arXiv subjects

Weifeng Su

Publications and source records attributed to Weifeng Su.

25 records · Page 2Linked to original sources

arXiv2024

Leveraging CORAL-Correlation Consistency Network for Semi-Supervised Left Atrium MRI Segmentation

Semi-supervised learning (SSL) has been widely used to learn from both a few labeled images and many unlabeled images to overcome the scarcity of labeled samples in medical image segmentation. Most current SSL-based segmentation methods use pixel values directly to identify similar features in labeled and unlabeled data. They usually fail to accurately capture the intricate attachment structures in the left atrium, such as the areas of inconsistent density or exhibit outward curvatures, adding to the complexity of the task. In this paper, we delve into this issue and introduce an effective solution, CORAL(Correlation-Aligned)-Correlation Consistency Network (CORN), to capture the global structure shape and local details of Left Atrium. Diverging from previous methods focused on each local pixel value, the CORAL-Correlation Consistency Module (CCM) in the CORN leverages second-order statistical information to capture global structural features by minimizing the distribution discrepancy between labeled and unlabeled samples in feature space. Yet, direct construction of features from unlabeled data frequently results in ``Sample Selection Bias'', leading to flawed supervision. We thus further propose the Dynamic Feature Pool (DFP) for the CCM, which utilizes a confidence-based filtering strategy to remove incorrectly selected features and regularize both teacher and student models by constraining the similarity matrix to be consistent. Extensive experiments on the Left Atrium dataset have shown that the proposed CORN outperforms previous state-of-the-art semi-supervised learning methods.

cs.CV↗

arXiv2024

M$^3$-Impute: Mask-guided Representation Learning for Missing Value Imputation

Missing values are a common problem that poses significant challenges to data analysis and machine learning. This problem necessitates the development of an effective imputation method to fill in the missing values accurately, thereby enhancing the overall quality and utility of the datasets. Existing imputation methods, however, fall short of explicitly considering the `missingness' information in the data during the embedding initialization stage and modeling the entangled feature and sample correlations during the learning process, thus leading to inferior performance. We propose M$^3$-Impute, which aims to explicitly leverage the missingness information and such correlations with novel masking schemes. M$^3$-Impute first models the data as a bipartite graph and uses a graph neural network to learn node embeddings, where the refined embedding initialization process directly incorporates the missingness information. They are then optimized through M$^3$-Impute's novel feature correlation unit (FRU) and sample correlation unit (SRU) that effectively captures feature and sample correlations for imputation. Experiment results on 25 benchmark datasets under three different missingness settings show the effectiveness of M$^3$-Impute by achieving 20 best and 4 second-best MAE scores on average.

cs.LG↗

arXiv2023

Towards Cooperative Federated Learning over Heterogeneous Edge/Fog Networks

Federated learning (FL) has been promoted as a popular technique for training machine learning (ML) models over edge/fog networks. Traditional implementations of FL have largely neglected the potential for inter-network cooperation, treating edge/fog devices and other infrastructure participating in ML as separate processing elements. Consequently, FL has been vulnerable to several dimensions of network heterogeneity, such as varying computation capabilities, communication resources, data qualities, and privacy demands. We advocate for cooperative federated learning (CFL), a cooperative edge/fog ML paradigm built on device-to-device (D2D) and device-to-server (D2S) interactions. Through D2D and D2S cooperation, CFL counteracts network heterogeneity in edge/fog networks through enabling a model/data/resource pooling mechanism, which will yield substantial improvements in ML model training quality and network resource consumption. We propose a set of core methodologies that form the foundation of D2D and D2S cooperation and present preliminary experiments that demonstrate their benefits. We also discuss new FL functionalities enabled by this cooperative framework such as the integration of unlabeled data and heterogeneous device privacy into ML model training. Finally, we describe some open research directions at the intersection of cooperative edge/fog and FL.

cs.DC↗

arXiv2021

Retrospective Interference Regeneration Schemes for Relay-Aided K-user MIMO Downlink Networks

