Wentao Zhu

Runqi Wang, Linlin Yang, Baochang Zhang et al.

Research on the generalization ability of deep neural networks (DNNs) has recently attracted a great deal of attention. However, due to their complex architectures and large numbers of parameters, measuring the generalization ability of specific DNN models remains an open challenge. In this paper, we propose to use multiple factors to measure and rank the relative generalization of DNNs based on a new concept of confidence dimension (CD). Furthermore, we provide a feasible framework in our CD to theoretically calculate the upper bound of generalization based on the conventional Vapnik-Chervonenk dimension (VC-dimension) and Hoeffding's inequality. Experimental results on image classification and object detection demonstrate that our CD can reflect the relative generalization ability for different DNNs. In addition to full-precision DNNs, we also analyze the generalization ability of binary neural networks (BNNs), whose generalization ability remains an unsolved problem. Our CD yields a consistent and reliable measure and ranking for both full-precision DNNs and BNNs on all the tasks.

Zhenhua Huang, Kunhao Li, Shaojie Wang et al.

Despite the Graph Neural Networks' (GNNs) proficiency in analyzing graph data, achieving high-accuracy and interpretable predictions remains challenging. Existing GNN interpreters typically provide post-hoc explanations disjointed from GNNs' predictions, resulting in misrepresentations. Self-explainable GNNs offer built-in explanations during the training process. However, they cannot exploit the explanatory outcomes to augment prediction performance, and they fail to provide high-quality explanations of node features and require additional processes to generate explainable subgraphs, which is costly. To address the aforementioned limitations, we propose a self-explained and self-supervised graph neural network (SES) to bridge the gap between explainability and prediction. SES comprises two processes: explainable training and enhanced predictive learning. During explainable training, SES employs a global mask generator co-trained with a graph encoder and directly produces crucial structure and feature masks, reducing time consumption and providing node feature and subgraph explanations. In the enhanced predictive learning phase, mask-based positive-negative pairs are constructed utilizing the explanations to compute a triplet loss and enhance the node representations by contrastive learning.

Shuwei Shao, Zhongcai Pei, Weihai Chen et al.

Recently, self-supervised learning technology has been applied to calculate depth and ego-motion from monocular videos, achieving remarkable performance in autonomous driving scenarios. One widely adopted assumption of depth and ego-motion self-supervised learning is that the image brightness remains constant within nearby frames. Unfortunately, the endoscopic scene does not meet this assumption because there are severe brightness fluctuations induced by illumination variations, non-Lambertian reflections and interreflections during data collection, and these brightness fluctuations inevitably deteriorate the depth and ego-motion estimation accuracy. In this work, we introduce a novel concept referred to as appearance flow to address the brightness inconsistency problem. The appearance flow takes into consideration any variations in the brightness pattern and enables us to develop a generalized dynamic image constraint. Furthermore, we build a unified self-supervised framework to estimate monocular depth and ego-motion simultaneously in endoscopic scenes, which comprises a structure module, a motion module, an appearance module and a correspondence module, to accurately reconstruct the appearance and calibrate the image brightness. Extensive experiments are conducted on the SCARED dataset and EndoSLAM dataset, and the proposed unified framework exceeds other self-supervised approaches by a large margin. To validate our framework's generalization ability on different patients and cameras, we train our model on SCARED but test it on the SERV-CT and Hamlyn datasets without any fine-tuning, and the superior results reveal its strong generalization ability. Code will be available at: \url{https://github.com/ShuweiShao/AF-SfMLearner}.

Tianli Zhao, Xi Sheryl Zhang, Wentao Zhu et al.

For practical deep neural network design on mobile devices, it is essential to consider the constraints incurred by the computational resources and the inference latency in various applications. Among deep network acceleration related approaches, pruning is a widely adopted practice to balance the computational resource consumption and the accuracy, where unimportant connections can be removed either channel-wisely or randomly with a minimal impact on model accuracy. The channel pruning instantly results in a significant latency reduction, while the random weight pruning is more flexible to balance the latency and accuracy. In this paper, we present a unified framework with Joint Channel pruning and Weight pruning (JCW), and achieves a better Pareto-frontier between the latency and accuracy than previous model compression approaches. To fully optimize the trade-off between the latency and accuracy, we develop a tailored multi-objective evolutionary algorithm in the JCW framework, which enables one single search to obtain the optimal candidate architectures for various deployment requirements. Extensive experiments demonstrate that the JCW achieves a better trade-off between the latency and accuracy against various state-of-the-art pruning methods on the ImageNet classification dataset. Our codes are available at https://github.com/jcw-anonymous/JCW.

Wentao Zhu, Can Zhao, Wenqi Li et al.

