Youngmok Jung

Selected Publications

1. Preprint

   Generalizing deep variant callers via domain adaptation and semi-supervised learning

   Youngmok Jung, Jinwoo Park, Hwijoon Lim, Jeong Seok Lee, Young Seok Ju, and Dongsu Han

   preprint 2023

   Abs HTML Github

   Deep learning-based variant callers (DVCs) offer state-of-the-art performance in small variant detection from DNA sequencing data. However, their reliance on supervised learning and the subsequent need for extensive labeled data pose a potential hurdle to their generalizability across diverse sequencing methods with varying error profiles. Indeed, even minor discrepancies in error profiles can compromise the robustness of DVCs and impair the variant calling accuracy in the target sequencing method. To mitigate these challenges, we propose RUN-DVC, the first semi-supervised training approach for DVCs that presents two complementary training techniques to the conventional supervised training approach. RUN-DVC leverages semi-supervised learning techniques to learn error profiles from unlabeled datasets of the target sequencing method as well as a domain adaptation technique to aid semi-supervised learning by reducing the domain discrepancy due to different error profiles. We analyze and contrast RUN-DVC against the supervised training approach under various generalization scenarios using nine sequencing methods from Illumina, BGI, PacBio, and Oxford Nanopore sequencing platforms. Remarkably, RUN-DVC significantly improves the variant calling accuracy of DVC in the target sequencing method even with purely unlabeled datasets in the target domain and enables label-efficient generalization when partially labeled datasets are available. Our results suggest RUN-DVC is a promising semi-supervised training method for DVCs with the potential to broaden the use of DVC across diverse sequencing methods.

   Bioinformatics

   BWA-MEME: BWA-MEM emulated with a machine learning approach

   Youngmok Jung and Dongsu Han

   Oxford Bioinformatics, 2022

   Abs HTML Github

   The growing use of next-generation sequencing and enlarged sequencing throughput require efficient short-read alignment, where seeding is one of the major performance bottlenecks. The key challenge in the seeding phase is searching for exact matches of substrings of short reads in the reference DNA sequence. Existing algorithms, however, present limitations in performance due to their frequent memory accesses.

   This paper presents BWA-MEME, the first full-fledged short read alignment software that leverages learned indices for solving the exact match search problem for efficient seeding. BWA-MEME is a practical and efficient seeding algorithm based on a suffix array search algorithm that solves the challenges in utilizing learned indices for SMEM search which is extensively used in the seeding phase. Our evaluation shows that BWA-MEME achieves up to 3.45x speedup in seeding throughput over BWA-MEM2 by reducing the number of instructions by 4.60x, memory accesses by 8.77x, and LLC misses by 2.21x, while ensuring the identical SAM output to BWA-MEM2.

   SIGCOMM

   LiveNAS - Neural-Enhanced Live Streaming: Improving Live Video Ingest via Online Learning

   Jaehong Kim, (Co-first)Youngmok Jung, Hyunho Yeo, Juncheol Ye and Dongsu Han

   In Proceedings of ACM SIGCOMM conference, 2020

   Abs HTML

   Live video accounts for a significant volume of today’s Internet video. Despite a large number of efforts to enhance user quality of experience (QoE) both at the ingest and distribution side of live video, the fundamental limitations are that streamer’s upstream bandwidth and computational capacity limit the quality of experience of thousands of viewers. To overcome this limitation, we design LiveNAS, a new live video ingest framework that enhances the origin stream’s quality by leveraging computation at ingest servers. Our ingest server applies neural super-resolution on the original stream, while imposing minimal overhead on ingest clients. LiveNAS employs online learning to maximize the quality gain and dynamically adjusts the resource use to the real-time quality improvement. LiveNAS delivers high-quality live streams up to 4K resolution, outperforming WebRTC by 1.96 dB on average in Peak-Signal-to-Noise-Ratio on real video streams and network traces, which leads to 12%-69% QoE improvement for live stream viewers.

   Publications

   1. CoNEXT

      Co-optimizing for Flow Completion Time in Radio Access Network

      Jaehong Kim, Yunheon Lee, Hwijoon Lim, Youngmok Jung, Song Min Kim, and Dongsu Han

      In Proceedings of the 18th International Conference on emerging Networking EXperiments and Technologies, 2022

      ABS HTML

      Traffic from interactive applications demanding low latency has become dominant in cellular networks. However, existing schedulers of cellular network base stations fall short in delivering low latency when prior information (i.e., dedicated Quality of Service (QoS)) is unavailable; they become service agnostic and perform towards maximizing the radio resource utilization or user fairness. We identify a new opportunity of providing a better latency for those latency-sensitive traffic flows by additionally taking the Flow Completion Time (FCT) into account in downlink scheduling at the base stations. However, the key challenges are 1) it can bring a severe cost in optimization metrics of the existing scheduler and 2) it should work without prior knowledge of the traffic. To this end, we present OutRAN, a practical flow scheduler designed for Radio Access Network that co-optimizes the FCT and optimization objectives of the cellular scheduler. The resulting system does not require prior information. Through simulation and over-the-air evaluation, we demonstrate that OutRAN outperforms the legacy LTE/5G schedulers in FCT, which leads to the reduction in webpage load time of Android phones.

