arXiv subjects

Ke Ma

Publications and source records attributed to Ke Ma.

At least 109 records · Page 6Linked to original sources

arXiv2023

Precise Location Matching Improves Dense Contrastive Learning in Digital Pathology

Dense prediction tasks such as segmentation and detection of pathological entities hold crucial clinical value in computational pathology workflows. However, obtaining dense annotations on large cohorts is usually tedious and expensive. Contrastive learning (CL) is thus often employed to leverage large volumes of unlabeled data to pre-train the backbone network. To boost CL for dense prediction, some studies have proposed variations of dense matching objectives in pre-training. However, our analysis shows that employing existing dense matching strategies on histopathology images enforces invariance among incorrect pairs of dense features and, thus, is imprecise. To address this, we propose a precise location-based matching mechanism that utilizes the overlapping information between geometric transformations to precisely match regions in two augmentations. Extensive experiments on two pretraining datasets (TCGA-BRCA, NCT-CRC-HE) and three downstream datasets (GlaS, CRAG, BCSS) highlight the superiority of our method in semantic and instance segmentation tasks. Our method outperforms previous dense matching methods by up to 7.2% in average precision for detection and 5.6% in average precision for instance segmentation tasks. Additionally, by using our matching mechanism in the three popular contrastive learning frameworks, MoCo-v2, VICRegL, and ConCL, the average precision in detection is improved by 0.7% to 5.2%, and the average precision in segmentation is improved by 0.7% to 4.0%, demonstrating generalizability. Our code is available at https://github.com/cvlab-stonybrook/PLM_SSL.

cs.CV↗

arXiv2023

MAMMOTH-Subaru III. Ly$α$ Halo Extended to $\sim200$ kpc Identified by Stacking $\sim 3300$ Ly$α$ Emitters at $z=2.2-2.3$

In this paper, we present a Ly$α$ halo extended to $\sim200$ kpc identified by stacking $\sim 3300$ Ly$α$ emitters at $z=2.2-2.3$. We carry out imaging observations and data reduction with Subaru/Hyper Suprime-Cam (HSC). Our total survey area is $\sim12$ deg$^2$ and imaging depths are $25.5-27.0$ mag. Using the imaging data, we select 1,240 and 2,101 LAE candidates at $z=2.2$ and 2.3, respectively. We carry out spectroscopic observations of our LAE candidates and data reduction with Magellan/IMACS to estimate the contamination rate of our LAE candidates. We find that the contamination rate of our sample is low (8%). We stack our LAE candidates with a median stacking method to identify the Ly$α$ halo at $z=2$. We show that the Ly$α$ halo is extended to $\sim200$ kpc at a surface brightness level of $10^{-20}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$. Comparing to previous studies, our Ly$α$ halo is more extended at radii of $\sim25-100$ kpc, which is not likely caused by the contamination in our sample but by different redshifts and fields instead. To investigate how central galaxies affect surrounding LAHs, we divide our LAEs into subsamples based on the Ly$α$ luminosity ($L_{\rm Lyα}$), rest-frame Ly$α$ equivalent width (EW$_0$), and UV magnitude (M$_{\rm uv}$). We stack the subsamples and find that higher $L_{\rm Lyα}$, lower EW$_0$, and brighter M$_{\rm uv}$ cause more extended halos. Our results suggest that more massive LAEs generally have more extended Ly$α$ halos.

astro-ph.GA↗

arXiv2023

MAMMOTH-Subaru V. Effects of Cosmic Variance on Ly$α$ Luminosity Functions at $z=2.2-2.3$

Cosmic variance introduces significant uncertainties into galaxy number density properties when surveying the high-z Universe with a small volume, such uncertainties produce the field-to-field variance of galaxy number $σ_{g}$ in observational astronomy. This uncertainty significantly affects the Luminosity Functions (LF) measurement of Lya Emitters (LAEs). For most previous Lya LF studies, $σ_{g}$ is often adopted from predictions by cosmological simulations, but barely confirmed by observations. Measuring cosmic variance requires a huge sample over a large volume, exceeding the capabilities of most astronomical instruments. In this study, we demonstrate an observational approach for measuring the cosmic variance contribution for $z\approx2.2$ Lya LFs. The LAE candidates are observed using narrowband and broadband of the Subaru/Hyper Suprime-Cam (HSC), with 8 independent fields, making the total survey area $\simeq11.62$deg$^2$ and a comoving volume of $\simeq8.71\times10^6$Mpc$^3$. These eight fields are selected using the project of MAMMOTH. We report a best-fit Schechter function with parameters $α=-1.75$ (fixed), $L_{Lyα}^{*}=5.18_{-0.40}^{+0.43} \times 10^{42}$erg s$^{-1}$ and $ϕ_{Lya}^{*}=4.87_{-0.55}^{+0.54}\times10^{-4}$Mpc$^{-3}$ for the overall Lya LFs. After clipping out the regions that can bias the cosmic variance measurements, we calculate $σ_{g}$, by sampling LAEs within multiple pointings assigned on the field image. We investigate the relation between $σ_{g}$ and survey volume $V$, and fit a simple power law: $σ_g=k\times(\frac{V_{\rm eff}}{10^5 {\rm Mpc}^3})^β$. We find best-fit values of $-1.209_{-0.106}^{+0.106}$ for $β$ and $0.986_{-0.100}^{+0.108}$ for k. We compare our measurements with predictions from simulations and find that the cosmic variance of LAEs might be larger than that of general star-forming galaxies.

