Yi-Chun Lin

Master Thesis- Active metric learning for 3D shape classification • Solve the difficulty of many machine learning algorithms suffering from a lack of labeled data • Select the most informative data points as training dataset • Large Margin Nearest Neighbor classification (LMNN) in an adaptive manner to make it more efficient and accurate. • Instead of fixed k Nearest Neighbours, applying the stochastic neighbors assignments to select the optimal neighborhood size.

• Create an REST API for integrating the platform of 3D protein structure visualization (Aquaria.ws) with protein prediction system (PredictProtein Service) • Add and emphasize the additional protein information, such as Interaction Sites (ISIS) identifying, Solvent Accessibility and etc.

Code refactor on 3 sparse coding (SC) and dictionary learning (DL) related papers: • Transfer Sparse Coding for Robust Image Representation • Graph Regularized Sparse Coding for Image Representation • Stacked Multiscale Feature Learning for Domain Independent Medical ImageSegmentation

• Detect anomalies in German credit dataset • Detect spams in email corpus • Fill voids created due to missing or corrupt data based on the satellite images • Learn the diagnosis results whether malignant or benign from a very small amount of labeled dataset • Recommendation system challenge: to predict the scores based on 10 jokes from 5981 users.

• Develop and maintain the financial system. • Involve project management • Design the workflow of financial system. • Negotiate with clients. • Solve problems if clients propose

Machine Learning / Artificial Intelligence/ Computer Vision

Software Engineering