Deep Learning (DL) models are becoming larger, because the increase in model size might offer significant accuracy gain. To enable the training of large deep networks, data parallelism and model parallelism are two well-known approaches for parallel training. However, data parallelism does not help reduce memory footprint per device. In this work, we introduce Large deep 3D ConvNets with Automated Model Parallelism (LAMP) and investigate the impact of both input's and deep 3D ConvNets' size on segmentation accuracy. Through automated model parallelism, it is feasible to train large deep 3D ConvNets with a large input patch, even the whole image. Extensive experiments demonstrate that, facilitated by the automated model parallelism, the segmentation accuracy can be improved through increasing model size and input context size, and large input yields significant inference speedup compared with sliding window of small patches in the inference. Code is available\footnote{https://monai.io/research/lamp-automated-model-parallelism}.

Zhenhua Huang, Kunhao Li, Shaojie Wang et al.

Graph neural networks (GNNs) are widely applied in graph data modeling. However, existing GNNs are often trained in a task-driven manner that fails to fully capture the intrinsic nature of the graph structure, resulting in sub-optimal node and graph representations. To address this limitation, we propose a novel Graph structure Prompt Learning method (GPL) to enhance the training of GNNs, which is inspired by prompt mechanisms in natural language processing. GPL employs task-independent graph structure losses to encourage GNNs to learn intrinsic graph characteristics while simultaneously solving downstream tasks, producing higher-quality node and graph representations. In extensive experiments on eleven real-world datasets, after being trained by GPL, GNNs significantly outperform their original performance on node classification, graph classification, and edge prediction tasks (up to 10.28%, 16.5%, and 24.15%, respectively). By allowing GNNs to capture the inherent structural prompts of graphs in GPL, they can alleviate the issue of over-smooth and achieve new state-of-the-art performances, which introduces a novel and effective direction for GNN research with potential applications in various domains.

Wentao Zhu

Vision transformers (ViTs) have achieved promising results on a variety of Computer Vision tasks, however their quadratic complexity in the number of input tokens has limited their application specially in resource-constrained settings. Previous approaches that employ gradual token reduction to address this challenge assume that token redundancy in one layer implies redundancy in all the following layers. We empirically demonstrate that this assumption is often not correct, i.e., tokens that are redundant in one layer can be useful in later layers. We employ this key insight to propose a novel token propagation controller (TPC) that incorporates two different token-distributions, i.e., pause probability and restart probability to control the reduction and reuse of tokens respectively, which results in more efficient token utilization. To improve the estimates of token distributions, we propose a smoothing mechanism that acts as a regularizer and helps remove noisy outliers. Furthermore, to improve the training-stability of our proposed TPC, we introduce a model stabilizer that is able to implicitly encode local image structures and minimize accuracy fluctuations during model training. We present extensive experimental results on the ImageNet-1K dataset using DeiT, LV-ViT and Swin models to demonstrate the effectiveness of our proposed method. For example, compared to baseline models, our proposed method improves the inference speed of the DeiT-S by 250% while increasing the classification accuracy by 1.0%.

Wentao Zhu

Transformers have achieved promising results on a variety of tasks. However, the quadratic complexity in self-attention computation has limited the applications, especially in low-resource settings and mobile or edge devices. Existing works have proposed to exploit hand-crafted attention patterns to reduce computation complexity. However, such hand-crafted patterns are data-agnostic and may not be optimal. Hence, it is likely that relevant keys or values are being reduced, while less important ones are still preserved. Based on this key insight, we propose a novel deformable audio Transformer for audio recognition, named DATAR, where a deformable attention equipping with a pyramid transformer backbone is constructed and learnable. Such an architecture has been proven effective in prediction tasks,~\textit{e.g.}, event classification. Moreover, we identify that the deformable attention map computation may over-simplify the input feature, which can be further enhanced. Hence, we introduce a learnable input adaptor to alleviate this issue, and DATAR achieves state-of-the-art performance.

Runqi Wang, Xiaoyue Duan, Baochang Zhang et al.

A human's attention can intuitively adapt to corrupted areas of an image by recalling a similar uncorrupted image they have previously seen. This observation motivates us to improve the attention of adversarial images by considering their clean counterparts. To accomplish this, we introduce Associative Adversarial Learning (AAL) into adversarial learning to guide a selective attack. We formulate the intrinsic relationship between attention and attack (perturbation) as a coupling optimization problem to improve their interaction. This leads to an attention backtracking algorithm that can effectively enhance the attention's adversarial robustness. Our method is generic and can be used to address a variety of tasks by simply choosing different kernels for the associative attention that select other regions for a specific attack. Experimental results show that the selective attack improves the model's performance. We show that our method improves the recognition accuracy of adversarial training on ImageNet by 8.32% compared with the baseline. It also increases object detection mAP on PascalVOC by 2.02% and recognition accuracy of few-shot learning on miniImageNet by 1.63%.