astro-ph.GA↗

arXiv2022

OMSN and FAROS: OCTA Microstructure Segmentation Network and Fully Annotated Retinal OCTA Segmentation Dataset

The lack of efficient segmentation methods and fully-labeled datasets limits the comprehensive assessment of optical coherence tomography angiography (OCTA) microstructures like retinal vessel network (RVN) and foveal avascular zone (FAZ), which are of great value in ophthalmic and systematic diseases evaluation. Here, we introduce an innovative OCTA microstructure segmentation network (OMSN) by combining an encoder-decoder-based architecture with multi-scale skip connections and the split-attention-based residual network ResNeSt, paying specific attention to OCTA microstructural features while facilitating better model convergence and feature representations. The proposed OMSN achieves excellent single/multi-task performances for RVN or/and FAZ segmentation. Especially, the evaluation metrics on multi-task models outperform single-task models on the same dataset. On this basis, a fully annotated retinal OCTA segmentation (FAROS) dataset is constructed semi-automatically, filling the vacancy of a pixel-level fully-labeled OCTA dataset. OMSN multi-task segmentation model retrained with FAROS further certifies its outstanding accuracy for simultaneous RVN and FAZ segmentation.

eess.IV↗

arXiv2022

Gigapixel Whole-Slide Images Classification using Locally Supervised Learning

Histopathology whole slide images (WSIs) play a very important role in clinical studies and serve as the gold standard for many cancer diagnoses. However, generating automatic tools for processing WSIs is challenging due to their enormous sizes. Currently, to deal with this issue, conventional methods rely on a multiple instance learning (MIL) strategy to process a WSI at patch level. Although effective, such methods are computationally expensive, because tiling a WSI into patches takes time and does not explore the spatial relations between these tiles. To tackle these limitations, we propose a locally supervised learning framework which processes the entire slide by exploring the entire local and global information that it contains. This framework divides a pre-trained network into several modules and optimizes each module locally using an auxiliary model. We also introduce a random feature reconstruction unit (RFR) to preserve distinguishing features during training and improve the performance of our method by 1% to 3%. Extensive experiments on three publicly available WSI datasets: TCGA-NSCLC, TCGA-RCC and LKS, highlight the superiority of our method on different classification tasks. Our method outperforms the state-of-the-art MIL methods by 2% to 5% in accuracy, while being 7 to 10 times faster. Additionally, when dividing it into eight modules, our method requires as little as 20% of the total gpu memory required by end-to-end training. Our code is available at https://github.com/cvlab-stonybrook/local_learning_wsi.

cs.CV↗

arXiv2022

A Tale of HodgeRank and Spectral Method: Target Attack Against Rank Aggregation Is the Fixed Point of Adversarial Game

Rank aggregation with pairwise comparisons has shown promising results in elections, sports competitions, recommendations, and information retrieval. However, little attention has been paid to the security issue of such algorithms, in contrast to numerous research work on the computational and statistical characteristics. Driven by huge profits, the potential adversary has strong motivation and incentives to manipulate the ranking list. Meanwhile, the intrinsic vulnerability of the rank aggregation methods is not well studied in the literature. To fully understand the possible risks, we focus on the purposeful adversary who desires to designate the aggregated results by modifying the pairwise data in this paper. From the perspective of the dynamical system, the attack behavior with a target ranking list is a fixed point belonging to the composition of the adversary and the victim. To perform the targeted attack, we formulate the interaction between the adversary and the victim as a game-theoretic framework consisting of two continuous operators while Nash equilibrium is established. Then two procedures against HodgeRank and RankCentrality are constructed to produce the modification of the original data. Furthermore, we prove that the victims will produce the target ranking list once the adversary masters the complete information. It is noteworthy that the proposed methods allow the adversary only to hold incomplete information or imperfect feedback and perform the purposeful attack. The effectiveness of the suggested target attack strategies is demonstrated by a series of toy simulations and several real-world data experiments. These experimental results show that the proposed methods could achieve the attacker's goal in the sense that the leading candidate of the perturbed ranking list is the designated one by the adversary.

