Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Steven C. H. HoiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works297 from public data
- Deep Convolutional Neural Network Based Regression Approach for Estimation of Remaining Useful Life968
A novel deep Convolutional Neural Network (CNN) based regression approach for estimating the Remaining Useful Life (RUL) of a subsystem or a component using sensor data, which has many real world applications.
- Online learning: A comprehensive survey871
This survey aims to provide a comprehensive survey of the online machine learning literatures through a systematic review of basic ideas and key principles and a proper categorization of different algorithms and techniques.
- Rumor Detection on Social Media with Bi-Directional Graph Convolutional Networks863
A novel bi-directional graph model is proposed, named Bi-Directional Graph Convolutional Networks (Bi-GCN), to explore both characteristics of rumors by operating on both top-down and bottom-up propagation of rumors.
- Efficient Test-Time Model Adaptation without Forgetting740
An active sample selection criterion is proposed to identify reliable and non-redundant samples, on which the model is updated to minimize the entropy loss for test-time adaptation, and a Fisher regularizer is introduced to constrain important model parameters from drastic changes.
- Towards Stable Test-Time Adaptation in Dynamic Wild World617
This paper proposes a sharpness-aware and reliable entropy minimization method, called SAR, for further stabilizing TTA from two aspects: 1) remove partial noisy samples with large gradients, 2) encourage model weights to go to a flat minimum so that the model is robust to the remaining noisy samples.
- Graph Convolutional Networks for Temporal Action Localization572
This paper builds an action proposal graph, where each proposal is represented as a node and their relations between two proposals as an edge and applies the GCNs over the graph to model the relations among different proposals and learn powerful representations for the action classification and localization.
- 472
- Mastering Complex Control in MOBA Games with Deep Reinforcement Learning432
A deep reinforcement learning framework to tackle the problem of complex action control in the Multi-player Online Battle Arena (MOBA) 1v1 games is presented, which is of low coupling and high scalability, which enables efficient explorations at large scale.
- Stochastic Optimization with Importance Sampling for Regularized Loss Minimization393
Stochastic optimization with importance sampling is studied, which improves the convergence rate by reducing the stochastic variance and the convergence rates with the proposed importance sampling methods can be significantly improved under suitable conditions both for prox-SGD and for proxies-SDCA.
- Neighborhood Regularized Logistic Matrix Factorization for Drug-Target Interaction Prediction391
The proposed NRLMF method focuses on modeling the probability that a drug would interact with a target by logistic matrix factorization, where the properties of drugs and targets are represented by drug-specific and target-specific latent vectors, respectively.
- 328
- 322
- Scaling Diffusion Language Models via Adaptation from Autoregressive Models299
A suite of DLMs capable of generating fluent text, performing in-context learning, filling in the middle without prompt re-ordering, and following instructions is released, revealing that these models outperform earlier DLMs and are competitive with their AR counterparts.
- Online Feature Selection and Its Applications282
This article investigates the problem of online feature selection (OFS) in which an online learner is only allowed to maintain a classifier involved only a small and fixed number of features, and presents novel algorithms to solve each of the two problems.
- Adversarial Sparse Transformer for Time Series Forecasting265
Adversarial Sparse Transformer (AST) is proposed, a new time series forecasting model based on Generative Adversarial Networks (GANs), based on the ground-truth provided during training and replaced by the network’s own one-step ahead output during inference during inference.
- Hierarchical graph learning for protein–protein interaction243
HIGH-PPI can interpret the modes of action of PPIs by identifying important binding and catalytic sites precisely and is a domain-knowledge-driven and interpretable framework for PPI prediction studies.
- Online AUC Maximization241
This work develops online learning algorithm for maximizing Area Under the ROC curve (AUC), a metric that is widely used for measuring the classification performance for imbalanced data distributions, and presents two algorithms for online AUC maximization with theoretic performance guarantee.
- 240
- LIBOL: a library for online learning algorithms236
LIBOL is an open-source library for large-scale online learning, which consists of a large family of efficient and scalable state-of-the-art online learning algorithms for large -scale online classification tasks.
- Online ARIMA Algorithms for Time Series Prediction216
This paper proposes online learning algorithms for estimating ARIMA models under relaxed assumptions on the noise terms, which is suitable to a wider range of applications and enjoys high computational efficiency.
- Online Transfer Learning205
This work proposes a novel machine learning framework called "Online Transfer Learning" (OTL), which aims to attack an online learning task on a target domain by transferring knowledge from some source domain, and investigates two different settings of OTL: (i) OTL on homogeneous domains of common feature space, and (ii) O TL across heterogeneous domainsof different feature spaces.
- Collaborative Unsupervised Domain Adaptation for Medical Image Diagnosis195
This work innovatively proposes a Collaborative Unsupervised Domain Adaptation algorithm, which conducts transferability-aware adaptation and conquers label noise in a collaborative way and theoretically analyze the generalization performance of the proposed method, and also empirically evaluate it on both medical and general images.
- Online multimodal deep similarity learning with application to image retrieval175
This paper proposes a novel framework of online multimodal deep similarity learning (OMDSL), which aims to optimally integrate multiple deep neural networks pretrained with stacked denoising autoencoder to improve similarity search in multimedia information retrieval tasks.
- RetroXpert: Decompose Retrosynthesis Prediction like A Chemist152
A novel template-free algorithm for automatic retrosynthetic expansion inspired by how chemists approach Retrosynthesis prediction is devised, outperforming the state-of-the-art baselines by a significant margin and provides chemically reasonable interpretation.
- Margin Learning Embedded Prediction for Video Anomaly Detection with A Few Anomalies149
A Margin Learning Embedded Prediction (MLEP) framework that helps with the detection of abnormal events, even for anomalies that have never been previously observed, and pushes the application of anomaly detection towards real scenarios.
- GCN-MF146
This paper proposes a new framework for disease-gene association task by combining Graph Convolutional Network (GCN) and matrix factorization, named GCN-MF, and demonstrates that the proposed deep learning algorithm outperforms all other state-of-the-art methods on most of metrics.
- 145
- Confidence Weighted Mean Reversion Strategy for Online Portfolio Selection144
Inspired by the mean reversion principle in finance and confidence weighted online learning technique in machine learning, Confidence Weighted Mean Reversion (CWMR) models the portfolio vector as a Gaussian distribution, and sequentially updates the distribution by following themean reversion trading principle.
- Cost-sensitive online active learning with application to malicious URL detection143
An extensive set of experiments are conducted to examine the empirical performance of the proposed CSOAL algorithms for a large-scale challenging malicious URL detection task, in which encouraging results showed that the proposed technique by querying an extremely small-sized labeled data can achieve better or highly comparable classification performance in comparison to the state-of-the-art cost-insensitive and cost-sensitive online classification algorithms.