Shuwei Shao, Ran Li, Zhongcai Pei et al.

Depth estimation attracts widespread attention in the computer vision community. However, it is still quite difficult to recover an accurate depth map using only one RGB image. We observe a phenomenon that existing methods tend to fail in different cases, caused by differences in network architecture, loss function and so on. In this work, we investigate into the phenomenon and propose to integrate the strengths of multiple weak depth predictor to build a comprehensive and accurate depth predictor, which is critical for many real-world applications, e.g., 3D reconstruction. Specifically, we construct multiple base (weak) depth predictors by utilizing different Transformer-based and convolutional neural network (CNN)-based architectures. Transformer establishes long-range correlation while CNN preserves local information ignored by Transformer due to the spatial inductive bias. Therefore, the coupling of Transformer and CNN contributes to the generation of complementary depth estimates, which are essential to achieve a comprehensive depth predictor. Then, we design mixers to learn from multiple weak predictions and adaptively fuse them into a strong depth estimate. The resultant model, which we refer to as Transformer-assisted depth ensembles (TEDepth). On the standard NYU-Depth-v2 and KITTI datasets, we thoroughly explore how the neural ensembles affect the depth estimation and demonstrate that our TEDepth achieves better results than previous state-of-the-art approaches. To validate the generalizability across cameras, we directly apply the models trained on NYU-Depth-v2 to the SUN RGB-D dataset without any fine-tuning, and the superior results emphasize its strong generalizability.

Holger R. Roth, Dong Yang, Wenqi Li et al.

Building robust deep learning-based models requires diverse training data, ideally from several sources. However, these datasets cannot be combined easily because of patient privacy concerns or regulatory hurdles, especially if medical data is involved. Federated learning (FL) is a way to train machine learning models without the need for centralized datasets. Each FL client trains on their local data while only sharing model parameters with a global server that aggregates the parameters from all clients. At the same time, each client's data can exhibit differences and inconsistencies due to the local variation in the patient population, imaging equipment, and acquisition protocols. Hence, the federated learned models should be able to adapt to the local particularities of a client's data. In this work, we combine FL with an AutoML technique based on local neural architecture search by training a "supernet". Furthermore, we propose an adaptation scheme to allow for personalized model architectures at each FL client's site. The proposed method is evaluated on four different datasets from 3D prostate MRI and shown to improve the local models' performance after adaptation through selecting an optimal path through the AutoML supernet.

Dong Yang, Ziyue Xu, Wenqi Li et al.

The recent outbreak of COVID-19 has led to urgent needs for reliable diagnosis and management of SARS-CoV-2 infection. As a complimentary tool, chest CT has been shown to be able to reveal visual patterns characteristic for COVID-19, which has definite value at several stages during the disease course. To facilitate CT analysis, recent efforts have focused on computer-aided characterization and diagnosis, which has shown promising results. However, domain shift of data across clinical data centers poses a serious challenge when deploying learning-based models. In this work, we attempt to find a solution for this challenge via federated and semi-supervised learning. A multi-national database consisting of 1704 scans from three countries is adopted to study the performance gap, when training a model with one dataset and applying it to another. Expert radiologists manually delineated 945 scans for COVID-19 findings. In handling the variability in both the data and annotations, a novel federated semi-supervised learning technique is proposed to fully utilize all available data (with or without annotations). Federated learning avoids the need for sensitive data-sharing, which makes it favorable for institutions and nations with strict regulatory policy on data privacy. Moreover, semi-supervision potentially reduces the annotation burden under a distributed setting. The proposed framework is shown to be effective compared to fully supervised scenarios with conventional data sharing instead of model weight sharing.

Yufang Huang, Wentao Zhu, Deyi Xiong et al.

Unsupervised text style transfer is full of challenges due to the lack of parallel data and difficulties in content preservation. In this paper, we propose a novel neural approach to unsupervised text style transfer, which we refer to as Cycle-consistent Adversarial autoEncoders (CAE) trained from non-parallel data. CAE consists of three essential components: (1) LSTM autoencoders that encode a text in one style into its latent representation and decode an encoded representation into its original text or a transferred representation into a style-transferred text, (2) adversarial style transfer networks that use an adversarially trained generator to transform a latent representation in one style into a representation in another style, and (3) a cycle-consistent constraint that enhances the capacity of the adversarial style transfer networks in content preservation. The entire CAE with these three components can be trained end-to-end. Extensive experiments and in-depth analyses on two widely-used public datasets consistently validate the effectiveness of proposed CAE in both style transfer and content preservation against several strong baselines in terms of four automatic evaluation metrics and human evaluation.

