Berkeley Lab – CS Seminar

Date:
Monday, January 6, 2025

Time:
11:00am – 12:00pm

Location:
 Bldg. 59, Room 4102 

Speakers(s) and Affiliations(s):
Nishchay Suri
NASA Ames Quantum AI Lab (QuAIL), USRA

Title:
Error Mitigation Beyond Symmetry Verification and Open Quantum System Simulation using QSVT

Abstract:
In the first part, we present a general condition to obtain subspaces that decay uniformly in a system governed by the Lindblad master equation and use them to perform error-mitigated quantum computation. The expectation values of dynamics encoded in such subspaces are unbiased estimators of noise-free expectation values. We apply our theory to a system of qubits and qudits undergoing relaxation with varying decay rates and show that such subspaces can be used to eliminate bias up to first-order variations in the decay rates without requiring full knowledge of noise. Since such a bias cannot be corrected through standard symmetry verification, our method can improve error mitigation in dual-rail qubits and, given partial knowledge of noise, can perform better than probabilistic error cancellation. [Phys. Rev. A 110, 042621 (2024)]

In the second part, we tackle the challenges of simulating large non-unitary open quantum systems on a quantum computer. We analyze complexities for promising methods such as the Sz.-Nagy dilation and linear combination of unitaries that can simulate open systems by the probabilistic realization of non-unitary operators, requiring multiple calls to both the encoding and state preparation oracles. We propose a fully quantum method to decompose a d dimensional Kraus operator into two unitaries that can be implemented using the quantum singular value transformation algorithm, avoiding the classically expensive SVD decomposition requiring O(d^3) overhead. [Quantum 7, 1002 (2023)]

Virtual Information:
Join Zoom Meeting
https://lbnl.zoom.us/j/95626474536?pwd=IaE8mMjTDo8yYOXgN6r47yaKDFxCtS.1

Meeting ID: 956 2647 4536
Passcode: 820946

Host of Seminar:
Bert de Jong
Applied Computing for Scientific Discovery
Applied Mathematics and Computational Research Division
Lawrence Berkeley National Laboratory

Berkeley Lab – CS Seminar

Date:
Tuesday, January 7, 2025

Time:
11:00am – 12:00pm

Location:
Bldg. 50A, Room 3107  

Speakers(s) and Affiliations(s):
Andy Wan
University of California Merced

Title:
Structure-preserving discretizations and their applications

Abstract:
Many models from science and engineering possess fundamental structures which are important to preserve in order for accurate and stable long-term predictions. For instance, preserving conserved quantities, such as energy, mass and momentum, are fundamental in many physical systems. Moreover, preserving dissipative quantities, such as entropy or Lyapunov functions, are also essential for predicting correct asymptotic limits.

In this talk, we will survey a recent new class of conservative integrators, called the Discrete Multiplier Method (DMM), and its extension, Minimal Norm DMM. We will discuss various applications to many-body systems, geodesic flow, and particle methods in fluids and kinetic models. Moreover, we will showcase a promising application of DMM in high dimensional computational statistics. In particular, we will introduce Conservative Hamiltonian Monte Carlo using DMM to improve sampling efficacy of Hamiltonian Monte Carlo. If time permits, we will also dis cuss how structure-preserving can improve long-term predictions for Physics-Informed Neural Networks.

Bio:
Andy Wan is an Assistant Professor in the department of Applied Mathematics at the University of California, Merced (UC Merced). He received his Ph.D. from Polytechnique Montreal in 2014, and was a postdoctoral fellow at McGill University from 2014 to 2018. Prior to joining UC Merced in 2024, he was an Assistant Professor and later an Associate Professor at the University of Northern British Columbia in Canada. His research interests are in numerical analysis, scientific computing, and scientific machine learning. He focuses on structure-preserving discretizations, specifically in the theory and development of conservative integrators, as well as their applications to mathematical sciences, computational statistics and machine learning. He is currently a co-investigator of the 2024-2027 Collaborative Research Group on “Structure-Preserving Discretizations and their Applications” supported by the Pacific Institute for the Mathematical Sciences.

Virtual Information:
Join Zoom Meeting
https://lbnl.zoom.us/j/94079180352?pwd=taaEtXyFikIB596n65Vo0F7gDTGSL4.1

Meeting ID: 940 7918 0352
Passcode: 286639

Host of Seminar:
Andy Nonaka
Center for Computational Sciences & Engineering (CCSE)
Applied Mathematics and Computational Research Division
Lawrence Berkeley National Laboratory

Berkeley Lab – CS Seminar

Date:
Wednesday, January 8, 2025

Time:
10:00am – 11:00am

Location:
Bldg. 59, Room 4022

Speakers(s) and Affiliations(s):
Diyi Liu
Department of Mathematics
University of Minnesota Twin Cities

Title:
Toward Early Fault-Tolerant Quantum Simulation: An End-to-End Perspective

Abstract:
With recent advancements in quantum hardware and the implementation of quantum error correction codes, the prospect of developing an early fault-tolerant quantum computer within the next decade has become increasingly realistic. The potential applications of such a quantum computer, particularly in solving complex problems in quantum chemistry and quantum many-body systems, are attracting significant attention. In this talk, we aim to provide an introduction to quantum simulation from an end-to-end perspective. We will begin with an introduction to quantum computers and quantum simulation, spanning multiple scales from hardware to scientific applications. Recent progress in compilation techniques, input models, and quantum algorithms will be discussed. In particular, we will highlight recent advancements in block encoding for quantum many-body systems.

