Ning Ma

Ming Gu, Gaoming Yang, Sheng Zhou et al.

Graph clustering is a fundamental task in graph analysis, and recent advances in utilizing graph neural networks (GNNs) have shown impressive results. Despite the success of existing GNN-based graph clustering methods, they often overlook the quality of graph structure, which is inherent in real-world graphs due to their sparse and multifarious nature, leading to subpar performance. Graph structure learning allows refining the input graph by adding missing links and removing spurious connections. However, previous endeavors in graph structure learning have predominantly centered around supervised settings, and cannot be directly applied to our specific clustering tasks due to the absence of ground-truth labels. To bridge the gap, we propose a novel method called \textbf{ho}mophily-enhanced structure \textbf{le}arning for graph clustering (HoLe). Our motivation stems from the observation that subtly enhancing the degree of homophily within the graph structure can significantly improve GNNs and clustering outcomes. To realize this objective, we develop two clustering-oriented structure learning modules, i.e., hierarchical correlation estimation and cluster-aware sparsification. The former module enables a more accurate estimation of pairwise node relationships by leveraging guidance from latent and clustering spaces, while the latter one generates a sparsified structure based on the similarity matrix and clustering assignments. Additionally, we devise a joint optimization approach alternating between training the homophily-enhanced structure learning and GNN-based clustering, thereby enforcing their reciprocal effects. Extensive experiments on seven benchmark datasets of various types and scales, across a range of clustering metrics, demonstrate the superiority of HoLe against state-of-the-art baselines.

Shu Yang, Zhiyuan Cai, Luyang Luo et al.

Capitalizing on image-level pre-trained models for various downstream tasks has recently emerged with promising performance. However, the paradigm of "image pre-training followed by video fine-tuning" for high-dimensional video data inevitably poses significant performance bottlenecks. Furthermore, in the medical domain, many surgical video tasks encounter additional challenges posed by the limited availability of video data and the necessity for comprehensive spatial-temporal modeling. Recently, Parameter-Efficient Image-to-Video Transfer Learning has emerged as an efficient and effective paradigm for video action recognition tasks, which employs image-level pre-trained models with promising feature transferability and involves cross-modality temporal modeling with minimal fine-tuning. Nevertheless, the effectiveness and generalizability of this paradigm within intricate surgical domain remain unexplored. In this paper, we delve into a novel problem of efficiently adapting image-level pre-trained models to specialize in fine-grained surgical phase recognition, termed as Parameter-Efficient Image-to-Surgical-Video Transfer Learning. Firstly, we develop a parameter-efficient transfer learning benchmark SurgPETL for surgical phase recognition, and conduct extensive experiments with three advanced methods based on ViTs of two distinct scales pre-trained on five large-scale natural and medical datasets. Then, we introduce the Spatial-Temporal Adaptation module, integrating a standard spatial adapter with a novel temporal adapter to capture detailed spatial features and establish connections across temporal sequences for robust spatial-temporal modeling. Extensive experiments on three challenging datasets spanning various surgical procedures demonstrate the effectiveness of SurgPETL with STA.

Ke Liu, Feng Liu, Haishuai Wang et al.

$\textit{Implicit neural representations}$ (INRs) aim to learn a $\textit{continuous function}$ (i.e., a neural network) to represent an image, where the input and output of the function are pixel coordinates and RGB/Gray values, respectively. However, images tend to consist of many objects whose colors are not perfectly consistent, resulting in the challenge that image is actually a $\textit{discontinuous piecewise function}$ and cannot be well estimated by a continuous function. In this paper, we empirically investigate that if a neural network is enforced to fit a discontinuous piecewise function to reach a fixed small error, the time costs will increase exponentially with respect to the boundaries in the spatial domain of the target signal. We name this phenomenon the $\textit{exponential-increase}$ hypothesis. Under the $\textit{exponential-increase}$ hypothesis, learning INRs for images with many objects will converge very slowly. To address this issue, we first prove that partitioning a complex signal into several sub-regions and utilizing piecewise INRs to fit that signal can significantly speed up the convergence. Based on this fact, we introduce a simple partition mechanism to boost the performance of two INR methods for image reconstruction: one for learning INRs, and the other for learning-to-learn INRs. In both cases, we partition an image into different sub-regions and dedicate smaller networks for each part. In addition, we further propose two partition rules based on regular grids and semantic segmentation maps, respectively. Extensive experiments validate the effectiveness of the proposed partitioning methods in terms of learning INR for a single image (ordinary learning framework) and the learning-to-learn framework.

