Our code is at https://github.com/DeepGraphLearning/S3F.
Irregularly sampled multivariate time series (ISMTS) are prevalent in reality. Most existing methods treat ISMTS as synchronized regularly sampled time series with missing values, neglecting that the irregularities are primarily attributed to variations in sampling rates. In this paper, we introduce a novel perspective that irregularity is essentially relative in some senses. With sampling rates artificially determined from low to high, an irregularly sampled time series can be transformed into a hierarchical set of relatively regular time series from coarse to fine. We observe that additional coarse-grained relatively regular series not only mitigate the irregularly sampled challenges to some extent but also incorporate broad-view temporal information, thereby serving as a valuable asset for representation learning. Therefore, following the philosophy of learning that Seeing the big picture first, then delving into the details, we present the Multi-Scale and Multi-Correlation Attention Network (MuSiCNet) combining multiple scales to iteratively refine the ISMTS representation. Specifically, within each scale, we explore time attention and frequency correlation matrices to aggregate intra- and inter-series information, naturally enhancing the representation quality with richer and more intrinsic details. While across adjacent scales, we employ a representation rectification method containing contrastive learning and reconstruction results adjustment to further improve representation consistency. MuSiCNet is an ISMTS analysis framework that competitive with SOTA in three mainstream tasks consistently, including classification, interpolation, and forecasting.
Early detection of surgical complications allows for timely therapy and proactive risk mitigation. Machine learning (ML) can be leveraged to identify and predict patient risks for postoperative complications. We developed and validated the effectiveness of predicting postoperative complications using a novel surgical Variational Autoencoder (surgVAE) that uncovers intrinsic patterns via cross-task and cross-cohort presentation learning. This retrospective cohort study used data from the electronic health records of adult surgical patients over four years (2018 - 2021). Six key postoperative complications for cardiac surgery were assessed: acute kidney injury, atrial fibrillation, cardiac arrest, deep vein thrombosis or pulmonary embolism, blood transfusion, and other intraoperative cardiac events. We compared prediction performances of surgVAE against widely-used ML models and advanced representation learning and generative models under 5-fold cross-validation. 89,246 surgeries (49% male, median (IQR) age: 57 (45-69)) were included, with 6,502 in the targeted cardiac surgery cohort (61% male, median (IQR) age: 60 (53-70)). surgVAE demonstrated superior performance over existing ML solutions across all postoperative complications of cardiac surgery patients, achieving macro-averaged AUPRC of 0.409 and macro-averaged AUROC of 0.831, which were 3.4% and 3.7% higher, respectively, than the best alternative method (by AUPRC scores). Model interpretation using Integrated Gradients highlighted key risk factors based on preoperative variable importance. surgVAE showed excellent discriminatory performance for predicting postoperative complications and addressing the challenges of data complexity, small cohort sizes, and low-frequency positive events. surgVAE enables data-driven predictions of patient risks and prognosis while enhancing the interpretability of patient risk profiles.
Driven by recent breakthrough advances in neural representation learning, approximate near-neighbor (ANN) search over vector embeddings has emerged as a critical computational workload. With the introduction of the seminal Hierarchical Navigable Small World (HNSW) algorithm, graph-based indexes have established themselves as the overwhelmingly dominant paradigm for efficient and scalable ANN search. As the name suggests, HNSW searches a layered hierarchical graph to quickly identify neighborhoods of similar points to a given query vector. But is this hierarchy even necessary? A rigorous experimental analysis to answer this question would provide valuable insights into the nature of algorithm design for ANN search and motivate directions for future work in this increasingly crucial domain. To that end, we conduct an extensive benchmarking study covering more large-scale datasets than prior investigations of this question. We ultimately find that a flat navigable small world graph graph retains all of the benefits of HNSW on high-dimensional datasets, with latency and recall performance essentially \emph{identical} to the original algorithm but with less memory overhead. Furthermore, we go a step further and study \emph{why} the hierarchy of HNSW provides no benefit in high dimensions, hypothesizing that navigable small world graphs contain a well-connected, frequently traversed ``highway" of hub nodes that maintain the same purported function as the hierarchical layers. We present compelling empirical evidence that the \emph{Hub Highway Hypothesis} holds for real datasets and investigate the mechanisms by which the highway forms. The implications of this hypothesis may also provide future research directions in developing enhancements to graph-based ANN search.
