Yahan Yang, Ramneet Kaur, Souradeep Dutta, and Insup Lee won the Best Paper Award at the 13th ACM/IEEE International Conference on Cyber-Physical Systems (ICCPS 2022). Below is the title and abstract for their paper.
"Interpretable Detection of Distribution Shifts in Learning Enabled Cyber-Physical Systems"
Deep neural networks have allowed us to use high dimensional real world signals generated from sensors like camera and LiDAR. However, this comes with its potential perils. The pitfalls arise from possible over-fitting, and subsequent unsafe behavior when exposed to unknown environments. In this paper, our proposal is to build good representations for in-distribution data. We introduce the idea of a memory bank to store prototypical samples from the input space. We use these memories to compute probability density estimates using kernel density estimation techniques. We evaluate our technique on two challenging scenarios : a self-driving car setting implemented inside the simulator CARLA with image inputs, and an autonomous racing car navigation setting, with LiDAR inputs. An added benefit of using training samples as memories to detect out-of-distribution inputs is that the system is interpretable to a human operator. Explanation of this nature is generally hard to obtain from pure deep learning based alternatives.
Congratulations on your well-deserved achievement!
We’re thrilled to announce our newest postdoc, Kuk Jang, is joining PRECISE!
Kuk recently completed his doctorate in the Electrical and Systems Engineering Department and has been working with Dr. Rahul Mangharam in mLab since 2014. His research interests include domain adaptation and reinforcement learning for medical and robotics applications.
Kuk will be working with Drs. Insup Lee and Jim Weimer on data design for robust machine learning in our “MURI” and “Smart Alarms 2.0” projects. We are glad he decided to extend his time at Penn and look forward to seeing his continued contributions.
Alan Ismaiel, Ivan Ruchkin, Jason Shu, Oleg Sokolsky, and Insup Lee won the Best Contributed Theoretical Paper Award at the 54th Winter Simulation Conference (WSC 2021). Below is the title and abstract for their paper:
Title: Data Generation with PROSPECT: a Probability Specification Tool
Abstract: Stochastic simulations of complex systems often rely on sampling dependent discrete random variables. Currently, their users are limited in expressing their intention about how these variables are distributed and related to each other over time. This limitation leads the users to program complex and error-prone sampling algorithms. This paper introduces a way to specify, declaratively and precisely, a temporal distribution over discrete variables. Our tool PROSPECT infers and samples this distribution by solving a system of polynomial equations. The evaluation on three simulation scenarios shows that the declarative specifications are easier to write, 3x more succinct than imperative sampling programs, and are processed correctly by PROSPECT.
Congratulations on your well-deserved achievement!
StrokeDetectAI, a Python library for detecting onset of asymmetric movement in hospitalized patients with risk factors of stroke and those with no baseline asymmetric upper extremity weakness, has officially been licensed to Neuralert! This is a technology developed out of our Assured Autonomy and Smart Alarms projects.
PRECISE received an award from the National Science Foundation on their proposal entitled "Active sensing and personalized interventions for pandemic-induced social isolation".
Social connections are essential for individuals' health and the growth of a community. The onset of the COVID-19 pandemic greatly exacerbated the problems of social isolation, depriving older adults of their personal interactions with peers and often even with caregivers. Academic researchers (Teruo Higashino, Hajime Nagahara, Teruhiro Mizumoto, Viktor Erdlyi, Manabu Ikeda, Mamoru Hashimoto and Yasuyuki Gondo) from Osaka University and the PRECISE center at the University of Pennsylvania will partner together to develop a technology platform and tools to facilitate re-engagement of community members at high risk for social isolation in the US and in Japan. This project is a joint collaboration between the National Science Foundation and the Japan Science and Technology Agency.
We are proud to share that Xiayan Ji and Shuo Li have been selected as one of the finalists in the Wells Fargo Campus Analytics Challenge. During the competition, they were provided with a large dataset and instructions to develop a machine learning model to predict suspected elder fraud in the digital payments space.
EMBS 2021 Graduate, Celine Lee, talks with Justin Gottschlich, Principal AI Scientist and Director/Founder of Machine Programming Research at Intel Labs, about her journey to Penn and machine programming.
Congratulate to Jim, Co-Founder of Neuralert Technologies, LLC -- a wearable medical device company that detects symptoms of potential stroke, on getting FDA designation for device, identifies potential lead investor. Their system uses non-invasive, wearable devices which accurately detect stroke symptoms and then alerts clinical staff to more rapidly initiate stroke mitigating treatments.
Joe Devietti has been selected as the new CIS Undergraduate Chair at Penn Engineering. His appointment will start on 1 July 2021.
Joe is renowned for his research in using both hardware and software techniques to simplify multiprocessor programming. He is a successful entrepreneur and has been an amazing mentor to many undergraduate, Master's, and PhD students. As a Penn CIS undergraduate alumnus himself, he'll bring a unique perspective to Penn's undergraduate programs.
Nikolai Matni has been awarded a prestigious Google Research Scholar Award for his research on Roust Learning for Safe Control. The Research Scholar Program 2021 supports early-career professors who are pursuing world-class technical research in Computer Science, Engineering and related fields around the world. For 2021 Google have granted 77 awards, which included 86 principal investigators representing 15+ countries and over 50 universities. All proposals go through a stringent review process involving expert reviewers to assess them for merit, innovation and alignment with Google’s overall research policy.
Congratulations to Nikolai Matni on this well-deserved recognition by Google.