Awards

• Best Paper Award in Signal Processing, Taiwan Telecommunications Annual Symposium, 2022. 

• THIRD place in the EMBS Student Paper Competition, the 43rd Annual International Conference of Engineering in Medicine and Biology, 2021.

• Foundation For The Advancement Of Outstanding Scholarship (傑出人才發展基金會 第五屆「年輕學者創新獎」), 2017.

• CIEE Outstanding Young Electrical Engineer Award (中國電機工程學會「優秀青年電機工程師獎」), 2015.

• Best Paper Award for the 2014 GIW-ISCB-ASIA conference.

• Excellent Teaching Award for the College of EECS, NTHU, 2012.

• Reward for Academic Excellence, NTHU, 2012, 2014.

• Best Poster Award, International Workshop on Mathematical Issues in Information Sciences (MIIS), Xi’an, China, 2012.

• Nominated for the Excellent Teaching Award for NTHU, 2011, 2013.

• E. A. Reid Fellowship Award, UIUC, Spring 2008.

• Vodafone Fellowship, UIUC, Fall 2006 - Spring 2008.

Patent

1. C. Lin, E. Onggosanusi, and A. Dabak, Patent TI-63186-“Codebook Design and Pre-Coder Selection for Closed-Loop MIMO”. 

The proposed codebook in the patent was adopted in the standard for next-generation wireless communication systems, 3rd Generation Partnership Project Long Term Evolution (3GPP LTE). Standard specifications: 3GPP TS 36.211 v.2.0.0 (2007-09) release 8.

2. Che Lin, Patent I 779887 -"Dynamic BnB recommendation system."  

Our technology is based on recurrent neural networks with an attention mechanism to extract information and provide predictions, which can be applied to BnB and e-commerce recommender systems that do not require user personal information and login. 

3. Che Lin, Wei-Zhu Chen, Patent I 818259 -"A Deep Learning Approach for Push Blockage Prediction."
   
   Our technology integrates different deep learning models, including DNN, RNN, Attention Mechanism, and natural language processing techniques, to achieve precise blockade intention prediction. This technology has been preliminarily tested in our offline push notification data set. The deep learning model we proposed is highly effective and practical in predicting user blockage of push notifications. We are confident that this technology will significantly contribute to the ecosystem of precision marketing via push notifications. 

4. Che Lin, Pei-Ying Liu, Tung-Hung Su, Patent I 833566 -"Lightweight Self-attention Model for High Missing Rate Electronic Health Records."
   
   This technology uses cutting-edge deep learning technology that allows for accurate disease prediction based on the observed value and missing information from the patient's past electronic health records through the self-attention mechanism. We collaborated with the National Taiwan University Hospital (NTUH) to validate our algorithm. Our technology was evaluated using hepatocellular carcinoma data collected from its Integrated Medical Database. The proposed technology was demonstrated to perform much better than traditional machine learning models, such as gated recurrent units, random forests, and conventional risk scores widely used in the medical domain. 

5. Che Lin, Ming-Che Cheng, Tung-Hung Su, Patent I 858354 -"Deep STI: Deep Stochastic Time-series Imputation on Electronic Medical Records"
   
   This technology utilized innovative deep-learning techniques. It imputes missing values through learning the observed patterns in EHRs and simultaneously predicts patients’ disease with the imputed EHRs. To validate our algorithm, we collaborated with the National Taiwan University Hospital (NTUH). Our technology was evaluated in Hepatocellular carcinoma data collected from Integrated Medical Database, NTUH. It was demonstrated to have better performance than traditional statistical models such as logistic regression, random forest, and risk scores.  

6. Che Lin, Shih-Hao Huang, Patent I 874059-''A DEEP LEARNING MODEL SYSTEM AND METHOD THEREOF"
   
   This technology integrates different deep learning models, including DNN, CNN, Multi-Modal learning, and also uses the Word2vec method to learn the representation of pro tein sequence and ligand SMILEs data. We then extract the semantic features from both and concatenate them along with the simulated docking information to achieve highly accurate drug target interaction predicting and ranking. This technology has been prel iminarily tested on the serine hydrolase data set. We have strong evidences that demonstrated the effectiveness of our deep learning model.