Yiou Huang, Ning Ma, Weichu Zhao et al.

Imitation learning (IL) has shown strong potential for contact-rich precision insertion tasks. However, its practical deployment is often hindered by covariate shift and the need for continuous expert monitoring to recover from failures during execution. In this paper, we propose Trajectory Editing Residual Dataset Aggregation (TER-DAgger), a scalable and force-aware human-in-the-loop imitation learning framework that mitigates covariate shift by learning residual policies through optimization-based trajectory editing. This approach smoothly fuses policy rollouts with human corrective trajectories, providing consistent and stable supervision. Second, we introduce a force-aware failure anticipation mechanism that triggers human intervention only when discrepancies arise between predicted and measured end-effector forces, significantly reducing the requirement for continuous expert monitoring. Third, all learned policies are executed within a Cartesian impedance control framework, ensuring compliant and safe behavior during contact-rich interactions. Extensive experiments in both simulation and real-world precision insertion tasks show that TER-DAgger improves the average success rate by over 37\% compared to behavior cloning, human-guided correction, retraining, and fine-tuning baselines, demonstrating its effectiveness in mitigating covariate shift and enabling scalable deployment in contact-rich manipulation.

Fuwen Luo, Chi Chen, Zihao Wan et al. · tsinghua

Multimodal large language models (MLLMs) have demonstrated promising results in a variety of tasks that combine vision and language. As these models become more integral to research and applications, conducting comprehensive evaluations of their capabilities has grown increasingly important. However, most existing benchmarks fail to consider that, in certain situations, images need to be interpreted within a broader context. In this work, we introduce a new benchmark, named as CODIS, designed to assess the ability of models to use context provided in free-form text to enhance visual comprehension. Our findings indicate that MLLMs consistently fall short of human performance on this benchmark. Further analysis confirms that these models struggle to effectively extract and utilize contextual information to improve their understanding of images. This underscores the pressing need to enhance the ability of MLLMs to comprehend visuals in a context-dependent manner. View our project website at https://thunlp-mt.github.io/CODIS.

Anthony Hughes, Ning Ma, Nikolaos Aletras

In sensitive domains such as medical and legal, protecting sensitive information is critical, with protective laws strictly prohibiting the disclosure of personal data. This poses challenges for sharing valuable data such as medical reports and legal cases summaries. While language models (LMs) have shown strong performance in text summarization, it is still an open question to what extent they can provide privacy-preserving summaries from non-private source documents. In this paper, we perform a comprehensive study of privacy risks in LM-based summarization across two closed- and four open-weight models of different sizes and families. We experiment with both prompting and fine-tuning strategies for privacy-preservation across a range of summarization datasets including medical and legal domains. Our quantitative and qualitative analysis, including human evaluation, shows that LMs frequently leak personally identifiable information in their summaries, in contrast to human-generated privacy-preserving summaries, which demonstrate significantly higher privacy protection levels. These findings highlight a substantial gap between current LM capabilities and expert human expert performance in privacy-sensitive summarization tasks.

Jun Sun, Boyu Yang, Jiahao Zhang et al.

Pretrained Vision-Language-Action (VLA) policies have achieved strong single-step manipulation, but their inference remains largely memoryless, which is brittle in non-Markovian long-horizon settings with occlusion, state aliasing, and subtle post-action changes. Prior approaches inject history either by stacking frames, which scales visual tokens and latency while adding near-duplicate pixels, or by learning additional temporal interfaces that require (re-)training and may break the original single-frame inference graph. We present TempoFit, a training-free temporal retrofit that upgrades frozen VLAs through state-level memory. Our key insight is that prefix attention K/V already form a model-native, content-addressable runtime state; reusing them across timesteps introduces history without new tokens or trainable modules. TempoFit stores layer-wise FIFO prefix K/V at selected intermediate layers, performs parameter-free K-to-K retrieval with Frame-Gap Temporal Bias (FGTB), a fixed recency bias inspired by positional biases in NLP, to keep decisions present-dominant, and injects the retrieved context via pre-attention residual loading with norm-preserving rescaling to avoid distribution shift under frozen weights. On LIBERO-LONG, TempoFit improves strong pretrained backbones by up to +4.0% average success rate while maintaining near-real-time latency, and it transfers consistently to CALVIN and real-robot long-horizon tasks.