cs.LG↗

arXiv2022

Optimal, centralized dynamic curbside parking space zoning

In this paper we formulate a dynamic mixed integer program for optimally zoning curbside parking spaces subject to transportation policy-inspired constraints and regularization terms. First, we illustrate how given some objective of curb zoning valuation as a function of zone type (e.g., paid parking or bus stop), dynamically rezoning involves unrolling this optimization program over a fixed time horizon. Second, we implement two different solution methods that optimize for a given curb zoning value function. In the first method, we solve long horizon dynamic zoning problems via approximate dynamic programming. In the second method, we employ Dantzig-Wolfe decomposition to break-up the mixed-integer program into a master problem and several sub-problems that we solve in parallel; this decomposition accelerates the MIP solver considerably. We present simulation results and comparisons of the different employed techniques on vehicle arrival-rate data obtained for a neighborhood in downtown Seattle, Washington, USA

math.OC↗

arXiv2022

Prior-Guided Adversarial Initialization for Fast Adversarial Training

Fast adversarial training (FAT) effectively improves the efficiency of standard adversarial training (SAT). However, initial FAT encounters catastrophic overfitting, i.e.,the robust accuracy against adversarial attacks suddenly and dramatically decreases. Though several FAT variants spare no effort to prevent overfitting, they sacrifice much calculation cost. In this paper, we explore the difference between the training processes of SAT and FAT and observe that the attack success rate of adversarial examples (AEs) of FAT gets worse gradually in the late training stage, resulting in overfitting. The AEs are generated by the fast gradient sign method (FGSM) with a zero or random initialization. Based on the observation, we propose a prior-guided FGSM initialization method to avoid overfitting after investigating several initialization strategies, improving the quality of the AEs during the whole training process. The initialization is formed by leveraging historically generated AEs without additional calculation cost. We further provide a theoretical analysis for the proposed initialization method. We also propose a simple yet effective regularizer based on the prior-guided initialization,i.e., the currently generated perturbation should not deviate too much from the prior-guided initialization. The regularizer adopts both historical and current adversarial perturbations to guide the model learning. Evaluations on four datasets demonstrate that the proposed method can prevent catastrophic overfitting and outperform state-of-the-art FAT methods. The code is released at https://github.com/jiaxiaojunQAQ/FGSM-PGI.

cs.CV↗

arXiv2022

Visual attention analysis of pathologists examining whole slide images of Prostate cancer

We study the attention of pathologists as they examine whole-slide images (WSIs) of prostate cancer tissue using a digital microscope. To the best of our knowledge, our study is the first to report in detail how pathologists navigate WSIs of prostate cancer as they accumulate information for their diagnoses. We collected slide navigation data (i.e., viewport location, magnification level, and time) from 13 pathologists in 2 groups (5 genitourinary (GU) specialists and 8 general pathologists) and generated visual attention heatmaps and scanpaths. Each pathologist examined five WSIs from the TCGA PRAD dataset, which were selected by a GU pathology specialist. We examined and analyzed the distributions of visual attention for each group of pathologists after each WSI was examined. To quantify the relationship between a pathologist's attention and evidence for cancer in the WSI, we obtained tumor annotations from a genitourinary specialist. We used these annotations to compute the overlap between the distribution of visual attention and annotated tumor region to identify strong correlations. Motivated by this analysis, we trained a deep learning model to predict visual attention on unseen WSIs. We find that the attention heatmaps predicted by our model correlate quite well with the ground truth attention heatmap and tumor annotations on a test set of 17 WSIs by using various spatial and temporal evaluation metrics.

eess.IV↗

arXiv2022

LAS-AT: Adversarial Training with Learnable Attack Strategy

Adversarial training (AT) is always formulated as a minimax problem, of which the performance depends on the inner optimization that involves the generation of adversarial examples (AEs). Most previous methods adopt Projected Gradient Decent (PGD) with manually specifying attack parameters for AE generation. A combination of the attack parameters can be referred to as an attack strategy. Several works have revealed that using a fixed attack strategy to generate AEs during the whole training phase limits the model robustness and propose to exploit different attack strategies at different training stages to improve robustness. But those multi-stage hand-crafted attack strategies need much domain expertise, and the robustness improvement is limited. In this paper, we propose a novel framework for adversarial training by introducing the concept of "learnable attack strategy", dubbed LAS-AT, which learns to automatically produce attack strategies to improve the model robustness. Our framework is composed of a target network that uses AEs for training to improve robustness and a strategy network that produces attack strategies to control the AE generation. Experimental evaluations on three benchmark databases demonstrate the superiority of the proposed method. The code is released at https://github.com/jiaxiaojunQAQ/LAS-AT.