- 142
- GADBench: Revisiting and Benchmarking Supervised Graph Anomaly Detection132
The main finding is that tree ensembles with simple neighborhood aggregation can outperform the latest GNNs tailored for the GAD task, shedding light on the current progress of GAD.
- Cost-Sensitive Portfolio Selection via Deep Reinforcement Learning132
A novel two-stream portfolio policy network is devised to extract both price series patterns and asset correlations, while a new cost-sensitive reward function is developed to maximize the accumulated return and constrain both costs via reinforcement learning.
- 132
- 131
- Self-Introspective Decoding: Alleviating Hallucinations for Large Vision-Language Models118
This work develops the Context and Text-aware Token Selection (CT2S) strategy, which preserves only unimportant vision tokens after early layers of LVLMs to adaptively amplify text-informed hallucination during the auto-regressive decoding.
- OTL: A Framework of Online Transfer Learning114
This paper investigates a new machine learning framework called Online Transfer Learning (OTL) that aims to transfer knowledge from some source domain to an online learning task on a target domain, and proposes techniques to address two kinds of OTL tasks.
- 112
- Cost-Sensitive Online Classification110
This paper proposes two novel cost-sensitive online classification algorithms, which are designed to directly optimize two well-known cost-sensitive measures: maximization of weighted sum of sensitivity and specificity, and minimization of weighted misclassification cost.
- Online feature selection for mining big data107
This work investigates the problem of Online Feature Selection (OFS) in which the online learner is only allowed to maintain a classifier involved a small and fixed number of features, and presents an effective algorithm to solve the problem, and gives the theoretical analysis and empirical performance of the proposed algorithms.
- Test-Time Model Adaptation with Only Forward Passes101
Without using any backpropagation and altering model weights, FOA runs on quantized 8-bit ViT outperforms gradient-based TENT on full-precision 32-bit ViT, while achieving an up to 24-fold memory reduction on ImageNet-C.
- Online Multiple Kernel Similarity Learning for Visual Search100
This paper investigates an online kernel similarity learning framework for learning kernel-based proximity functions which goes beyond the conventional linear distance metric learning approaches and proposes a novel online multiple kernel similarity (OMKS) learning method which learns a flexible nonlinear proximity function with multiple kernels to improve visual similarity search in CBIR.
- Online Multiple Kernel Classification99
This work addresses a new research problem, termed Online Multiple Kernel Classification (OMKC), which learns a kernel-based prediction function by selecting a subset of predefined kernel functions in an online learning fashion, and develops stochastic selection strategies that randomly select a portion of kernels for combination and model updating, thus improving the learning efficiency.
- Double Updating Online Learning98
An extensive set of empirical evaluations for both binary and multi-class online learning tasks show that the proposed DUOL technique is considerably more effective than the state-of-the-art online learning algorithms.
- Repeat Buyer Prediction for E-Commerce97
This paper created profiles for users, merchants, brands, categories, items and their interactions via extensive feature engineering for repeat buyer prediction based on the sales data of the ``Double 11" shopping event in 2014 at Tmall.com.
- Towards Fast Adaptation of Neural Architectures with Meta Learning96
A novel Transferable Neural Architecture Search method based on meta-learning, which learns a meta-architecture that is able to adapt to a new task quickly through a few gradient steps, which makes the transferred architecture suitable for the specific task.
- Projection-free Distributed Online Learning in Networks96
The distributed online conditional gradient algorithm is proposed, which eschews the expensive projection operation needed in its counterpart algorithms by exploiting much simpler linear optimization steps and gives a regret bound for the proposed algorithm as a function of the network size and topology.
- Graph Convolutional Module for Temporal Action Localization in Videos93
It is claimed that the relations between action units play an important role in action localization, and a more powerful action detector should not only capture the local content of each action unit but also allow a wider field of view on the context related to it to enhance action localization.
- Relation-Aware Transformer for Portfolio Policy Learning91
Under a deep reinforcement learning paradigm for portfolio selection, a novel Relation-aware Transformer (RAT) is proposed that is structurally innovated to capture both sequential patterns and asset correlations for portfolio selection.
- Contextualized Point-of-Interest Recommendation89
This paper proposes a new framework for POI recommendation, which explicitly utilizes similarity with contextual information, and outperforms all the state-of-the-art methods.
- Divide-and-Conquer: Post-User Interaction Network for Fake News Detection on Social Media88
A fake news detection model named Post-User Interaction Network (PSIN) is proposed, which adopts a divide-and-conquer strategy to model the post-post, user-user and post-user interactions in social context effectively while maintaining their intrinsic characteristics.
- Fairness-guided Few-shot Prompting for Large Language Models87
This paper introduces a metric to evaluate the predictive bias of a fixed prompt against labels or a given attributes and proposes a novel search strategy based on the greedy search to identify the near-optimal prompt for improving the performance of in-context learning.
- Privacy Preserving Point-of-Interest Recommendation Using Decentralized Matrix Factorization86
Experimental results on two real-world datasets demonstrate that the Decentralized MF (DMF) framework for POI recommendation significantly improvements the recommendation performance in terms of precision and recall.
- Mining social images with distance metric learning for automated image tagging84
This paper proposes a machine learning framework for mining social images and investigates its application to automated image tagging, and develops an efficient stochastic gradient descent algorithm for solving the UDML optimization task and proves the convergence of the algorithm.
- 82
- DrugOOD: Out-of-Distribution Dataset Curator and Benchmark for AI-Aided Drug Discovery – a Focus on Affinity Prediction Problems with Noise Annotations81
DrugOOD offers an automated dataset curator with user-friendly customization scripts, rich domain annotations aligned with biochemistry knowledge, realistic noise level annotations, and rigorous benchmarking of SOTA OOD algorithms, as opposed to only providing fixed datasets.
- Online Multi-Modal Distance Metric Learning with Application to Image Retrieval77
A novel scheme of online multi-modal distance metric learning (OMDML) is investigated, which explores a unified two-level online learning scheme that learns to optimize a distance metric on each individual feature space; and then it learns to find the optimal combination of diverse types of features.
- Adaptive cost-sensitive online classification73
A family of cost-sensitive online classification algorithms with adaptive regularization with sketching technique is proposed, which significantly accelerates the computational speed with quite slight performance loss and applies to several online anomaly detection tasks from real world.
- Latent Diffusion Transformer for Probabilistic Time Series Forecasting72
A latent multivariate time series diffusion framework called Latent Diffusion Transformer (LDT), which consists of a symmetric statistics-aware autoencoder and a diffusion-based conditional generator, to implement high-dimensional multivariate time series forecasting into a problem of latent space time series generation.