Anthony Hughes, Vasisht Duddu, N. Asokan et al.

Language models (LMs) may memorize personally identifiable information (PII) from training data, enabling adversaries to extract it during inference. Existing defense mechanisms such as differential privacy (DP) reduce this leakage, but incur large drops in utility. Based on a comprehensive study using circuit discovery to identify the computational circuits responsible PII leakage in LMs, we hypothesize that specific PII leakage circuits in LMs should be responsible for this behavior. Therefore, we propose PATCH (Privacy-Aware Targeted Circuit PatcHing), a novel approach that first identifies and subsequently directly edits PII circuits to reduce leakage. PATCH achieves better privacy-utility trade-off than existing defenses, e.g., reducing recall of PII leakage from LMs by up to 65%. Finally, PATCH can be combined with DP to reduce recall of residual leakage of an LM to as low as 0.01%. Our analysis shows that PII leakage circuits persist even after the application of existing defense mechanisms. In contrast, PATCH can effectively mitigate their impact.

Xiaolei Xu, Chaoyue Niu, Guy J. Brown et al.

Obstructive sleep apnoea (OSA) is a prevalent condition with significant health consequences, yet many patients remain undiagnosed due to the complexity and cost of over-night polysomnography. Acoustic-based screening provides a scalable alternative, yet performance is limited by environmental noise and the lack of physiological context. Respiratory effort is a key signal used in clinical scoring of OSA events, but current approaches require additional contact sensors that reduce scalability and patient comfort. This paper presents the first study to estimate respiratory effort directly from nocturnal audio, enabling physiological context to be recovered from sound alone. We propose a latent-space fusion framework that integrates the estimated effort embeddings with acoustic features for OSA detection. Using a dataset of 157 nights from 103 participants recorded in home environments, our respiratory effort estimator achieves a concordance correlation coefficient of 0.48, capturing meaningful respiratory dynamics. Fusing effort and audio improves sensitivity and AUC over audio-only baselines, especially at low apnoea-hypopnoea index thresholds. The proposed approach requires only smartphone audio at test time, which enables sensor-free, scalable, and longitudinal OSA monitoring.

Hanqun Cao, Mutian He, Ning Ma et al.

DNA-encoded library (DEL) screening has revolutionized the detection of protein-ligand interactions through read counts, enabling rapid exploration of vast chemical spaces. However, noise in read counts, stemming from nonspecific interactions, can mislead this exploration process. We present DEL-Ranking, a novel distribution-correction denoising framework that addresses these challenges. Our approach introduces two key innovations: (1) a novel ranking loss that rectifies relative magnitude relationships between read counts, enabling the learning of causal features determining activity levels, and (2) an iterative algorithm employing self-training and consistency loss to establish model coherence between activity label and read count predictions. Furthermore, we contribute three new DEL screening datasets, the first to comprehensively include multi-dimensional molecular representations, protein-ligand enrichment values, and their activity labels. These datasets mitigate data scarcity issues in AI-driven DEL screening research. Rigorous evaluation on diverse DEL datasets demonstrates DEL-Ranking's superior performance across multiple correlation metrics, with significant improvements in binding affinity prediction accuracy. Our model exhibits zero-shot generalization ability across different protein targets and successfully identifies potential motifs determining compound binding affinity. This work advances DEL screening analysis and provides valuable resources for future research in this area.

Yifu Zhang, Hongru Li, Tao Yang et al.