Several self-supervised learning (SSL) approaches have shown that redundancy reduction in the feature embedding space is an effective tool for representation learning. However, these methods consider a narrow notion of redundancy, focusing on pairwise correlations between features. To address this limitation, we formalize the notion of embedding space redundancy and introduce redundancy measures that capture more complex, higher-order dependencies. We mathematically analyze the relationships between these metrics, and empirically measure these redundancies in the embedding spaces of common SSL methods. Based on our findings, we propose Self Supervised Learning with Predictability Minimization (SSLPM) as a method for reducing redundancy in the embedding space. SSLPM combines an encoder network with a predictor engaging in a competitive game of reducing and exploiting dependencies respectively. We demonstrate that SSLPM is competitive with state-of-the-art methods and find that the best performing SSL methods exhibit low embedding space redundancy, suggesting that even methods without explicit redundancy reduction mechanisms perform redundancy reduction implicitly.
Most real-world datasets consist of a natural hierarchy between classes or an inherent label structure that is either already available or can be constructed cheaply. However, most existing representation learning methods ignore this hierarchy, treating labels as permutation invariant. Recent work [Zeng et al., 2022] proposes using this structured information explicitly, but the use of Euclidean distance may distort the underlying semantic context [Chen et al., 2013]. In this work, motivated by the advantage of hyperbolic spaces in modeling hierarchical relationships, we propose a novel approach HypStructure: a Hyperbolic Structured regularization approach to accurately embed the label hierarchy into the learned representations. HypStructure is a simple-yet-effective regularizer that consists of a hyperbolic tree-based representation loss along with a centering loss, and can be combined with any standard task loss to learn hierarchy-informed features. Extensive experiments on several large-scale vision benchmarks demonstrate the efficacy of HypStructure in reducing distortion and boosting generalization performance especially under low dimensional scenarios. For a better understanding of structured representation, we perform eigenvalue analysis that links the representation geometry to improved Out-of-Distribution (OOD) detection performance seen empirically. The code is available at \url{https://github.com/uiuctml/HypStructure}.
We introduce a novel framework for representation learning in head pose estimation (HPE). Previously such a scheme was difficult due to head pose data sparsity, making triplet sampling infeasible. Recent progress in 3D generative adversarial networks (3D-aware GAN) has opened the door for easily sampling triplets (anchor, positive, negative). We perform contrastive learning on extensively augmented data including geometric transformations and demonstrate that contrastive learning allows networks to learn genuine features that contribute to accurate HPE. On the other hand, we observe that existing HPE works struggle to predict head poses as accurately when test image rotation matrices are slightly out of the training dataset distribution. Experiments show that our methodology performs on par with state-of-the-art models on standard test datasets and outperforms them when images are slightly rotated/ flipped or full range head pose. To the best of our knowledge, we are the first to deliver a true full range HPE model capable of accurately predicting any head pose including upside-down pose. Furthermore, we compared with other existing full-yaw range models and demonstrated superior results.
Fine-tuning pre-trained models has become invaluable in computer vision and robotics. Recent fine-tuning approaches focus on improving efficiency rather than accuracy by using a mixture of smaller learning rates or frozen backbones. To return the spotlight to model accuracy, we present PROFIT (Proximally Restricted Optimizer For Iterative Training), one of the first optimizers specifically designed for incrementally fine-tuning converged models on new tasks or datasets. Unlike traditional optimizers such as SGD or Adam, which make minimal assumptions due to random initialization, PROFIT leverages the structure of a converged model to regularize the optimization process, leading to improved results. By employing a simple temporal gradient orthogonalization process, PROFIT outperforms traditional fine-tuning methods across various tasks: image classification, representation learning, and large-scale motion prediction. Moreover, PROFIT is encapsulated within the optimizer logic, making it easily integrated into any training pipeline with minimal engineering effort. A new class of fine-tuning optimizers like PROFIT can drive advancements as fine-tuning and incremental training become increasingly prevalent, reducing reliance on costly model training from scratch.