- Detecting cyberattacks in industrial control systems using online learning algorithms72
Several state-of-the-art online learning algorithms categorically are introduced and illustrated, and a new cost-sensitive online learning algorithm is explored to solve the class-imbalance problem that is pervasive in industrial intrusion detection systems.
- Online Passive-Aggressive Active learning72
This paper proposes several variants of PAA algorithms to tackle three types of online learning tasks: binary classification, multi-class classification, and cost-sensitive classification and gives the mistake bounds of the proposed algorithms in theory, and conducts extensive experiments to evaluate the empirical performance.
- 71
- 70
- 70
- From Whole Slide Imaging to Microscopy: Deep Microscopy Adaptation Network for Histopathology Cancer Image Classification70
A novel Deep Microscopy Adaptation Network (DMAN) is proposed, which can classify MSIs effectively even without MSI annotations, by reducing domain discrepancies via adversarial learning and entropy minimization, and alleviating class imbalance with sample reweighting.
- 62
- Context-Aware Domain Adaptation in Semantic Segmentation61
This paper proposes a cross-attention mechanism based on self-att attention to capture context dependencies between two domains and adapt transferable context and designs two cross-domain attention modules to adapt context dependencies from both spatial and channel views.
- 61
- Supervised Learning Achieves Human-Level Performance in MOBA Games: A Case Study of Honor of Kings61
JueWu-SL, the first supervised-learning-based artificial intelligence program that achieves human-level performance in playing multiplayer online battle arena (MOBA) games, integrates the macro-strategy and the micromanagement of MOBA-game-playing into neural networks in a supervised and end-to-end manner.
- Deceptive Review Spam Detection via Exploiting Task Relatedness and Unlabeled Data58
This paper develops a multi-task learning method based on logistic regression (MTL-LR), which can boost the learning for a task by sharing the knowledge contained in the training signals of other related tasks, and proposes a novel semi-supervised multi- task learning method via Laplacian regularized logistics regression (SMTL-LLR) to further improve the review spam detection performance.
- Deep learning for human activity recognition56
This chapter proposes a systematic feature learning method that adopts a deep convolutional neural network to automate feature learning from the raw inputs in a systematic way and results in mutual enhancements in both feature learning and classification.
- A boosting algorithm for item recommendation with implicit feedback56
A boosting algorithm named AdaBPR (Adaptive Boosting Personalized Ranking) is proposed for top-N item recommendation using users' implicit feedback and demonstrates its effectiveness on three datasets compared with strong baseline algorithms.
- UMIX: Improving Importance Weighting for Subpopulation Shift via Uncertainty-Aware Mixup54
This work proposes a simple yet practical framework, called uncertainty-aware mixup (UMIX), to mitigate the overfitting issue in over-parameterized models by reweighting the ''mixed'' samples according to the sample uncertainty.
- 54
- Neighbour Interaction based Click-Through Rate Prediction via Graph-masked Transformer52
Neighbor-Interaction based CTR prediction (NI-CTR) is proposed, which considers this task under a Heterogeneous Information Network (HIN) setting and outperforms state-of-the-art CTR models significantly, and is deployed on the WeChat Official Account Platform with billions of users.
- Online Adaptive Asymmetric Active Learning for Budgeted Imbalanced Data52
A novel Online Adaptive Asymmetric Active (OA3) learning algorithm is proposed, based on a new asymmetric strategy (merging both the asymmetric losses and queries strategies), and second-order optimization, which is theoretically analyzed and empirically evaluated on four real-world online anomaly detection tasks.
- Online multi-task collaborative filtering for on-the-fly recommender systems49
A novel efficient and scalable online collaborative filtering technique for on-the-fly recommender systems, which is able to effectively online update the recommendation model from a sequence of rating observations is investigated.
- 48
- An Adaptive Gradient Method for Online AUC Maximization47
This paper proposes a novel algorithm of Adaptive Online AUC Maximization (AdaOAM), by applying an adaptive gradient method for exploiting the knowledge of historical gradients to perform more informative online learning.
- KunPeng46
KunPeng not only encapsulates the characteristics of data/model parallelism, load balancing, model sync-up, sparse representation, industrial fault-tolerance, etc., but also provides easy-to-use interface to empower users to focus on the core ML logics.
- Multi-Scale Attention Flow for Probabilistic Time Series Forecasting45
This work proposed a novel non-autoregressive deep learning model, called Multi-scale Attention Normalizing Flow (MANF), where it combines multi-scale attention with relative position information and the multivariate data distribution is represented by the conditioned normalizing flow.
- Towards Accurate and Compact Architectures via Neural Architecture Transformer45
A Neural Architecture Transformer (NAT) method is proposed which casts the optimization problem into a Markov Decision Process (MDP) and seeks to replace the redundant operations with more efficient operations, such as skip or null connection.
- Adapting Large Language Models for Content Moderation: Pitfalls in Data Engineering and Supervised Fine-tuning43
This paper presents a complete process, from data collection and construction to model training and overfitting elimination, for fine-tuning LLMs in vertical domain deployments, and reveals that incorporating reasoning processes during the fine-tuning of LLMs can effectively alleviate overfitting.
- COME: Test-time adaption by Conservatively Minimizing Entropy41
The Conservatively Minimize the Entropy (COME) is proposed, which is a simple drop-in replacement of traditional EM to elegantly address the limitation of overconfidence in test-time adaption methods.
- 41
- 41
- Are Large Language Models Really Robust to Word-Level Perturbations?40
This work proposes a novel rational evaluation approach that leverages pre-trained reward models as diagnostic tools to evaluate the longer conversation generated from more challenging open questions by LLMs, which it refers to as the Reward Model for Reasonable Robustness Evaluation (TREvaL).
- Second Order Online Collaborative Filtering39
Online collaborative ltering techniques for building live recommender systems where the CF model can evolve on-the-y over time are investigated, unlike the regular rst order CF algorithms.
- RetroComposer: Composing Templates for Template-Based Retrosynthesis Prediction38
This work proposes an innovative retrosynthesis prediction framework that can compose novel templates beyond training templates and can produce novel templates for 15 USPTO-50K test reactions that are not covered by training templates.
- WatME: Towards Lossless Watermarking Through Lexical Redundancy37
A novel approach leveraging linguistic prior knowledge of inherent lexical redundancy in LLM vocabularies to seamlessly integrate watermarks is introduced, which effectively prevents the unavailability of appropriate tokens and preserves the expressive power of LLMs.