      SIGCOMM

      NeuroScaler: Neural Video Enhancement at Scale

      Hyunho Yeo, Hwijoon Lim, Jaehong Kim, Youngmok Jung, Juncheol Ye, and Dongsu Han

      In Proceedings of ACM SIGCOMM conference, 2022

      Abs HTML GITHUB

      High-definition live streaming has experienced tremendous growth. However, the video quality of live video is often limited by the streamer's uplink bandwidth. Recently, neural-enhanced live streaming has shown great promise in enhancing the video quality by running neural super-resolution at the ingest server. Despite its benefit, it is too expensive to be deployed at scale. To overcome the limitation, we present NeuroScaler, a framework that delivers efficient and scalable neural enhancement for live streams. First, to accelerate end-to-end neural enhancement, we propose novel algorithms that significantly reduce the overhead of video super-resolution, encoding, and GPU context switching. Second, to maximize the overall quality gain, we devise a resource scheduler that considers the unique characteristics of the neural-enhancing workload. Our evaluation on a public cloud shows NeuroScaler reduces the overall cost by 22.3× and 3.0--11.1× compared to the latest per-frame and selective neural-enhancing systems, respectively.

      EuroSys

      Towards Timeout-less Transport in Commodity Datacenter Networks

      Hwijoon Lim, Wei Bai, Yibo Zhu, Youngmok Jung and Dongsu Han

      The 16th European Conference on Computer Systems, 2021

      Abs HTML Github

      Despite recent advances in datacenter networks, timeouts caused by congestion packet losses still remain a major cause of high tail latency. Priority-based Flow Control (PFC) was introduced to make the network lossless, but its Head-of-Line blocking nature causes various performance and management problems. In this paper, we ask if it is possible to design a network that achieves (near) zero timeout only using commodity hardware in datacenters. Our answer is TLT, an extension to existing transport designed to eliminate timeouts. We are inspired by the observation that only certain types of packet drops cause timeouts. Therefore, instead of blindly dropping (TCP) or not dropping packets at all (RoCEv2), TLT proactively drops some packets to ensure the delivery of more important ones, whose losses may cause timeouts. It classifies packets at the host and leverages color-aware thresholding, a feature widely supported by commodity switches, to proactively drop some less important packets. We implement TLT prototypes using VMA to test with real applications. Our testbed evaluation on Redis shows that TLT reduces 99%-ile FCT up to 91.7% on handling bursts of SET operations. In large-scale simulations, TLT augments diverse datacenter transports, from widely-used (TCP, DCTCP, DCQCN) to state-of-the-art (IRN and HPCC), by achieving up to 81% lower tail latency.

      Mobicom

      NEMO: Enabling Neural-enhanced Video Streaming on Commodity Mobile Devices

      Hyunho Yeo, Chan Ju Chong, Youngmok Jung, Juncheol Ye and Dongsu Han

      The 26th Annual International Conference on Mobile Computing and Networking, 2020

      Abs HTML Github

      The demand for mobile video streaming has experienced tremendous growth over the last decade. However, existing methods of video delivery fall short of delivering high-quality video. Recent advances in neural super-resolution have opened up the possibility of enhancing video quality by leveraging client-side computation. Unfortunately, mobile devices cannot benefit from this because it is too expensive in computation and power-hungry. To overcome the limitation, we present NEMO, a system that enables real-time video super-resolution on mobile devices. NEMO applies neural super-resolution to a few select frames and transfers the outputs to benefit the remaining frames. The frames to which super-resolution is applied are carefully chosen to maximize the overall quality gains. NEMO leverages fine-grained dependencies using information from the video codec and strives to provide guarantees in the quality degradation compared to per-frame super-resolution. Our evaluation using a full system implementation on Android shows NEMO improves the overall processing throughput by x11.5, reduces energy consumption by 88.6%, and maintains device temperatures at acceptable levels compared to per-frame super-resolution, while ensuring high video quality. Overall, this leads to a 31.2% improvement in quality of experience for mobile users.