Self Supervised learning (SSL) has demonstrated its effectiveness in feature learning from unlabeled data. Regarding this success, there have been some arguments on the role that mutual information plays within the SSL framework. Some works argued for increasing mutual information between representation of augmented views. Others suggest decreasing mutual information between them, while increasing task-relevant information. We ponder upon this debate and propose to revisit the core idea of SSL within the framework of partial information decomposition (PID). Thus, with SSL under PID we propose to replace traditional mutual information with the more general concept of joint mutual information to resolve the argument. Our investigation on instantiation of SSL within the PID framework leads to upgrading the existing pipelines by considering the components of the PID in the SSL models for improved representation learning. Accordingly we propose a general pipeline that can be applied to improve existing baselines. Our pipeline focuses on extracting the unique information component under the PID to build upon lower level supervision for generic feature learning and on developing higher-level supervisory signals for task-related feature learning. In essence, this could be interpreted as a joint utilization of local and global clustering. Experiments on four baselines and four datasets show the effectiveness and generality of our approach in improving existing SSL frameworks.
The success of self-supervised learning (SSL) has been the focus of multiple recent theoretical and empirical studies, including the role of data augmentation (in feature decoupling) as well as complete and dimensional representation collapse. While complete collapse is well-studied and addressed, dimensional collapse has only gain attention and addressed in recent years mostly using variants of redundancy reduction (aka whitening) techniques. In this paper, we further explore a complementary approach to whitening via feature decoupling for improved representation learning while avoiding representation collapse. In particular, we perform feature decoupling by early promotion of useful features via careful feature coloring. The coloring technique is developed based on a Bayesian prior of the augmented data, which is inherently encoded for feature decoupling. We show that our proposed framework is complementary to the state-of-the-art techniques, while outperforming both contrastive and recent non-contrastive methods. We also study the different effects of coloring approach to formulate it as a general complementary technique along with other baselines.
We present a novel representation learning method for downstream tasks like anomaly detection, unsupervised classification, and similarity searches in high-energy data sets. This enabled the discovery of a new extragalactic fast X-ray transient (FXT) in Chandra archival data, XRT 200515, a needle-in-the-haystack event and the first Chandra FXT of its kind. Recent serendipitous discoveries in X-ray astronomy, including FXTs from binary neutron star mergers and an extragalactic planetary transit candidate, highlight the need for systematic transient searches in X-ray archives. We introduce new event file representations, E-t maps and E-t-dt cubes, that effectively encode both temporal and spectral information, enabling the seamless application of machine learning to variable-length event file time series. Our unsupervised learning approach employs PCA or sparse autoencoders to extract low-dimensional, informative features from these data representations, followed by clustering in the embedding space with DBSCAN. New transients are identified within transient-dominant clusters or through nearest-neighbour searches around known transients, producing a catalogue of 3559 candidates (3447 flares and 112 dips). XRT 200515 exhibits unique temporal and spectral variability, including an intense, hard <10s initial burst, followed by spectral softening in an ~800s oscillating tail. We interpret XRT 200515 as either the first giant magnetar flare observed at low X-ray energies or the first extragalactic Type I X-ray burst from a faint, previously unknown low-mass X-ray binary in the LMC. Our method extends to data sets from other observatories such as XMM-Newton, Swift-XRT, eROSITA, Einstein Probe, and upcoming missions like AXIS.
Recent advancements in Spectral Graph Convolutional Networks (SpecGCNs) have led to state-of-the-art performance in various graph representation learning tasks. To exploit the potential of SpecGCNs, we analyze corresponding graph filters via polynomial interpolation, the cornerstone of graph signal processing. Different polynomial bases, such as Bernstein, Chebyshev, and monomial basis, have various convergence rates that will affect the error in polynomial interpolation. Although adopting Chebyshev basis for interpolation can minimize maximum error, the performance of ChebNet is still weaker than GPR-GNN and BernNet. \textbf{We point out it is caused by the Gibbs phenomenon, which occurs when the graph frequency response function approximates the target function.} It reduces the approximation ability of a truncated polynomial interpolation. In order to mitigate the Gibbs phenomenon, we propose to add the Gibbs damping factor with each term of Chebyshev polynomials on ChebNet. As a result, our lightweight approach leads to a significant performance boost. Afterwards, we reorganize ChebNet via decoupling feature propagation and transformation. We name this variant as \textbf{ChebGibbsNet}. Our experiments indicate that ChebGibbsNet is superior to other advanced SpecGCNs, such as GPR-GNN and BernNet, in both homogeneous graphs and heterogeneous graphs.