- Collaborative topic regression for online recommender systems: an online and Bayesian approach37
A novel inference algorithm, called the Online Bayesian Inference algorithm for CTR model, which is efficient and scalable for learning from data streams, and jointly optimize the combined objective function of both PMF and LDA in an online learning fashion.
- 36
- Exploring Robustness of Unsupervised Domain Adaptation in Semantic Segmentation36
This work comprehensively evaluates the robustness of existing UDA methods and proposes a robust UDA approach that maximizes the agreement between clean images and their adversarial examples by a contrastive loss in the output space and is resistant to adversarial attacks.
- Hierarchical Graph Capsule Network36
This paper innovatively proposes hierarchical graph capsule network (HGCN) that can jointly learn node embeddings and extract graph hierarchies and demonstrates the effectiveness of HGCN and the contribution of each component.
- Soft Confidence-Weighted Learning34
A family of Soft Confidence-Weighted (SCW) learning algorithms for both binary classification and multiclass classification tasks is proposed, which is the first family of online classification algorithms that enjoys four salient properties simultaneously: (1) large margin training, (2) confidence weighting, (3) capability to handle nonseparable data, and (4) adaptive margin.
- A Survey of Trustworthy Graph Learning: Reliability, Explainability, and Privacy Protection33
This survey provides a comprehensive review of recent leading approaches in the TwGL field from three dimensions, namely, reliability, explainability, and privacy protection, and gives a general categorization for existing work and review typical work for each category.
- 33
- Online Adaptive Asymmetric Active Learning With Limited Budgets33
A novel Online Adaptive Asymmetric Active learning algorithm is proposed, based on a new asymmetric strategy (merging both asymmetric losses and queries strategies), and second-order optimization, which significantly accelerates the computational speed with quite slight performance degradation.
- Second-Order Online Active Learning and Its Applications32
The SOAL framework is extended by proposing the Cost-sensitive Second-order Online Active Learning algorithm named “SOAL$_{CS}$ ”, which is devised by maximizing the sum of weighted sensitivity and specificity or minimizing the cost of weighted mistakes of different classes.
- IgGM: A Generative Model for Functional Antibody and Nanobody Design31
IgGM is introduced, a generative model for the de novo design of immunoglobulins with functional specificity that has demonstrated its effectiveness in not only predicting structures but also designing novel antibodies and nanobodies.
- Disturbance-immune weight sharing for neural architecture search31
This work proposes a new disturbance-immune update strategy for model updating that constrain the training of subsequent architectures in an orthogonal space via Orthogonal gradient descent to preserve the knowledge learned by previous architectures.
- Distributed multi-task classification: a decentralized online learning approach31
Theoretical analysis indicates that the proposed algorithm achieves a O(T) regret bound when compared with the best classifier in hindsight, which is further validated by experiments on both synthetic and real-world datasets.
- Online learning to rank for content-based image retrieval30
A new framework of learning to rank for CBIR is investigated, which aims to seek the optimal combination of different retrieval schemes by learning from large-scale training data in CBIR.
- Uncertainty-Calibrated Test-Time Model Adaptation Without Forgetting29
An Efficient Anti-Forgetting Test-Time Adaptation (EATA) method which develops an active sample selection criterion to identify reliable and non-redundant samples for test-time entropy minimization and devise a min-max entropy regularization to selectively increase and decrease predicting confidence for confidence re-calibration.
- SDformer: Similarity-driven Discrete Transformer For Time Series Generation28
This paper proposes a method based on discrete token modeling technique called Similarity-driven Discrete Transformer (SDformer), which utilizes a similarity-driven vector quantization method for learning high-quality discrete token representations of time series, followed by a discrete Transformer for data distribution modeling at the token level.
- Locally Connected Deep Learning Framework for Industrial-scale Recommender Systems28
A locally connected deep learning framework for recommender systems is proposed, which reduces the complexity of deep neural network (DNN) by two to three orders of magnitude and is extended using the idea of the recently proposed Wide&Deep model.
- Learning User Dependencies for Recommendation28
This paper proposes a novel recommendation method, named probabilistic relational matrix factorization (PRMF), which can automatically learn the dependencies between users to improve recommendation accuracy and proposes an alternating optimization algorithm to solve the optimization problem of PRMF.
- 28
- Online Passive Aggressive Active Learning and Its Applications27
The proposed PAA learning algorithms not only use the misclassified instances to update the classifier, but also exploit correctly classified examples with low prediction confidence, which shows clear advantages of the algorithms over the baselines.
- NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Prediction26
This work systematically explore the use of dual-channel speech data in the context of modern large language models, and introduces a novel generative modeling paradigm, Next-Token-Pair Prediction (NTPP), to enable speaker-independent dual-channel spoken dialogue learning using decoder-only architectures for the first time.
- tFold-Ab: Fast and Accurate Antibody Structure Prediction without Sequence Homologs26
This work proposes a fully end-to-end architecture for simultaneous prediction of backbone and side-chain conformations, namely tFold-Ab, which firstly predicts monomer structures of each chain, and then refines them into heavy-light chain complex structure prediction, which enables multi-level supervision for model training.
- PSMART: Parameter Server based Multiple Additive Regression Trees System25
A Parameter Server based Multiple Additive Regression Trees system that scales MART to hundreds of billions of samples and thousands of features in production.
- 25
- MARS: a motif-based autoregressive model for retrosynthesis prediction24
A novel end-to-end graph generation model for retrosynthesis prediction, which sequentially identifies the reaction center, generates the synthons, and adds motifs to the synthons to generate reactants, which achieves lower prediction complexity than adding atoms and demonstrates superior performance than adding leaving groups.
- SVIP: Sequence VerIfication for Procedures in Videos24
A simple but effective baseline based on the transformer encoder with a novel sequence alignment loss is introduced to better characterize long-term dependency between steps, which outperforms other action recognition methods.
- 23
- Value Penalized Q-Learning for Recommender Systems23
Value Penalized Q-learning (VPQ), a novel uncertainty-based offline RL algorithm that penalizes the unstable Q-values in the regression target using uncertainty-aware weights, achieving the conservative Q-function without the need of estimating the behavior policy, suitable for RS with a large number of items.
- Neural graph personalized ranking for Top-N Recommendation23
This work proposes neural graph personalized ranking (NGPR) which directly makes use of the user–item interaction information in embedding learning by incorporating the user-item interaction graph in embeding learning.
- Nonparametric Topic Modeling with Neural Inference23
This work proposes iTM-VAE, a nonparametric topic models with Auto-Encoding Variational Bayes, where the topics are treated as trainable parameters and the document-specific topic proportions are obtained by a stick-breaking construction and introduces a hyper-prior technique to model the uncertainty of the prior parameter.
