I am a research scientist at the SK Telecom Data R&D Center. Previously, I was in the Data Science Laboratory supervised by Professor Sungroh Yoon. My research interests lie in representation learning, and I examined a methodology unifying deep learning and manifold learning in my dissertation. I am recently interested in deep learning for anomaly detection and have a technical blog where I share my experience and thoughts about engineering. Also, I have been contributing to some open source projects including Keras.

Skills

  • Engineering Scalable Machine Learning, Deep Learning, Statistical Learning, Parallel Computing, Bioinformatics
  • Programming Python, Matlab, C, C++, CUDA, Scala, Spark, Java, JavaScript, LaTeX, PHP, HTML, CSS, SQL, Shell, R, ASP, Rails
  • Miscellaneous Photoshop, Illustrator, Premiere

Recent Publications

  • Taehoon Lee et al.
    "HiComet: A High-Throughput Comet Analysis Tool for Large-Scale DNA Damage Assessment"
    BMC Bioinformatics, vol. 19 (Suppl 1), no. 44, pp. 49-61, February 2018. [paper] [code]
  • Taehoon Lee, Joong-Ho Won, Johan Lim, and Sungroh Yoon
    "Large-scale Structured Sparsity via Parallel Fused Lasso on Multiple GPUs"
    Journal of Computational and Graphical Statistics, vol. 26, no. 4, pp. 851-864, October 2017. [paper]
  • Andre S. Yoon, Taehoon Lee et al.
    "Semi-supervised Learning with Deep Generative Models for Asset Failure Prediction"
    in Proceedings of KDD17 Workshop on Machine Learning for Prognostics and Health Management, Halifax, Nova Scotia, Canada, August 2017. [paper]
  • Taehoon Lee
    "Robust Feature Learning with Deep Neural Networks"
    PhD Dissertation, Seoul National University, Department of Electrical and Computer Engineering, May 2016. [paper] [slides] [template]
  • Taehoon Lee, Taesup Moon, Seung Jean Kim, and Sungroh Yoon
    "Regularization and Kernelization of the Maximin Correlation Approach"
    IEEE Access, vol. 4, pp. 1385-1392, April 2016. [paper] [page] [code]
  • Byunghan Lee, Taehoon Lee, Byunggook Na, and Sungroh Yoon
    "DNA-Level Splice Junction Prediction using Deep Recurrent Neural Networks"
    in Proceedings of NIPS Workshop on Machine Learning in Computational Biology, Montreal, Canada, December 2015. [paper]
  • Taehoon Lee, Minsuk Choi, and Sungroh Yoon
    "Manifold Regularized Deep Neural Networks using Adversarial Examples"
    arXiv:1511.06381 [cs.LG], November 2015. [paper] [code]
  • Seungmyung Lee, Hanjoo Kim, Siqi Tan, Taehoon Lee, Sungroh Yoon, and Rhiju Das
    "Automated band annotation for RNA structure probing experiments with numerous capillary electrophoresis profiles"
    Bioinformatics, vol. 31, no. 17, pp. 2808-2815, September 2015. [paper]
  • Taehoon Lee and Sungroh Yoon
    "Boosted Categorical Restricted Boltzmann Machine for Computational Prediction of Splice Junctions"
    in Proceedings of International Conference on Machine Learning (ICML), Lille, France, July 2015. [paper] [page] [slides] [poster]
  • Donghyeon Yu, Joong-Ho Won, Taehoon Lee, Johan Lim, and Sungroh Yoon
    "High-dimensional Fused Lasso Regression using Majorization-Minimization and Parallel Processing"
    Journal of Computational and Graphical Statistics, vol. 24, no. 1, pp. 121-153, March 2015. [paper]
  • Taehoon Lee et al.
    "Robust Classification of DNA Damage Patterns in Single Cell Gel Electrophoresis"
    in Proceedings of International Conference of the IEEE Engineering in Medicine and Biology Society, Osaka, Japan, July 2013. [paper] [poster]

Open Source Contributions

  • Contributing to popular projects
    May 2017 - Current
    Keras (Committer and Top 3 Contributor), TensorFlow, scikit-learn, Matplotlib
  • TensorNets
    High level network definitions with pre-trained weights in TensorFlow.
    Python, 2017. [link]
  • HiComet
    The official web server implementation for A High-Throughput Comet Analysis Tool for Large-Scale DNA Damage Assessment.
    MATLAB and JavaScript, 2017. [link]
  • RMCA
    The official implementation for Regularization and Kernelization of the Maximin Correlation Approach.
    MATLAB, 2016. [link]
  • easyDL
    EasyDL supports various deep learning model variants. You can easily configure parameters of all layers with a model signature.
    MATLAB, 2015. [link]
  • taehoonlee/caffe
    Caffe is one of the most popular deep learning frameworks. I added here new forward-backward steps and customized several layers for adversarial training and manifold regularization.
    C++ and CUDA, 2015. [link]

Recent Projects

  • Classification of Car Environment and Internal Car Data
    Nov 2015 - Apr 2016 with Hyundai Motor Group (Data Analytics 1 Team)
    keywords: time series, multi modal learning, recurrent neural networks, tensorflow
  • Classification of Overhead Console Signals using Convolutional Neural Networks
    Aug 2015 - Mar 2016 with Hyundai Motor Group (Vecture Technology Development Team)
    keywords: time series, spectrogram, convolutional neural networks, matlab
  • Development of Ontology Based Framework and Heterogeneous Bigdata Models for Personalized Medicine
    Jun 2015 - Feb 2016 with National Research Foundation
    keywords: biomedical imaging, genetic analysis, personalized medicine
  • Biomedical Signal Processing for ICT Convergence:
    Anomaly Detection and Sequence Classification with Recurrent Neural Networks

    Mar 2015 - Feb 2016 with Samsung Advanced Institute of Technology
    keywords: anomaly detection, sequence classification, recurrent neural networks, tensorflow, keras
  • Human-Level Lifelong Machine Learning
    Mar 2014 - Feb 2015 with Ministry of Science, ICT, and Future Planning
    keywords: deep learning, parallel computing
  • Biomedical Signal Processing for ICT Convergence:
    High-Dimensional Clustering and Visualization of Representations from Deep Learning

    Mar 2014 - Feb 2015 with Samsung Advanced Institute of Technology
    keywords: deep learning, convolutional neural networks, visualization, python, caffe
  • Medical Imaging for Diagnosis of Cerebral Palsy, Bone Age Estimation, and Multiple Epiphyseal Dysplasia
    Jul 2013 - Jun 2014 with Seoul National University Bundang Hospital (Department of Orthopaedic Surgery)
    keywords: medical imaging, deep learning, dictionary learning, classification, matlab
  • Biomedical Signal Processing for ICT Convergence:
    CAD for Diagnosing Cancer based on Medical Imaging and Genomic analysis

    Mar 2013 - Feb 2014 with Samsung Advanced Institute of Technology
    keywords: medical imaging, segmentation, matlab

Location

CENTROPOLIS, 26 Ujeongguk-ro, Jongno-gu, Seoul, 03161