Graph Transformers (GTs) have demonstrated significant advantages in graph representation learning through their global attention mechanisms. However, the self-attention mechanism in GTs tends to neglect the inductive biases inherent in graph structures, making it chanllenging to effectively capture essential structural information. To address this issue, we propose a novel approach that integrate graph inductive bias into self-attention mechanisms by leveraging quantum technology for structural encoding. In this paper, we introduce the Graph Quantum Walk Transformer (GQWformer), a groundbreaking GNN framework that utilizes quantum walks on attributed graphs to generate node quantum states. These quantum states encapsulate rich structural attributes and serve as inductive biases for the transformer, thereby enabling the generation of more meaningful attention scores. By subsequently incorporating a recurrent neural network, our design amplifies the model's ability to focus on both local and global information. We conducted comprehensive experiments across five publicly available datasets to evaluate the effectiveness of our model. These results clearly indicate that GQWformer outperforms existing state-of-the-art graph classification algorithms. These findings highlight the significant potential of integrating quantum computing methodologies with traditional GNNs to advance the field of graph representation learning, providing a promising direction for future research and applications.
Subgraph representation learning has been effective in solving various real-world problems. However, current graph neural networks (GNNs) produce suboptimal results for subgraph-level tasks due to their inability to capture complex interactions within and between subgraphs. To provide a more expressive and efficient alternative, we propose WLKS, a Weisfeiler-Lehman (WL) kernel generalized for subgraphs by applying the WL algorithm on induced $k$-hop neighborhoods. We combine kernels across different $k$-hop levels to capture richer structural information that is not fully encoded in existing models. Our approach can balance expressiveness and efficiency by eliminating the need for neighborhood sampling. In experiments on eight real-world and synthetic benchmarks, WLKS significantly outperforms leading approaches on five datasets while reducing training time, ranging from 0.01x to 0.25x compared to the state-of-the-art.
One-shot voice conversion (VC) aims to alter the timbre of speech from a source speaker to match that of a target speaker using just a single reference speech from the target, while preserving the semantic content of the original source speech. Despite advancements in one-shot VC, its effectiveness decreases in real-world scenarios where reference speeches, often sourced from the internet, contain various disturbances like background noise. To address this issue, we introduce Noro, a Noise Robust One-shot VC system. Noro features innovative components tailored for VC using noisy reference speeches, including a dual-branch reference encoding module and a noise-agnostic contrastive speaker loss. Experimental results demonstrate that Noro outperforms our baseline system in both clean and noisy scenarios, highlighting its efficacy for real-world applications. Additionally, we investigate the hidden speaker representation capabilities of our baseline system by repurposing its reference encoder as a speaker encoder. The results shows that it is competitive with several advanced self-supervised learning models for speaker representation under the SUPERB settings, highlighting the potential for advancing speaker representation learning through one-shot VC task.
Humans recognize objects after observing only a few examples, a remarkable capability enabled by their inherent language understanding of the real-world environment. Developing verbalized and interpretable representation can significantly improve model generalization in low-data settings. In this work, we propose Verbalized Representation Learning (VRL), a novel approach for automatically extracting human-interpretable features for object recognition using few-shot data. Our method uniquely captures inter-class differences and intra-class commonalities in the form of natural language by employing a Vision-Language Model (VLM) to identify key discriminative features between different classes and shared characteristics within the same class. These verbalized features are then mapped to numeric vectors through the VLM. The resulting feature vectors can be further utilized to train and infer with downstream classifiers. Experimental results show that, at the same model scale, VRL achieves a 24% absolute improvement over prior state-of-the-art methods while using 95% less data and a smaller mode. Furthermore, compared to human-labeled attributes, the features learned by VRL exhibit a 20% absolute gain when used for downstream classification tasks. Code is available at: https://github.com/joeyy5588/VRL/tree/main.
