arXiv subjects

Fan Yi

Publications and source records attributed to Fan Yi.

8 recordsLinked to original sources

arXiv2025

Explore and Establish Synergistic Effects Between Weight Pruning and Coreset Selection in Neural Network Training

Modern deep neural networks rely heavily on massive model weights and training samples, incurring substantial computational costs. Weight pruning and coreset selection are two emerging paradigms proposed to improve computational efficiency. In this paper, we first explore the interplay between redundant weights and training samples through a transparent analysis: redundant samples, particularly noisy ones, cause model weights to become unnecessarily overtuned to fit them, complicating the identification of irrelevant weights during pruning; conversely, irrelevant weights tend to overfit noisy data, undermining coreset selection effectiveness. To further investigate and harness this interplay in deep learning, we develop a Simultaneous Weight and Sample Tailoring mechanism (SWaST) that alternately performs weight pruning and coreset selection to establish a synergistic effect in training. During this investigation, we observe that when simultaneously removing a large number of weights and samples, a phenomenon we term critical double-loss can occur, where important weights and their supportive samples are mistakenly eliminated at the same time, leading to model instability and nearly irreversible degradation that cannot be recovered in subsequent training. Unlike classic machine learning models, this issue can arise in deep learning due to the lack of theoretical guarantees on the correctness of weight pruning and coreset selection, which explains why these paradigms are often developed independently. We mitigate this by integrating a state preservation mechanism into SWaST, enabling stable joint optimization. Extensive experiments reveal a strong synergy between pruning and coreset selection across varying prune rates and coreset sizes, delivering accuracy boosts of up to 17.83% alongside 10% to 90% FLOPs reductions.

cs.LG↗

arXiv2025

Automated, Cross-Layer Root Cause Analysis of 5G Video-Conferencing Quality Degradation

5G wireless networks are complex, leveraging layers of scheduling, retransmission, and adaptation mechanisms to maximize their efficiency. But these mechanisms interact to produce significant fluctuations in uplink and downlink capacity and latency. This markedly impacts the performance of real-time applications, such as video-conferencing, which are particularly sensitive to such fluctuations, resulting in lag, stuttering, distorted audio, and low video quality. This paper presents a cross-layer view of 5G networks and their impact on and interaction with video-conferencing applications. We conduct novel, detailed measurements of both Private CBRS and commercial carrier cellular network dynamics, capturing physical- and link-layer events and correlating them with their effects at the network and transport layers, and the video-conferencing application itself. Our two datasets comprise days of low-rate campus-wide Zoom telemetry data, and hours of high-rate, correlated WebRTC-network-5G telemetry data. Based on these data, we trace performance anomalies back to root causes, identifying 24 previously unknown causal event chains that degrade 5G video conferencing. Armed with this knowledge, we build Domino, a tool that automates this process and is user-extensible to future wireless networks and interactive applications.

cs.NI↗

arXiv2025

Phenotype-Guided Generative Model for High-Fidelity Cardiac MRI Synthesis: Advancing Pretraining and Clinical Applications

Cardiac Magnetic Resonance (CMR) imaging is a vital non-invasive tool for diagnosing heart diseases and evaluating cardiac health. However, the limited availability of large-scale, high-quality CMR datasets poses a major challenge to the effective application of artificial intelligence (AI) in this domain. Even the amount of unlabeled data and the health status it covers are difficult to meet the needs of model pretraining, which hinders the performance of AI models on downstream tasks. In this study, we present Cardiac Phenotype-Guided CMR Generation (CPGG), a novel approach for generating diverse CMR data that covers a wide spectrum of cardiac health status. The CPGG framework consists of two stages: in the first stage, a generative model is trained using cardiac phenotypes derived from CMR data; in the second stage, a masked autoregressive diffusion model, conditioned on these phenotypes, generates high-fidelity CMR cine sequences that capture both structural and functional features of the heart in a fine-grained manner. We synthesized a massive amount of CMR to expand the pretraining data. Experimental results show that CPGG generates high-quality synthetic CMR data, significantly improving performance on various downstream tasks, including diagnosis and cardiac phenotypes prediction. These gains are demonstrated across both public and private datasets, highlighting the effectiveness of our approach. Code is availabel at https://anonymous.4open.science/r/CPGG.

cs.CV↗

arXiv2024

Out-of-Distribution Detection using Neural Activation Prior

Out-of-distribution detection (OOD) is a crucial technique for deploying machine learning models in the real world to handle the unseen scenarios. In this paper, we first propose a simple yet effective Neural Activation Prior (NAP) for OOD detection. Our neural activation prior is based on a key observation that, for a channel before the global pooling layer of a fully trained neural network, the probability of a few neurons being activated with a large response by an in-distribution (ID) sample is significantly higher than that by an OOD sample. An intuitive explanation is that for a model fully trained on ID dataset, each channel would play a role in detecting a certain pattern in the ID dataset, and a few neurons can be activated with a large response when the pattern is detected in an input sample. Then, a new scoring function based on this prior is proposed to highlight the role of these strongly activated neurons in OOD detection. Our approach is plug-and-play and does not lead to any performance degradation on ID data classification and requires no extra training or statistics from training or external datasets. Notice that previous methods primarily rely on post-global-pooling features of the neural networks, while the within-channel distribution information we leverage would be discarded by the global pooling operator. Consequently, our method is orthogonal to existing approaches and can be effectively combined with them in various applications. Experimental results show that our method achieves the state-of-the-art performance on CIFAR benchmark and ImageNet dataset, which demonstrates the power of the proposed prior. Finally, we extend our method to Transformers and the experimental findings indicate that NAP can also significantly enhance the performance of OOD detection on Transformers, thereby demonstrating the broad applicability of this prior knowledge.