- 23
- 22
- DUOL: A Double Updating Approach for Online Learning22
It is shown that the mistake bound can be significantly improved by the proposed online learning method, termed Double Updating Online Learning, or DUOL for short, which is in general considerably more effective than the state-of-the-art online learning algorithms.
- 21
- Robust Online Multi-Task Learning with Correlative and Personalized Structures21
This paper proposes a robust online MTL framework that overcomes the restriction of task relatedness into a presumed structure via a single weight matrix by decomposing the weight matrix into two components: the first one captures the low-rank common structure among tasks via a nuclear norm and the second one identifies the personalized patterns of outlier tasks through a group lasso.
- 20
- PSSM-Distil: Protein Secondary Structure Prediction (PSSP) on Low-Quality PSSM by Knowledge Distillation with Contrastive Learning20
A novel PSSM-Distil framework for PSSP on low-quality P SSM, which not only enhances the PSSS feature at a lower level but also aligns the feature distribution at a higher level, and supports a pre-trained protein sequence language BERT model to provide auxiliary information.
- Robust Cost-Sensitive Learning for Recommendation with Implicit Feedback20
A robust framework is proposed by decomposing the U-I matrix into two components: a low-rank matrix that captures the common preference, and a sparse matrix that detects the user-specific preference of individuals that outperforms several state-of-the-art algorithms on benchmark recommendation datasets.
- Online Compact Convexified Factorization Machine20
An Online Compact Convex Factorization Machine (OCCFM) algorithm that eschews the projection operation with efficient linear optimization steps is proposed and it is proved that the developed algorithm achieves a sub-linear regret bound.
- Online Active Learning with Expert Advice20
A framework of active forecasters is proposed, which attempts to extend two fully supervised forecasters, Exponentially Weighted Average Forecaster and Greedy Forecaster, to tackle the task of online active learning (OAL) with expert advice.
- Sparse Passive-Aggressive Learning for Bounded Online Kernel Methods20
This article proposes a novel framework for bounded online kernel methods, named “Sparse Passive-Aggressive (SPA)” learning, which is able to yield a final output kernel-based hypothesis with a bounded number of support vectors, and theoretically proves that SPA achieves an optimal mistake bound in expectation, and empirically shows that it outperforms various budget online kernel learning algorithms.
- Cost sensitive online multiple kernel classification20
This paper investigates supervised machine learning techniques for mining data streams with application to online anomaly detection with a novel Cost-Sensitive Online Multiple Kernel Classi(cid:12)cation (CSOMKC) scheme and gives both theoretical and empirical analysis of the proposed algorithms.
- Decentralized Online Learning: Take Benefits from Others’ Data without Sharing Your Own to Track Global Trend19
This article studied when and why the communication can help the decentralized online learning to reduce the regret, and shows that decentralized online gradient enjoys a regret bound that suggests that people can get benefits from the randomness in the private data by exchanging private information.
- 18
- 18
- De novo design of epitope-specific antibodies via a structure-driven computational workflow17
The tFold System overcomes key limitations of existing methods by enabling rapid, high-throughput antibody discovery against user-defined epitopes, and designs epitope-specific antibodies, with their nanomolar affinities experimentally confirmed.
- Step-On-Feet Tuning: Scaling Self-Alignment of LLMs via Bootstrapping17
This pioneering exploration delves into the impact of bootstrapping self-alignment on large language models and reveals that bootstrapping self-alignment markedly surpasses the single-round approach, by guaranteeing data diversity from in-context learning.
- 17
- 17
- 16
- 16
- 15
- Data Analytics for Fog Computing by Distributed Online Learning with Asynchronous Update15
This work proposes a system in which data streams generated from distributed sources are digested almost locally, whereas a relatively small amount of distilled information is converged to a center, and a distributed machine learning algorithm using the online learning approach is devised.
- A Min-Max Optimization Framework For Online Graph Classification15
The theoretical result shows that an online algorithm learning on these selected labels can achieve comparable mistake bound with the fully-supervised online counterpart, and an aggressive algorithm is proposed, which can update the model even if no error occurs.
- Enhancing Jailbreak Attacks on LLMs via Persona Prompts14
This work proposes a genetic algorithm-based method that automatically crafts persona prompts to bypass LLM's safety mechanisms, and demonstrates synergistic effects when combined with existing attack methods, increasing success rates by 10-20%.
- 14
- Online Multitask Relative Similarity Learning14
A simple yet effective online learning algorithm for multi-task relative similarity learning and an active learning algorithm to save the labeling cost are developed.
- Online Sparse Passive Aggressive Learning with Kernels14
This paper theoretically proves that the proposed SPA algorithm achieves an optimal regret bound in expectation, and empirically shows that the new algorithm outperforms various bounded kernel-based online learning algorithms.
- 13
- 13
- Cost-Sensitive Double Updating Online Learning and Its Application to Online Anomaly Detection13
This paper proposes the Cost-Sensitive Double Updating Online Learning (CSDUOL) algorithms, which explores a recent double updating technique to tackle the online optimization task of cost-sensitive classification by maximizing the weighted sum or minimizing the weighted misclassification cost.
- 12
- Locally Linear Factorization Machines12
This work forms a joint optimization over the anchor points, local coding coordinates and FMs variables to minimize classification or regression risk and demonstrates that this approach achieves much better predictive accuracy than other competitive methods which employ LLFM with unsupervised anchor point learning and predefined local coding scheme.
- 12
- 11
- Probing the Safety Response Boundary of Large Language Models via Unsafe Decoding Path Generation11
It is argued that: even when an LLM appears to successfully block harmful queries, there may still be hidden vulnerabilities that could act as ticking time bombs and be exploited via prompt optimization such as soft prompts on images.
- 11
- A Unified Framework for Sparse Online Learning11
A unified online learning framework for the big data stream classification task is investigated and a unified Sparse Online Classification (SOC) framework is proposed, from which a second-order online learning algorithm and a cost-sensitive sparse onlinelearning algorithm are derived, which could successfully handle online anomaly detection tasks with the extremely unbalanced class distribution.
- Self-Bootstrapping for Versatile Test-Time Adaptation10
A self-bootstrapping scheme that optimizes prediction consistency between the test image and its deteriorated view is achieved through a self-bootstrapping scheme that achieves superior results across classification, segmentation, and 3D monocular detection tasks with both transformer and CNN models.
- 10
- Towards Geometric Normalization Techniques in SE(3) Equivariant Graph Neural Networks for Physical Dynamics Simulations10
A new normalization layer called GeoNorm is proposed, which can satisfy the SE(3) equivariance and simultaneously stabilize the training process and stabilizes the training dynamics of SE(3) equivariant GNNs on large systems.