Lesion segmentation of ultrasound medical images based on deep learning techniques is a widely used method for diagnosing diseases. Although there is a large amount of ultrasound image data in medical centers and other places, labeled ultrasound datasets are a scarce resource, and it is likely that no datasets are available for new tissues/organs. Transfer learning provides the possibility to solve this problem, but there are too many features in natural images that are not related to the target domain. As a source domain, redundant features that are not conducive to the task will be extracted. Migration between ultrasound images can avoid this problem, but there are few types of public datasets, and it is difficult to find sufficiently similar source domains. Compared with natural images, ultrasound images have less information, and there are fewer transferable features between different ultrasound images, which may cause negative transfer. To this end, a multi-source adversarial transfer learning network for ultrasound image segmentation is proposed. Specifically, to address the lack of annotations, the idea of adversarial transfer learning is used to adaptively extract common features between a certain pair of source and target domains, which provides the possibility to utilize unlabeled ultrasound data. To alleviate the lack of knowledge in a single source domain, multi-source transfer learning is adopted to fuse knowledge from multiple source domains. In order to ensure the effectiveness of the fusion and maximize the use of precious data, a multi-source domain independent strategy is also proposed to improve the estimation of the target domain data distribution, which further increases the learning ability of the multi-source adversarial migration learning network in multiple domains.

Ning Ma, Jiajun Bu, Lixian Lu et al.

Domain Adaptation has been widely used to deal with the distribution shift in vision, language, multimedia etc. Most domain adaptation methods learn domain-invariant features with data from both domains available. However, such a strategy might be infeasible in practice when source data are unavailable due to data-privacy concerns. To address this issue, we propose a novel adaptation method via hypothesis transfer without accessing source data at adaptation stage. In order to fully use the limited target data, a semi-supervised mutual enhancement method is proposed, in which entropy minimization and augmented label propagation are used iteratively to perform inter-domain and intra-domain alignments. Compared with state-of-the-art methods, the experimental results on three public datasets demonstrate that our method gets up to 19.9% improvements on semi-supervised adaptation tasks.

Ning Ma, Jiajun Bu, Zhen Zhang et al.

Present domain adaptation methods usually perform explicit representation alignment by simultaneously accessing the source data and target data. However, the source data are not always available due to the privacy preserving consideration or bandwidth limitation. Source-free domain adaptation aims to solve the above problem by performing domain adaptation without accessing the source data. The adaptation paradigm is receiving more and more attention in recent years, and multiple works have been proposed for unsupervised source-free domain adaptation. However, without utilizing any supervised signal and source data at the adaptation stage, the optimization of the target model is unstable and fragile. To alleviate the problem, we focus on semi-supervised domain adaptation under source-free setting. More specifically, we propose uncertainty-guided Mixup to reduce the representation's intra-domain discrepancy and perform inter-domain alignment without directly accessing the source data. Finally, we conduct extensive semi-supervised domain adaptation experiments on various datasets. Our method outperforms the recent semi-supervised baselines and the unsupervised variant also achieves competitive performance. The experiment codes will be released in the future.

Zehai Tu, Ning Ma, Jon Barker

Current hearing aids normally provide amplification based on a general prescriptive fitting, and the benefits provided by the hearing aids vary among different listening environments despite the inclusion of noise suppression feature. Motivated by this fact, this paper proposes a data-driven machine learning technique to develop hearing aid fittings that are customised to speech in different noisy environments. A differentiable hearing loss model is proposed and used to optimise fittings with back-propagation. The customisation is reflected on the data of speech in different noise with also the consideration of noise suppression. The objective evaluation shows the advantages of optimised custom fittings over general prescriptive fittings.

Zehai Tu, Ning Ma, Jon Barker

Hearing aids are expected to improve speech intelligibility for listeners with hearing impairment. An appropriate amplification fitting tuned for the listener's hearing disability is critical for good performance. The developments of most prescriptive fittings are based on data collected in subjective listening experiments, which are usually expensive and time-consuming. In this paper, we explore an alternative approach to finding the optimal fitting by introducing a hearing aid speech processing framework, in which the fitting is optimised in an automated way using an intelligibility objective function based on the HASPI physiological auditory model. The framework is fully differentiable, thus can employ the back-propagation algorithm for efficient, data-driven optimisation. Our initial objective experiments show promising results for noise-free speech amplification, where the automatically optimised processors outperform one of the well recognised hearing aid prescriptions.