cs.CV↗

arXiv2023

WaveFlex: A Smart Surface for Private CBRS Wireless Cellular Networks

We present the design and implementation of WaveFlex, the first smart surface that enhances Private LTE/5G networks operating under the shared-license framework in the Citizens Broadband Radio Service frequency band. WaveFlex works in the presence of frequency diversity: multiple nearby base stations operating on different frequencies, as dictated by a Spectrum Access System coordinator. It also handles time dynamism: due to the dynamic sharing rules of the band, base stations occasionally switch channels, especially when priority users enter the network. Finally, WaveFlex operates independently of the network itself, not requiring access to nor modification of the base station or mobile users, yet it remain compliant with and effective on prevailing cellular protocols. We have designed and fabricated WaveFlex on a custom multi-layer PCB, software defined radio-based network monitor, and supporting control software and hardware. Our experimental evaluation benchmarks an operational Private LTE network running at full line rate. Results demonstrate an 8.50 dB average SNR gain, and an average throughput gain of 4.36 Mbps for a single small cell, and 3.19 Mbps for four small cells, in a realistic indoor office scenario.

cs.NI↗

arXiv2023

Multiverse Transformer: 1st Place Solution for Waymo Open Sim Agents Challenge 2023

This technical report presents our 1st place solution for the Waymo Open Sim Agents Challenge (WOSAC) 2023. Our proposed MultiVerse Transformer for Agent simulation (MVTA) effectively leverages transformer-based motion prediction approaches, and is tailored for closed-loop simulation of agents. In order to produce simulations with a high degree of realism, we design novel training and sampling methods, and implement a receding horizon prediction mechanism. In addition, we introduce a variable-length history aggregation method to mitigate the compounding error that can arise during closed-loop autoregressive execution. On the WOSAC, our MVTA and its enhanced version MVTE reach a realism meta-metric of 0.5091 and 0.5168, respectively, outperforming all the other methods on the leaderboard.

cs.CV↗

arXiv2023

AI vs. Human -- Differentiation Analysis of Scientific Content Generation

Recent neural language models have taken a significant step forward in producing remarkably controllable, fluent, and grammatical text. Although studies have found that AI-generated text is not distinguishable from human-written text for crowd-sourcing workers, there still exist errors in AI-generated text which are even subtler and harder to spot. We primarily focus on the scenario in which scientific AI writing assistant is deeply involved. First, we construct a feature description framework to distinguish between AI-generated text and human-written text from syntax, semantics, and pragmatics based on the human evaluation. Then we utilize the features, i.e., writing style, coherence, consistency, and argument logistics, from the proposed framework to analyze two types of content. Finally, we adopt several publicly available methods to investigate the gap of between AI-generated scientific text and human-written scientific text by AI-generated scientific text detection models. The results suggest that while AI has the potential to generate scientific content that is as accurate as human-written content, there is still a gap in terms of depth and overall quality. The AI-generated scientific content is more likely to contain errors in factual issues. We find that there exists a "writing style" gap between AI-generated scientific text and human-written scientific text. Based on the analysis result, we summarize a series of model-agnostic and distribution-agnostic features for detection tasks in other domains. Findings in this paper contribute to guiding the optimization of AI models to produce high-quality content and addressing related ethical and security concerns.

cs.CL↗

arXiv2020

PBE-CC: Congestion Control via Endpoint-Centric, Physical-Layer Bandwidth Measurements

Wireless networks are becoming ever more sophisticated and overcrowded, imposing the most delay, jitter, and throughput damage to end-to-end network flows in today's internet. We therefore argue for fine-grained mobile endpoint-based wireless measurements to inform a precise congestion control algorithm through a well-defined API to the mobile's wireless physical layer. Our proposed congestion control algorithm is based on Physical-Layer Bandwidth measurements taken at the Endpoint (PBE-CC), and captures the latest 5G New Radio innovations that increase wireless capacity, yet create abrupt rises and falls in available wireless capacity that the PBE-CC sender can react to precisely and very rapidly. We implement a proof-of-concept prototype of the PBE measurement module on software-defined radios and the PBE sender and receiver in C. An extensive performance evaluation compares PBE-CC head to head against the leading cellular-aware and wireless-oblivious congestion control protocols proposed in the research community and in deployment, in mobile and static mobile scenarios, and over busy and quiet networks. Results show 6.3% higher average throughput than BBR, while simultaneously reducing 95th percentile delay by 1.8x.

cs.NI↗