- Dockerized Scalable Multi-model Car Reservation Prediction : Kubernetes + Flask + Gnuicorn + DNN (multi-loss, multi-target) + XGBoost + Random Forest + multi-label SVM (RBF) + Softmax Reg + Multilinear Reg (https://nbviewer.jupyter.org/github/ginochen/Dual-branched_DNN_SVM/blob/master/Dual_branched_DNN_SVM.ipynb)
- NLP spacCy + LSTM Amazon review sentiment analysis (binary classification)(https://github.com/ginochen/NLP_spaCy_cuDNNLSTM)
- DNN credit card fraud detection (multi-feature binary classification) (https://github.com/ginochen/DNN_anomaly_detection)
- LSTM multi-feature stock forecast with Keras (RNN sequence modeling) (https://github.com/ginochen/LSTM_stock)
- LSTM weather prediction (RNN sequence modeling) (https://github.com/ginochen/LSTM)
- SARIMA weather prediction (time-series forecast) (https://github.com/ginochen/ARIMA)
- CNN (YOLOv3) object detection and classification on Open Images (computer vision)(https://github.com/ginochen/openimagechallenge)
- Web scraping and commodity prices EDA (data acquisition, cleaning, EDA) (https://github.com/ginochen/web_scraping)
- Stochastic forecast system (dissertation-appendix, https://scholarlyrepository.miami.edu/oa_dissertations/2137/)
Projects
Education
Ph.D., Meteorology and Physical Oceanography, Univ. of Maimi RSMAS
Advisor: Ben Kirtman, Prof. of Atmospheric Sciences. (see Kirtman group)
M.A. Climate & Society, Columbia Univ. (2009)
Advisor: Suzana J. Camargo, Lamont Research Prof.
B.S. Biology, Tunghai Univ. (2004)
Research Interests
I study the physics of clouds, which is one of the most complex processes to accurately simulate in a global weather model. I applied mathematical tools to reproduce the physical processes in a cost effective way to avoid running expensive models. I have developed a novel forecast system, which has the ability to run low cost simulations and quantify the model uncertainties in a seamless fashion.
A more technical description of my PhD research:
I focus on quantifying the uncertainties of model cloud physics with the tools : stochastic spectral method (i.e., Polynomial Chaos Expansion (PCE) with Dimension Adaptive Sparse Quadrature (DASQ) rules), Bayesian inference (using Delayed Rejection Adaptive Metropolis (DRAM) MCMC), Sobol global sensitivity analysis, clustering method (e.g., Heirarchi- cal Cluster Analysis (HCA), k-means/k-medoids clustering, etc), spectral methods (e.g., PCA, Rotated EOF, FFT, discrete Chebyshev transform, etc), and stochastic process parameterization.
A more technical description of my PhD research:
I focus on quantifying the uncertainties of model cloud physics with the tools : stochastic spectral method (i.e., Polynomial Chaos Expansion (PCE) with Dimension Adaptive Sparse Quadrature (DASQ) rules), Bayesian inference (using Delayed Rejection Adaptive Metropolis (DRAM) MCMC), Sobol global sensitivity analysis, clustering method (e.g., Heirarchi- cal Cluster Analysis (HCA), k-means/k-medoids clustering, etc), spectral methods (e.g., PCA, Rotated EOF, FFT, discrete Chebyshev transform, etc), and stochastic process parameterization.
Publications
G Chen, BP Kirtman, M Iskandarani, 2015, Quarterly Journal of the Royal Meteorological Society
Long-lived mesoscale convective systems of super-parameterized CAM and the response of CAM
G Chen, BP Kirtman, 2018, Journal of Advances in Modeling Earth Systems
Mesoscale Organized Convection in Climate Models: A Stochastic Scheme and Uncertainty Quantification
G Chen, 2018, Open Access Dissertations. 2137.
Also available on Google Scholar or Researchgate.
Subscribe to:
Comments (Atom)