Project

User Agency in Streaming Media (Netflix)

Published:

​How does Netflix make you spend more time consuming their contents? How do features like autoplay affect you? Can we propose interface design that mediate tech addiction?

Physica: Tangible Physics Simulation via Swarm Robot Cluster for Improvisation with Tangible User Interface

Published:

​Through Tangible User Interface (TUI), people are able to experi- ence, manipulate, interact, and alter their surrounding environment. In our paper, we proposed a novel physics interactive sandbox, Phys- ica, which enables user to play around simple physics simulation models (e.g. kinetic motions) through direct sensations. By doing so, we aim to convert complex physics equations into understandable tangible patterns created by a collection of two-wheeled robots (i.e Toio). Unlike traditional Graphical User Interface (GUI) which typically is hosted on different physics-related websites, our project provides an on-demand and real-time interactive tool to reduce users’ mental workload and make physical science more accessible for students and the general public. Moreover, since the background simulation algorithm obeys physics laws, we naturally blend physical world with digital word and create a variety of interesting cross-interactions.

Machine Learning and the search for soft γ-ray pulsars

Published:

​A number of researchers has used machine learning technique to analyze and classify Gamma-Ray sources. However, few study has investigated whether we can have an algorithmic method to classify soft gamma-ray pulsars which have energy peaks that is under 1000 MeV. Therefore, our research focuses on using unsupervised learning machine learning to detect whether our 4FGL dataset contains different clusters. Ideally, one of the clusters should be soft gamma-ray pulsars.

Topos

Published:

A diverse and depolarized social network for people to match and join the topics and activities they love (similar to reddit+eventbrite but not exactly the same). We encourage our users to meet new friends offline, while we also have a lot of exciting features to ensure a great online experience.

Deep Learning Methods in Detecting Prostate Cancer

Published:

​Contemporary approaches to prostate cancer diagnosis are fallible and inefficient, often requiring multiple specialists to confirm a single diagnosis. Computer-automated decision-making systems can reduce the rate of misdiagnosis and increase reproducibility. In particular, convolutional neural networks (CNNs) are the leading algorithm for image recognition.