Self-supervised learning (SSL) has achieved great success in speech-related tasks. While Transformer and Conformer architectures have dominated SSL backbones, encoders like Zipformer, which excel in automatic speech recognition (ASR), remain unexplored in SSL. Concurrently, inefficiencies in data processing within existing SSL training frameworks, such as fairseq, pose challenges in managing the growing volumes of training data. To address these issues, we propose k2SSL, an open-source framework that offers faster, more memory-efficient, and better-performing self-supervised speech representation learning, focusing on downstream ASR tasks. The optimized HuBERT and proposed Zipformer-based SSL systems exhibit substantial reductions in both training time and memory usage during SSL training. Experiments on LibriSpeech demonstrate that Zipformer Base significantly outperforms HuBERT and WavLM, achieving up to a 34.8% relative WER reduction compared to HuBERT Base after fine-tuning, along with a 3.5x pre-training speedup in GPU hours. When scaled to 60k hours of LibriLight data, Zipformer Large exhibits remarkable efficiency, matching HuBERT Large's performance while requiring only 5/8 pre-training steps.
In recent years, graph representation learning has undergone a paradigm shift, driven by the emergence and proliferation of graph neural networks (GNNs) and their heterogeneous counterparts. Heterogeneous GNNs have shown remarkable success in extracting low-dimensional embeddings from complex graphs that encompass diverse entity types and relationships. While meta-path-based techniques have long been recognized for their ability to capture semantic affinities among nodes, their dependence on manual specification poses a significant limitation. In contrast, matrix-focused methods accelerate processing by utilizing structural cues but often overlook contextual richness. In this paper, we challenge the current paradigm by introducing ontology as a fundamental semantic primitive within complex graphs. Our goal is to integrate the strengths of both matrix-centric and meta-path-based approaches into a unified framework. We propose perturbation Ontology-based Graph Attention Networks (POGAT), a novel methodology that combines ontology subgraphs with an advanced self-supervised learning paradigm to achieve a deep contextual understanding. The core innovation of POGAT lies in our enhanced homogeneous perturbing scheme designed to generate rigorous negative samples, encouraging the model to explore minimal contextual features more thoroughly. Through extensive empirical evaluations, we demonstrate that POGAT significantly outperforms state-of-the-art baselines, achieving a groundbreaking improvement of up to 10.78\% in F1-score for the critical task of link prediction and 12.01\% in Micro-F1 for the critical task of node classification.
The remarkable progress in deep learning (DL) showcases outstanding results in various computer vision tasks. However, adaptation to real-time variations in data distributions remains an important challenge. Test-Time Training (TTT) was proposed as an effective solution to this issue, which increases the generalization ability of trained models by adding an auxiliary task at train time and then using its loss at test time to adapt the model. Inspired by the recent achievements of contrastive representation learning in unsupervised tasks, we propose ReC-TTT, a test-time training technique that can adapt a DL model to new unseen domains by generating discriminative views of the input data. ReC-TTT uses cross-reconstruction as an auxiliary task between a frozen encoder and two trainable encoders, taking advantage of a single shared decoder. This enables, at test time, to adapt the encoders to extract features that will be correctly reconstructed by the decoder that, in this phase, is frozen on the source domain. Experimental results show that ReC-TTT achieves better results than other state-of-the-art techniques in most domain shift classification challenges.
Developing effective path representations has become increasingly essential across various fields within intelligent transportation. Although pre-trained path representation learning models have shown improved performance, they predominantly focus on the topological structures from single modality data, i.e., road networks, overlooking the geometric and contextual features associated with path-related images, e.g., remote sensing images. Similar to human understanding, integrating information from multiple modalities can provide a more comprehensive view, enhancing both representation accuracy and generalization. However, variations in information granularity impede the semantic alignment of road network-based paths (road paths) and image-based paths (image paths), while the heterogeneity of multi-modal data poses substantial challenges for effective fusion and utilization. In this paper, we propose a novel Multi-modal, Multi-granularity Path Representation Learning Framework (MM-Path), which can learn a generic path representation by integrating modalities from both road paths and image paths. To enhance the alignment of multi-modal data, we develop a multi-granularity alignment strategy that systematically associates nodes, road sub-paths, and road paths with their corresponding image patches, ensuring the synchronization of both detailed local information and broader global contexts. To address the heterogeneity of multi-modal data effectively, we introduce a graph-based cross-modal residual fusion component designed to comprehensively fuse information across different modalities and granularities. Finally, we conduct extensive experiments on two large-scale real-world datasets under two downstream tasks, validating the effectiveness of the proposed MM-Path. The code is available at: https://github.com/decisionintelligence/MM-Path.