- Privacy Matters: Vertical Federated Linear Contextual Bandits for Privacy Protected Recommendation10
A customized encryption scheme named orthogonal matrix-based mask mechanism (O3M), a tailored component for contextual bandits by carefully exploiting their shared structure, can ensure privacy protection while avoiding expensive conventional cryptographic techniques.
- ImDrug: A Benchmark for Deep Imbalanced Learning in AI-aided Drug Discovery10
ImDrug is introduced, a comprehensive benchmark with an open-source Python library which consists of 4 imbalance settings, 11 AI-ready datasets, 54 learning tasks and 16 baseline algorithms tailored for imbalanced learning that demonstrates that the existing algorithms fall short of solving medicinal and pharmaceutical challenges in the data imbalance scenario.
- Distributed Collaborative Hashing and Its Applications in Ant Financial10
DCH has comparable performance but has both fast convergence speed in offline model training procedure and realtime efficiency in online recommendation procedure and the encouraging performance of DCH is also shown for several real-world applications in Ant Financial.
- Online Adaptive Passive-Aggressive Methods for Non-Negative Matrix Factorization and Its Applications10
A novel family of online Adaptive Passive-Aggressive (APA) learning algorithms for NMF overcomes two critical limitations of NN-PA by exploiting second-order information to enhance PA in making more informative updates at each iteration, and achieving the parameter auto-selection by exploring the idea of online learning with expert advice in deciding the optimal combination of the key parameters in NMF.
- 10
- Learning Relative Similarity from Data Streams: Active Online Learning Approaches10
This work proposes a novel framework of active online similarity learning and proposes two new algorithms: (i)~PAAS: Passive-Aggressive Active Similarity learning; (ii)~CWAS: Confidence-Weighted Active Similarities learning, and these algorithms will prove their mistake bounds in theory.
- BDUOL: Double Updating Online Learning on a Fixed Budget10
This paper presents a new framework of budget kernel-based online learning based on a recently proposed online learning method called "Double Updating Online Learning" (DUOL), which has shown state-of-the-art performance as compared with the other traditional kernel- based online learning algorithms.
- 9
- PointNSP: Autoregressive 3D Point Cloud Generation with Next-Scale Level-of-Detail Prediction9
Inspired by the level-of-detail (LOD) principle in shape modeling, PointNSP is proposed, a coarse-to-fine generative framework that preserves global shape structure at low resolutions and progressively refines fine-grained geometry at higher scales through a next-scale prediction paradigm.
- Measuring Diversity in Synthetic Datasets9
DCScore is introduced, a novel method for measuring synthetic dataset diversity from a classification perspective that formulates diversity evaluation as a sample classification task, leveraging mutual relationships among samples.
- A Generative Foundation Model for Antibody Design9
Extensive evaluation through both in silico benchmarks and in vitro experiments demonstrates that IgGM consistently generates antibodies or nanobodies with high measured affinity, underscore IgGM’s versatility and effectiveness as a powerful tool for next-generation antibody discovery and optimization.
- 9
- 9
- 9
- Mitigating Sample Selection Bias with Robust Domain Adaption in Multimedia Recommendation9
A debiasing framework named DAMCAR is proposed, which introduces Domain Adaptation to mitigate SSB in Multimedia CAscade Recommendation systems and obtains a retrieval model that maintains stable performance during inference through a hybrid training mechanism.
- Confidence Weighted Multitask Learning9
A confidence weighted multitask learning algorithm, which maintains a Gaussian distribution over each task model to guide online learning process, which is able to achieve promising learning efficacy, while simultaneously minimizing the labeling cost.
- 9
- 9
- Efficient Parallel Training Methods for Spiking Neural Networks with Constant Time Complexity8
Experimental results show that FPT effectively simulates the dynamics of original LIF neurons, significantly reducing computational time without sacrificing accuracy, which makes FPT a scalable and efficient solution for real-world applications, particularly for long-term tasks.
- 8
- 8
- 8
- Boost Test-Time Performance with Closed-Loop Inference8
A general Closed-Loop Inference (CLI) method that first devise a filtering criterion to identify those hard-classified test samples that need additional inference loops and construct looped inference, so that the original erroneous predictions on these hard test samples can be corrected with little additional effort.
- 8
- TLRec: A Transfer Learning Framework to Enhance Large Language Models for Sequential Recommendation Tasks7
TLRec, a transfer learning framework aimed at enhancing LLMs for sequential recommendation tasks, focuses on text inputs to mitigate the challenge of limited transferability across diverse domains, offering promising advantages over traditional recommendation models that heavily depend on unique identities.
- Pareto-aware Neural Architecture Generation for Diverse Computational Budgets7
A Pareto-aware Neural Architecture Generator (PNAG) which only needs to be trained once and dynamically produces the Pare to optimal architecture for any given budget via inference, which not only greatly reduces the overall search cost but also improves the search results.
- 7
- Bandit Online Learning on Graphs via Adaptive Optimization7
This work proposes a bandit online algorithm on graphs that derives per-instance confidence region of the prediction, from which the model can be learned adaptively to minimize the online regret.
- Robust Asymmetric Recommendation via Min-Max Optimization7
A robust asymmetric recommendation model that integrates cost-sensitive learning with capped unilateral loss into a joint objective function, which can be optimized by an iteratively weighted approach and which demonstrates the effectiveness of the algorithm on benchmark recommendation datasets.
- 7
- Sparse Online Relative Similarity Learning7
Several Sparse Online Relative Similarity (SORS) learning algorithms are introduced, which learn a sparse model during the learning process, so that the memory and computational cost can be significantly reduced.
- Genesis: Evolving Attack Strategies for LLM Web Agent Red-Teaming6
Genesis is a novel agentic framework composed of three modules: Attacker, Scorer, and Strategist that dynamically uncovers effective strategies from interaction logs and compiles them into a continuously growing strategy library, which is then re-deployed to enhance the Attacker's effectiveness.
- Spurious Feature Eraser: Stabilizing Test-Time Adaptation for Vision-Language Foundation Model6
This work introduces a test-time prompt tuning paradigm that optimizes a learnable prompt, thereby compelling the model to exploit invariant features while disregarding decision shortcuts during the inference phase, and conducts comparative analysis of the proposed method against various approaches which validates the significant superiority.
- 6
- 6
- High-dimensional Similarity Learning via Dual-sparse Random Projection6
This paper adopts a dual-sparse regularized random projection method that introduces a sparse regularizer into the reduced dual problem, as the original dual solution is a sparse one and applying a sparseRegularizer in the reduced space relaxes the low-rank assumption.