cs.CV↗

arXiv2021

Deep Learning for Beam-Management: State-of-the-Art, Opportunities and Challenges

Benefiting from huge bandwidth resources, millimeter-wave (mmWave) communications provide one of the most promising technologies for next-generation wireless networks. To compensate for the high pathloss of mmWave signals, large-scale antenna arrays are required both at the base stations and user equipment to establish directional beamforming, where beam-management is adopted to acquire and track the optimal beam pair having the maximum received power. Naturally, narrow beams are required for achieving high beamforming gain, but they impose enormous training overhead and high sensitivity to blockages. As a remedy, deep learning (DL) may be harnessed for beam-management. First, the current state-of-the-art is reviewed, followed by the associated challenges and future research opportunities. We conclude by highlighting the associated DL design insights and novel beam-management mechanisms.

cs.IT↗

arXiv2021

Advancing COVID-19 Diagnosis with Privacy-Preserving Collaboration in Artificial Intelligence

Artificial intelligence (AI) provides a promising substitution for streamlining COVID-19 diagnoses. However, concerns surrounding security and trustworthiness impede the collection of large-scale representative medical data, posing a considerable challenge for training a well-generalised model in clinical practices. To address this, we launch the Unified CT-COVID AI Diagnostic Initiative (UCADI), where the AI model can be distributedly trained and independently executed at each host institution under a federated learning framework (FL) without data sharing. Here we show that our FL model outperformed all the local models by a large yield (test sensitivity /specificity in China: 0.973/0.951, in the UK: 0.730/0.942), achieving comparable performance with a panel of professional radiologists. We further evaluated the model on the hold-out (collected from another two hospitals leaving out the FL) and heterogeneous (acquired with contrast materials) data, provided visual explanations for decisions made by the model, and analysed the trade-offs between the model performance and the communication costs in the federated training process. Our study is based on 9,573 chest computed tomography scans (CTs) from 3,336 patients collected from 23 hospitals located in China and the UK. Collectively, our work advanced the prospects of utilising federated learning for privacy-preserving AI in digital health.

cs.AI↗

arXiv2021

Transactive Energy System Deployment over Insecure Communication Links

In this paper, the privacy and security issues associated with the transactive energy system (TES) deployment over insecure communication links are addressed. In particular, it is ensured that (1) individual agents' bidding information is kept private throughout hierarchical market-based interactions; and (2) any extraneous data injection attack can be quickly and easily detected. An implementation framework is proposed to enable the cryptography-based enhancement of privacy and security for the deployment of any general hierarchical systems including TESs. Under the proposed framework, a unified cryptography-based approach is developed to achieve both privacy and security simultaneously. Specifically, privacy preservation is realized by an enhanced Paillier encryption scheme, where a block design is proposed to significantly improve computational efficiency. Attack detection is further achieved by an enhanced Paillier digital signature scheme, where a stamp-concatenation mechanism is proposed to enable detection of data replace and reorder attacks. Simulation results verify the effectiveness of the proposed cyber-resilient design for transactive energy systems.

cs.CR↗

arXiv2021

Deep Learning Assisted Calibrated Beam Training for Millimeter-Wave Communication Systems

Huge overhead of beam training imposes a significant challenge in millimeter-wave (mmWave) wireless communications. To address this issue, in this paper, we propose a wide beam based training approach to calibrate the narrow beam direction according to the channel power leakage. To handle the complex nonlinear properties of the channel power leakage, deep learning is utilized to predict the optimal narrow beam directly. Specifically, three deep learning assisted calibrated beam training schemes are proposed. The first scheme adopts convolution neural network to implement the prediction based on the instantaneous received signals of wide beam training. We also perform the additional narrow beam training based on the predicted probabilities for further beam direction calibrations. However, the first scheme only depends on one wide beam training, which lacks the robustness to noise. To tackle this problem, the second scheme adopts long-short term memory (LSTM) network for tracking the movement of users and calibrating the beam direction according to the received signals of prior beam training, in order to enhance the robustness to noise. To further reduce the overhead of wide beam training, our third scheme, an adaptive beam training strategy, selects partial wide beams to be trained based on the prior received signals. Two criteria, namely, optimal neighboring criterion and maximum probability criterion, are designed for the selection. Furthermore, to handle mobile scenarios, auxiliary LSTM is introduced to calibrate the directions of the selected wide beams more precisely. Simulation results demonstrate that our proposed schemes achieve significantly higher beamforming gain with smaller beam training overhead compared with the conventional and existing deep-learning based counterparts.