Ning Ma, Jiajun Bu, Jieyu Yang et al.

Graph classification aims to extract accurate information from graph-structured data for classification and is becoming more and more important in graph learning community. Although Graph Neural Networks (GNNs) have been successfully applied to graph classification tasks, most of them overlook the scarcity of labeled graph data in many applications. For example, in bioinformatics, obtaining protein graph labels usually needs laborious experiments. Recently, few-shot learning has been explored to alleviate this problem with only given a few labeled graph samples of test classes. The shared sub-structures between training classes and test classes are essential in few-shot graph classification. Exiting methods assume that the test classes belong to the same set of super-classes clustered from training classes. However, according to our observations, the label spaces of training classes and test classes usually do not overlap in real-world scenario. As a result, the existing methods don't well capture the local structures of unseen test classes. To overcome the limitation, in this paper, we propose a direct method to capture the sub-structures with well initialized meta-learner within a few adaptation steps. More specifically, (1) we propose a novel framework consisting of a graph meta-learner, which uses GNNs based modules for fast adaptation on graph data, and a step controller for the robustness and generalization of meta-learner; (2) we provide quantitative analysis for the framework and give a graph-dependent upper bound of the generalization error based on our framework; (3) the extensive experiments on real-world datasets demonstrate that our framework gets state-of-the-art results on several few-shot graph classification tasks compared to baselines.

Mengyuan Lee, Ning Ma, Guanding Yu et al.

Generalized Benders decomposition (GBD) is a globally optimal algorithm for mixed integer nonlinear programming (MINLP) problems, which are NP-hard and can be widely found in the area of wireless resource allocation. The main idea of GBD is decomposing an MINLP problem into a primal problem and a master problem, which are iteratively solved until their solutions converge. However, a direct implementation of GBD is time- and memory-consuming. The main bottleneck is the high complexity of the master problem, which increases over the iterations. Therefore, we propose to leverage machine learning (ML) techniques to accelerate GBD aiming at decreasing the complexity of the master problem. Specifically, we utilize two different ML techniques, classification and regression, to deal with this acceleration task. In this way, a cut classifier and a cut regressor are learned, respectively, to distinguish between useful and useless cuts. Only useful cuts are added to the master problem and thus the complexity of the master problem is reduced. By using a resource allocation problem in device-to-device communication networks as an example, we validate that the proposed method can reduce the computational complexity of GBD without loss of optimality and has strong generalization ability. The proposed method is applicable for solving various MINLP problems in wireless networks since the designs are invariant for different problems.

Ning Ma, Jose A. Gonzalez, Guy J. Brown

Despite there being clear evidence for top-down (e.g., attentional) effects in biological spatial hearing, relatively few machine hearing systems exploit top-down model-based knowledge in sound localisation. This paper addresses this issue by proposing a novel framework for binaural sound localisation that combines model-based information about the spectral characteristics of sound sources and deep neural networks (DNNs). A target source model and a background source model are first estimated during a training phase using spectral features extracted from sound signals in isolation. When the identity of the background source is not available, a universal background model can be used. During testing, the source models are used jointly to explain the mixed observations and improve the localisation process by selectively weighting source azimuth posteriors output by a DNN-based localisation system. To address the possible mismatch between training and testing, a model adaptation process is further employed on-the-fly during testing, which adapts the background model parameters directly from the noisy observations in an iterative manner. The proposed system therefore combines model-based and data-driven information flow within a single computational framework. The evaluation task involved localisation of a target speech source in the presence of an interfering source and room reverberation. Our experiments show that by exploiting model-based information in this way, sound localisation performance can be improved substantially under various noisy and reverberant conditions.

Ning Ma, Tobias May, Guy J. Brown

This paper presents a novel machine-hearing system that exploits deep neural networks (DNNs) and head movements for robust binaural localisation of multiple sources in reverberant environments. DNNs are used to learn the relationship between the source azimuth and binaural cues, consisting of the complete cross-correlation function (CCF) and interaural level differences (ILDs). In contrast to many previous binaural hearing systems, the proposed approach is not restricted to localisation of sound sources in the frontal hemifield. Due to the similarity of binaural cues in the frontal and rear hemifields, front-back confusions often occur. To address this, a head movement strategy is incorporated in the localisation model to help reduce the front-back errors. The proposed DNN system is compared to a Gaussian mixture model (GMM) based system that employs interaural time differences (ITDs) and ILDs as localisation features. Our experiments show that the DNN is able to exploit information in the CCF that is not available in the ITD cue, which together with head movements substantially improves localisation accuracies under challenging acoustic scenarios in which multiple talkers and room reverberation are present.