- Unified Locally Linear Classifiers With Diversity-Promoting Anchor Points6
Diversified and Unified Locally Linear Support Vector Machine (DU-LLSVM) is the first principled method that directly learns sparse local coding and can be easily generalized to other supervised learning models.
- Efficient multi-class selective sampling on graphs6
A unified framework which operates directly on the multi- class problem without reducing it to a collection of binary tasks is suggested, which makes active learning practically feasible for multi-class problems, while the one-vs-all strategy cannot.
- Matrix eigen-decomposition via doubly stochastic riemannian optimization6
A Doubly Stochastic Riemannian Gradient EIGenSolver, DSRG-EIGS is proposed, where the double stochasticity comes from the generalization of the Stochastic Euclidean gradient ascent and the stochastics coordinate ascent to RiemANNian manifolds, which induces a greatly reduced complexity per iteration and enables the algorithm to completely avoid the matrix inversion.
- Adapt in the Wild: Test-Time Entropy Minimization with Sharpness and Feature Regularization5
This paper proposes a sharpness-aware and reliable entropy minimization method, called SAR, for stabilizing TTA from two aspects: to remove partial noisy samples with large gradients, and encourage model weights to go to a flat minimum so that the model is robust to the remaining noisy samples.
- DRFLM: Distributionally Robust Federated Learning with Inter-client Noise via Local Mixup5
A general framework to solve the above two challenges simultaneously, using distributionally robust optimization to mitigate the negative effects caused by data heterogeneity paradigm to sample clients based on a learnable distribution at each iteration is proposed.
- 5
- Robust Factorization Machine: A Doubly Capped Norms Minimization5
A new scheme of Robust Factorization Machine (RFM) is proposed by exploring a doubly capped norms minimization approach, which employs both a capped squared trace norm in achieving a tighter approximation of the rank minimization and a capped `1norm loss to enhance the robustness of the empirical loss minimization from noisy data.
- Decentralized Optimization with Edge Sampling5
A decentralized distributed algorithm with stochastic communication among nodes, building on a sampling method called “edge sampling”, which is shown to be unbiased, communication-efficient and suffers from lower sampling variances.
- Learning Correlative and Personalized Structure for Online Multi-Task Classification5
This paper proposes a general online MTL framework that overcomes this restriction by decomposing the weight matrix into two components: the first component captures the correlative structure among tasks in a low-rank subspace, and the second component identifies the personalized patterns for the outlier tasks.
- Learning Relative Similarity by Stochastic Dual Coordinate Ascent5
Theoretically, this paper proves the optimal linear convergence rate for the proposed SDCA algorithm, beating the well-known sublinear convergence rate by the previous best metric learning algorithms.
- Kernel based online learning5
This thesis investigates effective DUOL algorithms for both binary and multi-class online classification tasks and proposes a family of cost- sensitive online classification algorithms by exploiting the DUOL techniques, which are designed to directly optimize two well-known cost-sensitive measures.
- SeqPE: Transformer With Sequential Position Encoding4
Experiments across language modeling, longcontext question answering, and 2D image classification demonstrate that SeqPE not only surpasses strong baselines in perplexity, exact match (EM), and accuracy—particularly under context length extrapolation—but also enables seamless generalization to multi-dimensional inputs without requiring manual architectural redesign.
- SyNDock: N Rigid Protein Docking via Learnable Group Synchronization4
SyNDock is introduced, an automated framework that swiftly assembles precise multimeric complexes within seconds, showcasing performance that can potentially surpass or be on par with recent advanced approaches.
- Reweighted Mixup for Subpopulation Shift4
This work proposes a simple yet practical framework, called reweighted mixup (RMIX), to mitigate the overfitting issue in over-parameterized models by conducting importance weighting on the ''mixed'' samples by leveraging reweighting in mixup.
- ETDock: A Novel Equivariant Transformer for Protein-Ligand Docking4
This work proposes an equivariant transformer neural network for protein-ligand docking pose prediction that can achieve state-of-the-art performance and employs an iterative optimization approach based on the predicted distance matrix to generate refined ligand poses.
- 4
- Structure-aware Mathematical Expression Recognition with Sequence-Level Modeling4
This work proposes a Structure-Aware Sequence-Level (SASL) model for MER to better represent and recognize the visual content of formula images, and proposes a structure-aware module to capture the relationship among different symbols.
- 4
- 4
- 4
- 4
- MSDformer: Multi-Scale Discrete Transformer for Time Series Generation3
This work proposes a novel multi-scale DTM-based time series generation method, called Multi-Scale Discrete Transformer (MSDformer), which employs a multi-scale time series tokenizer to learn discrete token representations at multiple scales, which jointly characterize the complex nature of time series data.
- Learning Long-Range Action Representation by Two-Stream Mamba Pyramid Network for Figure Skating Assessment3
A two-stream Mamba pyramid network that aligns with actual judging criteria to predict TES and PCS by separating visual-feature based TES evaluation stream from audio-visual-feature based PCS evaluation stream and introduces a multi-level fusion mechanism to guarantee that video-based features remain unaffected when assessing TES.
- Online parallel multi-task relationship learning via alternating direction method of multipliers3
This study proposes a novel OMTL framework based on the alternating direction multiplier method (ADMM), a recent breakthrough in optimization suitable for the distributed computing environment because of its decomposable and easy-to-implement nature.
- 3
- Curriculum Graph Poisoning3
This paper revisits the bi-level optimization problem in graph poisoning and proposes a novel graph poisoning method, termed Curriculum Graph Poisoning (CuGPo), inspired by curriculum learning, which achieves state-of-the-art performance inGraph poisoning attacks.
- Geometric Graph Learning for Protein Mutation Effect Prediction3
A new graph learning framework, Hierarchical Graph Invariant Network (HGIN), a fine-grained and data-efficient graph neural encoder for encoding protein structures and predicting the mutation effect on protein properties, which outperforms 9 state-of-the-art approaches on 3 protein datasets.
- 3
- Collaborative online ranking algorithms for multitask learning3
This paper considers a case in which the information that is learned for a task can be used to enhance the learning of other tasks and proposes a collaborative multitask ranking method that learns several ranking models for each of the related tasks together.
- Temporal Kernel Descriptors for Learning with Time-sensitive Patterns3
A novel multiple-kernel learning view is offered and Temporal Kernel Descriptors which utilize Kernel functions to comprehensively detect temporal patterns by deriving relationship of instances with the time features are developed.