eess.SP↗

arXiv2021

Poisoning Attack against Estimating from Pairwise Comparisons

As pairwise ranking becomes broadly employed for elections, sports competitions, recommendations, and so on, attackers have strong motivation and incentives to manipulate the ranking list. They could inject malicious comparisons into the training data to fool the victim. Such a technique is called poisoning attack in regression and classification tasks. In this paper, to the best of our knowledge, we initiate the first systematic investigation of data poisoning attacks on pairwise ranking algorithms, which can be formalized as the dynamic and static games between the ranker and the attacker and can be modeled as certain kinds of integer programming problems. To break the computational hurdle of the underlying integer programming problems, we reformulate them into the distributionally robust optimization (DRO) problems, which are computationally tractable. Based on such DRO formulations, we propose two efficient poisoning attack algorithms and establish the associated theoretical guarantees. The effectiveness of the suggested poisoning attack strategies is demonstrated by a series of toy simulations and several real data experiments. These experimental results show that the proposed methods can significantly reduce the performance of the ranker in the sense that the correlation between the true ranking list and the aggregated results can be decreased dramatically.

cs.LG↗

arXiv2021

HAIR: Head-mounted AR Intention Recognition

Human teams exhibit both implicit and explicit intention sharing. To further development of human-robot collaboration, intention recognition is crucial on both sides. Present approaches rely on a vast sensor suite on and around the robot to achieve intention recognition. This relegates intuitive human-robot collaboration purely to such bulky systems, which are inadequate for large-scale, real-world scenarios due to their complexity and cost. In this paper we propose an intention recognition system that is based purely on a portable head-mounted display. In addition robot intention visualisation is also supported. We present experiments to show the quality of our human goal estimation component and some basic interactions with an industrial robot. HAIR should raise the quality of interaction between robots and humans, instead of such interactions raising the hair on the necks of the human coworkers.

cs.RO↗

arXiv2021

Deep Learning Assisted mmWave Beam Prediction with Prior Low-frequency Information

Huge overhead of beam training poses a significant challenge to mmWave communications. To address this issue, beam tracking has been widely investigated whereas existing methods are hard to handle serious multipath interference and non-stationary scenarios. Inspired by the spatial similarity between low-frequency and mmWave channels in non-standalone architectures, this paper proposes to utilize prior low-frequency information to predict the optimal mmWave beam, where deep learning is adopted to enhance the prediction accuracy. Specifically, periodically estimated low-frequency channel state information (CSI) is applied to track the movement of user equipment, and timing offset indicator is proposed to indicate the instant of mmWave beam training relative to low-frequency CSI estimation. Meanwhile, long-short term memory networks based dedicated models are designed to implement the prediction. Simulation results show that our proposed scheme can achieve higher beamforming gain than the conventional methods while requiring little overhead of mmWave beam training.

eess.SP↗

arXiv2021

AdaSpring: Context-adaptive and Runtime-evolutionary Deep Model Compression for Mobile Applications

There are many deep learning (e.g., DNN) powered mobile and wearable applications today continuously and unobtrusively sensing the ambient surroundings to enhance all aspects of human lives. To enable robust and private mobile sensing, DNN tends to be deployed locally on the resource-constrained mobile devices via model compression. The current practice either hand-crafted DNN compression techniques, i.e., for optimizing DNN-relative performance (e.g., parameter size), or on-demand DNN compression methods, i.e., for optimizing hardware-dependent metrics (e.g., latency), cannot be locally online because they require offline retraining to ensure accuracy. Also, none of them have correlated their efforts with runtime adaptive compression to consider the dynamic nature of the deployment context of mobile applications. To address those challenges, we present AdaSpring, a context-adaptive and self-evolutionary DNN compression framework. It enables the runtime adaptive DNN compression locally online. Specifically, it presents the ensemble training of a retraining-free and self-evolutionary network to integrate multiple alternative DNN compression configurations (i.e., compressed architectures and weights). It then introduces the runtime search strategy to quickly search for the most suitable compression configurations and evolve the corresponding weights. With evaluation on five tasks across three platforms and a real-world case study, experiment outcomes show that AdaSpring obtains up to 3.1x latency reduction, 4.2 x energy efficiency improvement in DNNs, compared to hand-crafted compression techniques, while only incurring <= 6.2ms runtime-evolution latency.

cs.LG↗