      OSDI

      Neural Adaptive Content-aware Internet Video Delivery

      Hyunho Yeo, Youngmok Jung, Jaehong Kim, Jinwoo Shin, and Dongsu Han

      13th USENIX Symposium on Operating Systems Design and Implementation, 2018

      Abs HTML Github

      Internet video streaming has experienced tremendous growth over the last few decades. However, the quality of existing video delivery critically depends on the bandwidth resource. Consequently, user quality of experience (QoE) suffers inevitably when network conditions become unfavorable. We present a new video delivery framework that utilizes client computation and recent advances in deep neural networks (DNNs) to reduce the dependency for delivering high-quality video. The use of DNNs enables us to enhance the video quality independent to the available bandwidth. We design a practical system that addresses several challenges, such as client heterogeneity, interaction with bitrate adaptation, and DNN transfer, in enabling the idea. Our evaluation using 3G and broadband network traces shows the proposed system outperforms the current state of the art, enhancing the average QoE by 43.08% using the same bandwidth budget or saving 17.13% of bandwidth while providing the same user QoE.

   Experiences

   1. Enhancing genome analysis pipeline with AI and ML techniques

      BWA-MEME: Machine-learning Enhanced Read Alignment Software
      • BWA MEM is a industry-standard alignment software for next-generation sequencing data developed by the Broad Institute of MIT and Harvard.
      • We developed BWA-MEME that achieves up to 3.45x speedup in seeding throughput over its' predecessor, BWA-MEM2 from Intel and Broad institute, while ensuring identical output.
      • BWA-MEME is now operational in the production environments of several institutions, projected to lower alignment costs by 35%. This efficiency translates into millions of $ in cost reductions for projects on a million-genome scale.
      Generalizing Deep Variant Callers via Domain Adaptation and Semi-supervised learning
      • Deploying deep learning-based variant callers (DVCs) to a sequencing method with varying error profiles necessitates generalization which is challenging due to their reliance on extensive labeled data.
      • We developed a generalization framework that enables deep learning-based variant callers (e.g., Google DeepVariant, Clair3) to accommodate diverse sequencing methods, leveraging semi-supervised learning and domain adaptation techniques.
      • RUN-DVC improves SNP F1-score and INDEL F1-score by up to 6.40 %p and 9.36 %p over the supevised training approach using only unlabeled data or achieves the same variant calling accuracy using merely half of the labeled data.
      

      
      

      Designing AI-enhanced high-performance networked systems

      Live-NAS: Stream high-quality live videos even when the network becomes congested
      • Developed neural-enhanced live video streaming system (LiveNAS) that incorporates 1) Online training and inference system for super-resolution DNN model during live video streaming. 2) Bandwidth allocation algorithm to maximize user Quality of Experience (QoE)
      • LiveNAS delivers live video with the same quality as Google WebRTC using only 45.9% bandwidth on average. LiveNAS system is built on top of Google WebRTC C++ code.
      NAS: Neural Adaptive Streaming
      • Developed an adaptive bit-rate streaming algorithm based on Reinforcement Learning (a3c) for NAS, a video-on-demand system such as Youtube or Netflix.
      • NAS enhanced the average QoE by 43.08% using the same bandwidth budget or saving 17.13% of bandwidth while providing the same user QoE.
      TLT: Timeout-less Transport in Commodity Datacenter Networks
      • Implemented data-center network protocols (TLT, PFC) in NS-3 network simulator. TLT protocol was also implemented in Linux kernel and tested in real-world testbed. TLT augments diverse datacenter transports, from widely-used (TCP, DCTCP, DCQCN) to state-of-the-art (IRN and HPCC), by achieving up to 81% lower tail latency.
      

      Opensource

      Main Contributor

      RUN-DVC

      https://github.com/kaist-ina/RUN-DVC
      - Generalizing deep learning-based variant caller via domain adaptation and semi-supervised learning.

      BWA-MEME

      https://github.com/kaist-ina/BWA-MEME
      - Short-read alignment software, released for benefit of research community.

      Collaborative Projects

      NAS: Neural Adaptive Content-aware Internet Video Delivery

      https://github.com/kaist-ina/NAS_public

      NEMO: Enabling Neural-enhanced Video Streaming on Commodity Mobile Devices

      https://github.com/kaist-ina/nemo

      TLT: Towards Timeout-less Transport in Commodity Datacenter Networks

      https://github.com/kaist-ina/ns3-tlt-rdma-public

      NeuroScaler: Neural Video Enhancement at Scale

      https://github.com/kaist-ina/neuroscaler-public

      Skills

      Programming Languages: Python, C++, Rust, SQL, Shell scripting

      Frameworks: Tensorflow, Pytorch, Spark, Hadoop, Python Django, Node.js, Cloudstack