A5103258769- Image and Video Quality Assessment
- Caching and Content Delivery
- Network Packet Processing and Optimization
- Internet Traffic Analysis and Secure E-voting
- Network Security and Intrusion Detection
- Video Coding and Compression Technologies
- Advanced Neural Network Applications
- Peer-to-Peer Network Technologies
- Video Surveillance and Tracking Methods
- Algorithms and Data Compression
- Anomaly Detection Techniques and Applications
- Visual Attention and Saliency Detection
- Network Traffic and Congestion Control
- Software-Defined Networks and 5G
- Topic Modeling
- Cloud Computing and Resource Management
- Data Stream Mining Techniques
- Adversarial Robustness in Machine Learning
- Advanced Computing and Algorithms
- Privacy-Preserving Technologies in Data
- CCD and CMOS Imaging Sensors
- IoT and Edge/Fog Computing
- Software System Performance and Reliability
- Natural Language Processing Techniques
- Advanced Optical Sensing Technologies
University of Chicago
2018-2025
Second Affiliated Hospital & Yuying Children's Hospital of Wenzhou Medical University
2024
Hohai University
2023-2024
Changchun University of Technology
2022-2024
Wenzhou Medical University
2024
University of Klagenfurt
2023
Ege University
2023
Rutgers, The State University of New Jersey
2023
University of Illinois Chicago
2019-2023
Dana-Farber Cancer Institute
2023
Many commercial video players rely on bitrate adaptation logic to adapt the in response changing network conditions. Past measurement studies have identified issues with today's respect three key metrics---efficiency, fairness, and stability---when multiple bitrate-adaptive share a bottleneck link. Unfortunately, our current understanding of why these effects occur how they can be mitigated is quite limited.
Recently, machine learning has been used in every possible field to leverage its amazing power. For a long time, the net-working and distributed computing system is key infrastructure provide efficient computational resource for learning. Networking itself can also benefit from this promising technology. This article focuses on application of Machine Learning techniques (MLN), which not only help solve intractable old network questions but stimulate new applications. In article, we summarize...
Applying deep convolutional neural networks (NN) to video data at scale poses a substantial systems challenge, as improving inference accuracy often requires prohibitive cost in computational resources. While it is promising balance resource and by selecting suitable NN configuration (e.g., the resolution frame rate of input video), one must also address significant dynamics configuration's impact on analytics accuracy. We present Chameleon, controller that dynamically picks best...
Bitrate adaptation is critical in ensuring good users' quality-of-experience (QoE) Internet video delivery system. Several efforts have argued that accurate throughput prediction can dramatically improve (1) initial bitrate selection for low startup delay and high resolution; (2) midstream QoE. However, prior ef- forts did not systematically quantify real-world predictability or develop algorithms. To bridge this gap, paper makes three key technical contributions: First, we analyze the...
Modern video players today rely on bit-rate adaptation in order to respond changing network conditions. Past measurement studies have identified issues with today's commercial when multiple bit-rate-adaptive share a bottleneck link respect three metrics: fairness, efficiency, and stability. Unfortunately, our current understanding of why these effects occur how they can be mitigated is quite limited. In this paper, we present principled analyze several through the lens an abstract player...
Video traffic already represents a significant fraction of today's and is projected to exceed 90% in the next five years. In parallel, user expectations for high quality viewing experience (e.g., low startup delays, buffering, bitrates) are continuously increasing. Unlike traditional workloads that either require latency short web transfers) or average throughput large file transfers), video requires sustained performance over extended periods time tens minutes). This imposes fundamentally...
Video streaming is crucial for AI applications that gather videos from sources to servers inference by deep neural nets (DNNs). Unlike traditional video optimizes visual quality, this new type of permits aggressive compression/pruning pixels not relevant achieving high DNN accuracy. However, much potential left unrealized, because current protocols are driven the source (camera) where compute rather limited. We advocate protocol should be real-time feedback server-side DNN. Our insight...
Name-based route lookup is a key function for Named Data Networking (NDN). The NDN names are hierarchical and have variable unbounded lengths, which much longer than IPv4/6 address, making fast name challenging issue. In this paper, we propose an effective Name Component Encoding (NCE) solution with the following two techniques: (1) A code allocation mechanism developed to achieve memory-efficient encoding components, (2) We apply improved State Transition Arrays accelerate longest prefix...
Streaming 360° videos requires more bandwidth than non-360° videos. This is because current solutions assume that users perceive the quality of in same way they means demand must be proportional to size user's field view. However, we found several quality-determining factors unique videos, which can help reduce demand. They include moving speed a viewpoint (center view), recent change video luminance, and difference depth-of-fields visual objects around viewpoint.
Interactive real-time streaming applications such as audio-video conferencing, online gaming and app streaming, place stringent requirements on the network in terms of delay, jitter, packet loss. Many these inherently involve client-to-client communication, which is particularly challenging since performance need to be met while traversing public wide-area (WAN). This different from typical situation cloud-to-client where WAN can often bypassed by moving a communication end-point cloud...
Sketch-based measurement has emerged as a promising alternative to the traditional sampling-based network approaches due its high accuracy and resource efficiency. While there have been various designs around sketches, they focus on measuring one particular flow key, it is infeasible support many keys based these sketches. In this work, we take significant step towards supporting arbitrary partial key queries, where only need specify full range of possible that are interest before starts, in...
Live video delivery is expected to reach a peak of 50 Tbps this year. This surging popularity fundamentally changing the Internet landscape. CDNs must meet users' demands for fast join times, high bitrates, and low buffering ratios, while minimizing their own cost responding issues in real-time. Wide-area latency, loss, failures, as well varied workloads ("mega-events" long-tail), make meeting these challenging.
Cameras are deployed at scale with the purpose of searching and tracking objects interest (e.g., a suspected person) through camera network on live videos. Such cross-camera analytics is data compute intensive, whose costs grow number cameras time. We present Spatula, cost-efficient system that enables scaling edge boxes to large networks by leveraging spatial temporal correlations. While such correlations have been used in computer vision community, Spatula uses them drastically reduce...
Driven by advances in computer vision and the falling costs of camera hardware, organizations are deploying video cameras en masse for spatial monitoring their physical premises. Scaling analytics to massive deployments, however, presents a new mounting challenge, as compute cost grows proportionally number feeds. This paper is driven simple question: can we scale such way that sublinearly, or even remains constant, deploy more cameras, while inference accuracy stable, improves. We believe...
Emerging deep learning-based video analytics tasks demand computation-intensive neural networks and powerful computing resources on the cloud to achieve high accuracy. Due latency requirement limited network bandwidth, edge-cloud systems adaptively compress data strike a balance between overall accuracy bandwidth consumption. However, degraded leads another issue of poor <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tail accuracy</i> ,...
As deep reinforcement learning (RL) showcases its strengths in networking, pitfalls are also coming to the public's attention. Training on a wide range of network environments leads suboptimal performance, whereas training narrow distribution results poor generalization.
Motivated by the increasing scale of data, we see a growing need high performance distributed machine learning systems. Many research works are being proposed to improve performance.