Hector E. Romero, Ning Ma, Guy J. Brown et al.

Sleep-disordered breathing (SDB) is a serious and prevalent condition, and acoustic analysis via consumer devices (e.g. smartphones) offers a low-cost solution to screening for it. We present a novel approach for the acoustic identification of SDB sounds, such as snoring, using bottleneck features learned from a corpus of whole-night sound recordings. Two types of bottleneck features are described, obtained by applying a deep autoencoder to the output of an auditory model or a short-term autocorrelation analysis. We investigate two architectures for snore sound detection: a tandem system and a hybrid system. In both cases, a `language model' (LM) was incorporated to exploit information about the sequence of different SDB events. Our results show that the proposed bottleneck features give better performance than conventional mel-frequency cepstral coefficients, and that the tandem system outperforms the hybrid system given the limited amount of labelled training data available. The LM made a small improvement to the performance of both classifiers.

Paolo Vecchiotti, Ning Ma, Stefano Squartini et al.

A novel end-to-end binaural sound localisation approach is proposed which estimates the azimuth of a sound source directly from the waveform. Instead of employing hand-crafted features commonly employed for binaural sound localisation, such as the interaural time and level difference, our end-to-end system approach uses a convolutional neural network (CNN) to extract specific features from the waveform that are suitable for localisation. Two systems are proposed which differ in the initial frequency analysis stage. The first system is auditory-inspired and makes use of a gammatone filtering layer, while the second system is fully data-driven and exploits a trainable convolutional layer to perform frequency analysis. In both systems, a set of dedicated convolutional kernels are then employed to search for specific localisation cues, which are coupled with a localisation stage using fully connected layers. Localisation experiments using binaural simulation in both anechoic and reverberant environments show that the proposed systems outperform a state-of-the-art deep neural network system. Furthermore, our investigation of the frequency analysis stage in the second system suggests that the CNN is able to exploit different frequency bands for localisation according to the characteristics of the reverberant environment.

Ning Ma, Xin Zhao, Mark Bolin

We propose an end-to-end solution, from watermark feature generation to metric design, for effectively demoting watermarked images surfed by a real world image search engine. We use a few fundamental techniques to obtain effective watermark features of images in the image search index, and utilize the signals in a commercial search engine to improve the image search quality. We collect a diverse and large set (about 1M) of images with human labels indicating whether the image contains visible watermark. We train a few deep convolutional neural networks to extract watermark information from the raw images. The deep CNN classifiers we trained can achieve high accuracy on the watermark test data set. We also analyze the images based on their domains to get watermark information from a domain-based watermark classifier. We design a new novel hybrid metric which includes the relevance, image attractiveness and watermark information all together. We demonstrate that using these watermark signals together with the new metric in image search ranker can significantly demote the watermarked images during the online image ranking.

Ning Ma, Alexey Volkov, Aleksandr Livshits et al.

We propose a new framework to rank image attractiveness using a novel pairwise deep network trained with a large set of side-by-side multi-labeled image pairs from a web image index. The judges only provide relative ranking between two images without the need to directly assign an absolute score, or rate any predefined image attribute, thus making the rating more intuitive and accurate. We investigate a deep attractiveness rank net (DARN), a combination of deep convolutional neural network and rank net, to directly learn an attractiveness score mean and variance for each image and the underlying criteria the judges use to label each pair. The extension of this model (DARN-V2) is able to adapt to individual judge's personal preference. We also show the attractiveness of search results are significantly improved by using this attractiveness information in a real commercial search engine. We evaluate our model against other state-of-the-art models on our side-by-side web test data and another public aesthetic data set. With much less judgments (1M vs 50M), our model outperforms on side-by-side labeled data, and is comparable on data labeled by absolute score.