- Online learning of ARIMA for time series prediction3
This paper proposes on- line learning algorithms for estimating ARIMA models un-der relaxed assumptions on the noise terms, which is suit- able to a wider range of applications and enjoys high com-putational efficiency.
- 2
- 2
- 2
- 2
- Enhancing Neural Subset Selection: Integrating Background Information into Set Representations2
This work proposes a simple yet effective information aggregation module designed to merge the representations of subsets and supersets from a permutation invariance perspective, underscoring the practicality and potency of the proposed strategies in real-world contexts.
- Exploring the Generalization Capabilities of AID-based Bi-level Optimization2
The uniform stability of AID-based methods is ascertain, which achieves similar results to a single-level nonconvex problem, and the generalization ability of AID-based bi-level optimization methods is given.
- Variational Bayes for Federated Continual Learning2
Federated Bayesian Neural Network is introduced, a versatile and efficacious framework employing a variational Bayesian neural network across all clients, which continually integrates knowledge from local and historical data distributions into a single model, adeptly learning from new data distributions while retaining performance on historical distributions.
- Deploying Offline Reinforcement Learning with Human Feedback2
An alternative framework that involves a human supervising the RL models and providing additional feedback in the online deployment phase and develops two approaches that demonstrate the effectiveness of these approaches for robot locomotion control and traffic light control tasks through empirical validation.
- Attention Paper: How Generative AI Reshapes Digital Shadow Industry?2
A technical analysis of the challenges and opportunities of AIGC from upstream, midstream, and downstream paths of black/shadow industries and suggest future directions for improving existing risk control systems are provided.
- Predicting Protein-Ligand Binding Affinity with Equivariant Line Graph Network2
A novel Equivariant Line Graph Network (ELGN) is proposed for affinity prediction of 3D protein ligand complexes that firstly adds a super node to the 3D complex, and then builds a line graph based on the3D complex.
- Learning Neural Set Functions Under the Optimal Subset Oracle2
This work presents a principled yet practical maximum likelihood learning framework, termed as EquiVSet1, that simultaneously meets the following desiderata of learning set functions under the OS oracle: i) permutation invariance of the set mass function being modeled; ii) permission of varying ground set; iii) fully differentiability; iv) minimum prior; and v) scalability
- FusionRetro: Molecule Representation Fusion via In-Context Learning for Retrosynthetic Planning2
This work develops and proposes Metro: Memory-Enhanced Transformer for RetrOsynthetic planning, a model that dramatically outperforms existing single-step retrosynthesis models by at least 10.7% in top-1 accuracy and can be directly used for synthetic accessibility analysis.
- Robust Offline Reinforcement Learning with Gradient Penalty and Constraint Relaxation2
Experimental results show that the proposed techniques effectively tame the non-optimal trajectories for policy constraint offline RL methods, evaluated on a set of contaminated D4RL Mujoco and Adroit datasets.
- 2
- 2
- Tackling Non-Stationarity in HVAC Control with TimeGPT-Enhanced Deep Reinforcement Learning1
Simulation studies on the BEAR platform demonstrate that TimeGPT's zero-shot forecasting capability offers superior performance, computational efficiency, and deployment simplicity, and integrating these forecasts with conventional HVAC state variables accelerates DRL policy convergence and reduces controlled temperature fluctuations within a non-stationary environment.
- Hexagon-Net: Heterogeneous Cross-View Aligned Graph Attention Networks for Implied Volatility Surface Prediction1
A novel end-to-end heterogeneous cross(x)-view aligned graph attention network (Hexagon-Net) is proposed, which aligns historical IVS data, learns distinctive IVS patterns, propagates predictive information, and forecasts future IVS movements simultaneously.
- Thermal-Delay Aware Deep Reinforcement Learning for HVAC System Control1
The Thermal Delay Embedding Network (TDE-Net) is proposed, a novel representation learning method that explicitly captures the historical influence of control actions on the current state that significantly outperforms baseline DRL approaches in maintaining thermal comfort, reducing temperature fluctuations, and minimizing energy consumption.
- 1
- DFWLayer: Differentiable Frank-Wolfe Optimization Layer1
Experimental results demonstrate that the DFWLayer not only attains competitive accuracy in solutions and gradients but also consistently adheres to constraints.
- 1
- 1
- 1
- Vertical Federated Linear Contextual Bandits1
The proposed protocols can perfectly recover the service quality of centralized bandit algorithms while achieving a satisfactory runtime efficiency, which is theoretically proved and analyzed in this paper.
- 1
- Aggregated Gradient Langevin Dynamics1
The theoretic results indicate that methods in AGLD possess the merits of both the low per-iteration computational complexity and the short mixture time, and empirical studies demonstrate that the framework allows to derive novel schemes to generate high-quality samples for large-scale Bayesian posterior learning tasks.
- 1
- 1
- –
- –
- –
- ANCHOR: Taming Entropy Dynamics for Stable and Efficient Reasoning of Large Language Models–
It is shown that policy entropy bounds both the policy gradient and probability update norms; consequently, entropy collapse effectively stops reward signal backpropagation, preventing further policy learning regardless of data quality.
- –
- –
- –
- –
- Efficient Distributed Sparse Relative Similarity Learning–
This article proposes an Efficient Distributed Sparse Relative Similarity Learning framework, and effectively reduces the number of updates through synchronous delay while maintaining a large batch size, addressing the issue of high computational cost.
- –
- –
- –
- –
- –
- SR-R$^2$KAC: Improving Single Image Defocus Deblurring–
An efficient deep learning method for single image defocus deblurring (SIDD) by further exploring inverse kernel properties and adding identity shortcuts between atrous convolutions to simulate residual deconvolutions.
- Quantized Adaptive Subgradient Algorithms and Their Applications–
This work proposes quantized composite mirror descent adaptive subgradient (QCMD adagrad) and quantized regularized dual average adaptive subgradient (QRDA adagrad) and quantized regularized dual average adaptive subgradient for distributed training to balance the communication costs, model capacity and model performance simultaneously.
- GPN: A Joint Structural Learning Framework for Graph Neural Networks–
This paper proposes Generative Predictive Network (GPN), a GNN-based joint learning framework that simultaneously learns the graph structure and the downstream task, and develops a bilevel optimization framework for this joint learning task, in which the upper optimization and the lower optimization are both instantiated with GNNs.
- –
- –
- –
- –
- –
- –
- –
- A Local Online Learning Approach for Non-linear Data–
A Soft Confidence-Weighted Local Online Learning (SCW-LOL) algorithm, which extends the single hyperplane SCW to the case with multiple local hyperplanes, which learns an online convergence boundary, overall achieving the best performance over almost all datasets, without any kernel modeling and parameter tuning.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.