Liyue Shen, Wentao Zhu, Xiaosong Wang et al.

Multi-domain data are widely leveraged in vision applications taking advantage of complementary information from different modalities, e.g., brain tumor segmentation from multi-parametric magnetic resonance imaging (MRI). However, due to possible data corruption and different imaging protocols, the availability of images for each domain could vary amongst multiple data sources in practice, which makes it challenging to build a universal model with a varied set of input data. To tackle this problem, we propose a general approach to complete the random missing domain(s) data in real applications. Specifically, we develop a novel multi-domain image completion method that utilizes a generative adversarial network (GAN) with a representational disentanglement scheme to extract shared skeleton encoding and separate flesh encoding across multiple domains. We further illustrate that the learned representation in multi-domain image completion could be leveraged for high-level tasks, e.g., segmentation, by introducing a unified framework consisting of image completion and segmentation with a shared content encoder. The experiments demonstrate consistent performance improvement on three datasets for brain tumor segmentation, prostate segmentation, and facial expression image completion respectively.

Zhuoqian Yang, Wentao Zhu, Wayne Wu et al.

We present a lightweight video motion retargeting approach TransMoMo that is capable of transferring motion of a person in a source video realistically to another video of a target person. Without using any paired data for supervision, the proposed method can be trained in an unsupervised manner by exploiting invariance properties of three orthogonal factors of variation including motion, structure, and view-angle. Specifically, with loss functions carefully derived based on invariance, we train an auto-encoder to disentangle the latent representations of such factors given the source and target video clips. This allows us to selectively transfer motion extracted from the source video seamlessly to the target video in spite of structural and view-angle disparities between the source and the target. The relaxed assumption of paired data allows our method to be trained on a vast amount of videos needless of manual annotation of source-target pairing, leading to improved robustness against large structural variations and extreme motion in videos. We demonstrate the effectiveness of our method over the state-of-the-art methods. Code, model and data are publicly available on our project page (https://yzhq97.github.io/transmomo).

Wenqi Li, Fausto Milletarì, Daguang Xu et al.

Due to medical data privacy regulations, it is often infeasible to collect and share patient data in a centralised data lake. This poses challenges for training machine learning algorithms, such as deep convolutional networks, which often require large numbers of diverse training examples. Federated learning sidesteps this difficulty by bringing code to the patient data owners and only sharing intermediate model training updates among them. Although a high-accuracy model could be achieved by appropriately aggregating these model updates, the model shared could indirectly leak the local training examples. In this paper, we investigate the feasibility of applying differential-privacy techniques to protect the patient data in a federated learning setup. We implement and evaluate practical federated learning systems for brain tumour segmentation on the BraTS dataset. The experimental results show that there is a trade-off between model performance and privacy protection costs.

Qing Han, Wentao Zhu, Yang Shi

Today, detection of anomalous events in civil infrastructures (e.g. water pipe breaks and leaks) is time consuming and often takes hours or days. Pipe breakage as one of the most frequent types of failure of water networks often causes community disruptions ranging from temporary interruptions in services to extended loss of business and relocation of residents. In this project, we design and implement a two-phase approach for leak event identification, which leverages dynamic data from multiple information sources including IoT sensing data (pressure values and/or flow rates), geophysical data (water systems), and human inputs (tweets posted on Twitter). In the approach, a high order Conditional Random Field (CRF) is constructed that enforces predictions based on IoT observations consistent with human inputs to improve the performance of event identifications. Considering the physical water network as a graph, a CRF model is built and learned by the Structured Support Vector Machine (SSVM) using node features such as water pressure and flow rate. After that, we built the high order CRF system by enforcing twitter leakage detection information. An optimal inference algorithm is proposed for the adapted high order CRF model. Experimental results show the effectiveness of our system.

Wentao Zhu, Jun Miao, Laiyun Qing

Extreme learning machine (ELM) is an extremely fast learning method and has a powerful performance for pattern recognition tasks proven by enormous researches and engineers. However, its good generalization ability is built on large numbers of hidden neurons, which is not beneficial to real time response in the test process. In this paper, we proposed new ways, named "constrained extreme learning machines" (CELMs), to randomly select hidden neurons based on sample distribution. Compared to completely random selection of hidden nodes in ELM, the CELMs randomly select hidden nodes from the constrained vector space containing some basic combinations of original sample vectors. The experimental results show that the CELMs have better generalization ability than traditional ELM, SVM and some other related methods. Additionally, the CELMs have a similar fast learning speed as ELM.