[Colloquium] Reminder - Joshua Viszlai MS Presentation/Feb 28, 2024

[Megan Woodward]

This is an announcement of Joshua Viszlai's MS Presentation
===============================================
Candidate: Joshua Viszlai

Date: Wednesday, February 28, 2024

Time:  9:30 am CST

Location: JCL 390

Title: Matching Generalized-Bicycle Codes to Neutral Atoms for Low-Overhead Fault-Tolerance

Abstract: Despite the necessity of fault-tolerant quantum systems built on error correcting codes, many popular codes, such as the surface code, have prohibitively large overheads. This has spurred significant interest in quantum LDPC codes which can reduce this overhead by an order of magnitude. Unfortunately, this reduction comes at the cost of many long-range interactions. The difficulty in implementing these long range interactions raises questions about the practical viability of quantum LDPC codes.

In this work we identify quantum LDPC codes that can be used in practice by carefully matching their features with underlying hardware. We target a specific family, known as generalized- bicycle codes, which have the favorable property of repeated check structure. We find this property can be exploited by quantum computers made from atom arrays, enabling an efficient implementation. Using numerical simulations, we evaluate the performance of generalized-bicycle codes in atom arrays and find comparable logical error rates to surface codes, but with an order of magnitude lower qubit overhead.

We then quantify the utility of this implementation in the context of a complete fault tolerant architecture. We compare the cost of implementing benchmark programs in a standard, surface code only architecture and a recently proposed mixed architecture where data is stored in qLDPC memory but is loaded to surface codes for computation. We treat the mixed architecture as a simple memory hierarchy and analyze the dependence of performance on program structure and compilation. Overall, for most programs of interest we find the mixed architecture leads to reduced costs compared to a standard, surface code only architecture. Additionally, we point out important compiler optimizations that emphasize qubit reuse and perform data prefetching to further improve the performance of the mixed architecture.

Advisors: Fred Chong

Committee Members: Fred Chong, Hannes Bernien, and Robert Rand





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20240228/89eeb90d/attachment.html>