Randy Katz, Professor

Closed (1) CellMate: A Responsive and Accurate Vision-based Appliance Identification System

Closed. This professor is continuing with Spring 2018 apprentices on this project; no new apprentices needed for Fall 2018.

Identifying and interacting with smart appliances has been challenging in the burgeoning smart building era. Existing identification methods require either cumbersome query statements or the deployment of additional infrastructure. There is no platform that abstracts sophisticated computer vision technologies to provide an easy visual identification interface, which is the most intuitive way for a human. CellMate is a new kind of visual appliance identification system that leverages advantages of different computer vision technologies and organizes them to optimize single image queries for fast response, high accuracy, and scalability.

Undergraduate students will learn the state-of-art of vision-based localization algorithms and implementations, design and implement our research prototypes, collect data, discuss with graduate students to tackle problems and invent algorithms, and contribute to research papers. Students will be given specific engineering tasks, and are expected to meet with the research supervisor once or twice a week.

Day-to-day supervisor for this project: Kaifei Chen, Graduate Student

Qualifications: C++ (required), Android programing (required), Algorithms (required), Operating System (required), Networking (required), Computer Vision (desirable), Machine Learning (desirable)

Weekly Hours: more than 12 hrs
Related website: https://github.com/SoftwareDefinedBuildings/CellMate

Closed (2) Social Games for Sustainability in Transportation

Check back for status

As UN Secretary General Ban Ki-Moon said in 1999, "Climate Change is the defining challenge of our age". However, behavioral economics indicates that it is a particularly hard problem for our civilization to tackle. It is complex and does not yield simple, easily implementable solutions. The threat is diffuse and distant, while the solutions involve immediate sacrifice. While the Paris accords have yielded significant progress in definining GHG reduction goals (US: 17\% below 2005 by 2020), the additional reduction required by 2050 (US: 80\% below 2005) is even more daunting.
This has lead to a renewed interest in serious games that can be used to motivate people towards societally important goals.

In this semester, we plan to have a small group of 4-5 students explore social games for sustainability in the context of sustainable transportation.

Concretely, they will build on an ongoing research project in the UC Berkeley EECS department (https://e-mission.eecs.berkeley.edu/) which automatically detects and models individual travel patterns, across all modes, and end to end. They will use this platform and work with the UC Berkeley Campus Sustainability office to design and implement a study on shifting commute mode share to campus. The study will allow participants to define their own goals and groups and meet them through a mixture of collaboration and competition.

The students will:
- design interventions (notifications/reminders, visualizations, goal tracking)
- define a mechanism and metrics to measure the effectiveness of the interventions
- implement a relevant set of interventions, and
- evaluate the interventions according to the metrics defined

At the end of the class, students will be expected to jointly write a 6 page extended abstract and submit it to the Computer Human Interaction (CHI) Late-Breaking Work track (http://chi2017.acm.org/lbw.html).

Day-to-day supervisor for this project: K. Shankari, Ph.D. candidate

Weekly Hours: 3-5 hrs

Open (3) Serverless Machine Learning

Open. Apprentices needed for the fall semester. Enter your application on the web beginning August 15th. The deadline to apply is Monday, August 27th at 9 AM.

This project is focused on providing a scalable and high-performance machine learning (ML) framework to run on serverless lambda functions (e.g., AWS Lambda) in the cloud. We want to bring the benefits of serverless computing to ML researchers and practitioners.

A recent trend towards serverless cloud abstractions promises highly scalable and elastic compute with small lambda functions paired with remote storage services (e.g., AWS S3). This new building block for the cloud can be leveraged to provide: a) better resource utilization, 2) VM-less deployments that require little management and 3) higher-level abstractions for users of ML algorithms. Our project aims to provide high-level Python abstractions backed by a high-performance ML backend to run machine learning workloads in the cloud. We plan to open-source our system in the next few months.

This work will be developed in the context of the RISELab research with a PhD student. Each URAP student will interact closely with a PhD student and contribute to his/her ongoing research.

Undergraduate will take responsibility in one of 2 roles: 1) distributed backend or 2) Python and visualization development. All undergrads participate in the design, implementation and evaluation of the system. The student will learn about system architecture and performance evaluation, and will contribute to research publications associated with this work. Co-authorship of research papers is highly likely.

Day-to-day supervisor for this project: Joao Carreira, Graduate Student

Qualifications: Role 1) C/C++, strong programming skills Role 2) Python and Data Visualization (e.g., Plotly, D3) Skills

Weekly Hours: more than 12 hrs

Closed (4) RISEing emission

Check back for status

e-mission is an open source platform for human travel data. It is an application in the RISE lab but, for historical reasons, does not use the RISE stack. We need to port it from the current ad-hoc stack to the RISE stack.

- Figure out relevant RISE projects
- Assess their maturity level
- For reasonably mature functionality replace current implementation with projects from the RISE stack.

You will get hands-on experience working with the latest and greatest RISE stack.

Day-to-day supervisor for this project: K. Shankari, Ph.D. candidate

Qualifications: Must have completed CS 162. Please report your grade in CS 162 as part of your application. Please also indicate your system-building experience (if any) along with your contribution for group projects.

Weekly Hours: more than 12 hrs