To accommodate the explosive growth of the Internet-of-Things (IoT), incorporating interference alignment (IA) into existing multiple access (MA) schemes is under investigation. However, when it is applied in MIMO networks to improve the system compacity, the incoming problem regarding information delay arises which does not meet the requirement of low-latency. Therefore, in this paper, we first propose a new metric, degree of delay (DoD), to quantify the issue of information delay, and characterize DoD for three typical transmission schemes, i.e., TDMA, beamforming based TDMA (BD-TDMA), and retrospective interference alignment (RIA). By analyzing DoD in these schemes, its value mainly depends on three factors, i.e., delay sensitive factor, size of data set, and queueing delay slot. The first two reflect the relationship between quality of service (QoS) and information delay sensitivity, and normalize time cost for each symbol, respectively. These two factors are independent of the transmission schemes, and thus we aim to reduce the queueing delay slot to improve DoD. Herein, three novel joint IA schemes are proposed for MIMO downlink networks with different number of users. That is, hybrid antenna array based partial interference elimination and retrospective interference regeneration scheme (HAA-PIE-RIR), HAA based improved PIE and RIR scheme (HAA-IPIE-RIR), and HAA based cyclic interference elimination and RIR scheme (HAA-CIE-RIR). Based on the first scheme, the second scheme extends the application scenarios from $2$-user to $K$-user while causing heavy computational burden. The third scheme relieves such computational burden, though it has certain degree of freedom (DoF) loss due to insufficient utilization of space resources.

cs.IT↗

arXiv2021

LenAtten: An Effective Length Controlling Unit For Text Summarization

Fixed length summarization aims at generating summaries with a preset number of words or characters. Most recent researches incorporate length information with word embeddings as the input to the recurrent decoding unit, causing a compromise between length controllability and summary quality. In this work, we present an effective length controlling unit Length Attention (LenAtten) to break this trade-off. Experimental results show that LenAtten not only brings improvements in length controllability and ROGUE scores but also has great generalization ability. In the task of generating a summary with the target length, our model is 732 times better than the best-performing length controllable summarizer in length controllability on the CNN/Daily Mail dataset.

cs.CL↗

arXiv2017

Using lab notebooks to examine students' engagement in modeling in an upper-division electronics lab course

We demonstrate how students' use of modeling can be examined and assessed using student notebooks collected from an upper-division electronics lab course. The use of models is a ubiquitous practice in undergraduate physics education, but the process of constructing, testing, and refining these models is much less common. We focus our attention on a lab course that has been transformed to engage students in this modeling process during lab activities. The design of the lab activities was guided by a framework that captures the different components of model-based reasoning, called the Modeling Framework for Experimental Physics. We demonstrate how this framework can be used to assess students' written work and to identify how students' model-based reasoning differed from activity to activity. Broadly speaking, we were able to identify the different steps of students' model-based reasoning and assess the completeness of their reasoning. Varying degrees of scaffolding present across the activities had an impact on how thoroughly students would engage in the full modeling process, with more scaffolded activities resulting in more thorough engagement with the process. Finally, we identified that the step in the process with which students had the most difficulty was the comparison between their interpreted data and their model prediction. Students did not use sufficiently sophisticated criteria in evaluating such comparisons, which had the effect of halting the modeling process. This may indicate that in order to engage students further in using model-based reasoning during lab activities, the instructor needs to provide further scaffolding for how students make these types of experimental comparisons. This is an important design consideration for other such courses attempting to incorporate modeling as a learning goal.

physics.ed-ph↗

arXiv2016

On the Degrees of Freedom of MIMO X Networks with Non-Cooperation Transmitters

Due to limited backhaul/feedback link capacity and channel state information (CSI) feedback delay, obtaining global and instantaneous channel state information at the transmitter (CSIT) is a main obstacle in practice. In this paper, novel transmission schemes are proposed for a class of interference networks that can achieve new trade-off regions between the sum of degrees of freedom (sum-DoF) and CSI feedback delay with distributed and temperately-delayed CSIT. More specifically, a distributed space-time interference alignment (STIA) scheme is proposed for the two-user multiple-input multiple-output (MIMO) X channel via a novel precoding method called Cyclic Zero-padding. The achieved sum-DoFs herein for certain antenna configurations are greater than the best known sum-DoFs in literature with delayed CSIT. Furthermore, we propose a distributed retrospective interference alignment (RIA) scheme that achieves more than 1 sum-DoF for the K-user single-input single-output (SISO) X network. Finally, we extend the distributed STIA to the MxN user multiple-input single-output (MISO) X network where each transmitter has N-1 antennas and each receiver has a single antenna, yielding the same sum-DoF as that in the global and instantaneous CSIT case. The discussion and the result of the MISO X network can be extended to the MIMO case due to spatial scale invariance property.

cs.IT↗