Virtual Information:
Join Zoom Meeting
https://lbnl.zoom.us/j/92916529508?pwd=cJl8XmYlmbwUjIVAADg6AqwlJICyfY.1

Meeting ID: 929 1652 9508
Passcode: 443837

Host of Seminar:
Bert de Jong
Applied Computing for Scientific Discovery
Applied Mathematics and Computational Research Division
Lawrence Berkeley National Laboratory

Berkeley Lab – CS Seminar

Date:
Thursday, January 9, 2025

Time:
10:00am – 11:00am

Location:
Bldg. 59, Room 4022

Speakers(s) and Affiliations(s):
Ahnaf Farhan
Amazon

Title:
Unifying Graphs and Text: LLM-GNN Applications Across Critical Domains

Abstract:
Large Language Models (LLMs) and Graph Neural Networks (GNNs) are revolutionizing AI by addressing challenges in both sequential and structured data. LLMs excel in capturing linguistic patterns and contextual relationships, driving advancements in text generation, machine translation, and knowledge extraction. GNNs specialize in modeling relationships within graph-structured data, such as social networks and molecular graphs. Integrating sequential data, like text, with graph-structured data presents unique challenges. This talk explores the synergy of LLMs and GNNs for tasks that combine sequential and graph-based data. In cybersecurity, I will demonstrate how these models could be utilized to bridge disconnected datasets, such as MITRE ATT&CK and Common Weakness Enumeration (CWE), offering enhanced insights for threat detection. Second, I will discuss a recently published paper introducing a multimodal framework integrating GNNs and transformer-based models to generate context-rich descriptions of protein functions.This talk highlights AI’s transformative potential in unifying text, graph, and sequence data for critical infrastructure and biological sciences.

Bio:
Dr. Ahnaf Farhan is an AI Software Development Engineer at Amazon, working on Large Language Models (LLMs) and Natural Language Processing (NLP) for code generation tasks. He earned his Ph.D. in Computer Science from the University of Texas at El Paso in 2023. He also completed a postdoctoral fellowship at UTEP focusing on integrating LLMs and Graph Neural Networks (GNNs) for cybersecurity applications. His research spans machine learning and AI, with a focus on applications in NLP and computer vision. Among his contributions, he developed a Retrieval-Augmented Generation (RAG) framework for LLMs, integrating graph structures and natural language descriptions from cybersecurity datasets to recommend system weaknesses and cyberattack patterns, enhancing cyber threat detection and mitigation. His other research works include temporal representation learning in text and vision data and predictive analysis using temporal representations. His work has been published in prestigious journals, conferences, and workshops, including Knowledge and Information Systems (KAIS), and IEEE Big Data.  

Virtual Information:
Join Zoom Meeting
https://lbnl.zoom.us/j/9619536364?pwd=WTVLb2ROenZQcTFBL0h1K2haSFpLdz09

Host of Seminar:
Kris Bouchard
Lead, Computational Biosciences
Scientific Data Division
Lawrence Berkeley National Laboratory

Please join Jonathan in person or virtually for the FY25 LDRD kick-off presentation.  

CS Area LDRD website (updated as we have more information)

If you have not used the new CS Area Website and have not signed in as staff below are the **NEW** instructions

• Go to cs.lbl.gov
• Scroll to the bottom in the dark blue section find CSA Staff Login
• Click that.
• On the next page a log in page with a [W] at the top 
• Click the “Login with OpenID Connect” (THIS IS A CRITICAL STEP) Do NOT fill in Username and password on the W page.
• Enter your LDAP this should be familiar
• Then the next odd step you will get a weird WordPress page this is normal.  
• In the upper Left corner click the words Computing Sciences
• Ok now you’re logged in as Staff
• You can go to the Staff Portal from the upper navigation and scroll down to LDRD.
• Once your logged in as staff this will hold your log in for a significant amount of time eg months… so you should not have to do this frequently.

──────────

Lisa Bruzdzinski is inviting you to a scheduled Zoom meeting.

Join Zoom Meeting
https://lbnl.zoom.us/j/95112537061?pwd=lsKJaUwRpe50eDbghps2XPKa80bbVr.1

Meeting ID: 951 1253 7061
Passcode: 743706