Amgen Scholars: Announcements of Opportunity
Related Pages 
Below are Announcements of Opportunity posted by Caltech faculty for the Amgen Scholars program.
Announcements of Opportunity are posted as they are received. Please check back regularly for new AO submissions! Remember: This is just one way that you can go about identifying a suitable project and/or mentor. For additional tips on identifying a mentor click here.
Please remember:
- Students pursuing Amgen must be U.S. citizens, U.S. permanent residents, or students with DACA status.
- Students pursuing Amgen must complete the 10-week program from June 21 - August 25, 2023. Students must commit to these dates. No exceptions will be made.
- Accepted students must live in provided Caltech housing.
<< Prev
Record
29 of
59
Next >>
Back To List
| Project: | Accelerating astrophysical transient classifiers with FPGAs | ||||||||
| Disciplines: | Astronomy, Data Science | ||||||||
| Mentor: |
Matthew Graham,
Prof., (PMA),
mjg@caltech.edu, |
||||||||
| Mentor URL: | https://www.astro.caltech.edu/~mjg (opens in new window) | ||||||||
| Background: |
The Zwicky Transient Facility (ZTF) surveys the observable night sky from Palomar Observatory every two nights and produces a real-time stream of notifications (alerts) of detected changes in the brightness or position of astrophysical sources. Among the hundreds of thousands of alerts published every night, there are rare events which are short lived (transient), such as the electromagnetic counterpart to a compact object merger detectable by LIGO, and these need to be identified as quickly as possible to ensure swift followup with other ground- and space-based facilities. Existing machine-learning algorithms run on regular computing hardware (CPUs/GPUs) but application specific integrated circuits (ASICs) offer the most low latency solution. The NSF-funded Accelerated AI Algorithms for Data-Driven Discovery Institute (A3D3) is a multi-institutional project aiming to develop customized AI solutions to process large data sets in real time in high-energy physics, multi-messenger astronomy (MMA), and systems neuroscience. The Caltech component is focused on developing the fastest optimal responses to detections of short-lived astronomical phenomena with limited resources. Part of this lies in porting and testing existing classification algorithms on new hardware. Low-latency solutions to MMA problems will become increasingly important as the Rubin Observatory Legacy Survey of Space and Time (LSST) comes online in early 2025 which will produce millions of alerts per night. |
||||||||
| Description: | This project involves taking existing deep learning algorithms that classify transient alerts developed for ZTF and porting them to run on a FPGA (Pynq Z2) using the hls4ml Python compiler. The performance of these algorithms will be assessed and compared to regular and other accelerated computing hardware to determine the optimal hardware solution. | ||||||||
| References: |
Deep learning for the Zwicky Transient Facility: real/bogus classification and identification of fast-moving objects: https://ui.adsabs.harvard.edu/abs/2020AAS...23538608D/abstract A3D3 website: https://a3d3.ai hls4ml website: https://fastmachinelearning.org/hls4ml/#:~:text=hls4ml is a Python package,configured for your use-case! |
||||||||
| Student Requirements: | Proficiency in Python | ||||||||
| Programs: |
This AO can be done under the following programs:
|
<< Prev Record 29 of 59 Next >> Back To List