Institution profile

Los Alamos National Laboratory

Academic institutionnorthamerica · us
Official website
Research library280linked papers
Opportunities0open roles
Selected work

Representative Papers

Randomized adiabatic quantum linear solver algorithm with optimal complexity scaling and detailed running costs

May 19, 2023

To address the suboptimal complexity $O(kappa cdot ext{poly}(1/varepsilon))$ of existing quantum linear solvers with respect to condition number $kappa$ and precision $varepsilon$, this paper proposes the first randomized adiabatic quantum algorithm achieving optimal complexity $O(kappa log(1/varepsilon))$. Methodologically, it replaces costly Hamiltonian simulation with a low-overhead random-walk operator, thereby eliminating classical preprocessing bottlenecks; integrates Poissonization-based sampling, block-encoding techniques, optimized circuit design, and a novel fused filtering scheme to enhance hardware compatibility. Theoretically, we derive a tight closed-form upper bound on runtime: for Hermitian matrices, only approximately $867kappa$ block-encoding queries are required at $varepsilon = 10^{-10}$—yielding an exponential improvement in precision dependence over prior art and constant-factor hardware acceleration.

19 citations4 influentialRead paper

Teacher-student training improves accuracy and efficiency of machine learning inter-atomic potentials

Feb 07, 2025

To address the high computational cost and memory footprint of machine-learned interatomic potentials (MLIPs) in large-scale molecular dynamics (MD) simulations, this work introduces knowledge distillation to MLIP training for the first time, proposing a teacher–student collaborative framework. A high-accuracy teacher model provides implicit supervision via atomic energy predictions, guiding the training of a lightweight student model with a customized, resource-efficient architecture. Remarkably, the student achieves superior accuracy to the teacher under identical training data. Experiments on benchmarks including QM9 demonstrate a 12% reduction in mean absolute error (MAE), a 2.3× speedup in MD simulation throughput, and a fivefold reduction in memory consumption relative to the teacher. The core contribution lies in adapting the knowledge distillation paradigm to the MLIP domain, enabling simultaneous optimization of predictive accuracy, computational efficiency, and memory efficiency.

3 citationsRead paper

Towards Million-Server Network Simulations on Just a Laptop

May 26, 2021arXiv.org

To address the challenges of assessing non-shortest-path diversity in large-scale interconnection networks and the poor scalability of conventional packet-level simulators, this paper proposes a lightweight simulation framework tailored for extreme-scale networks. By identifying memory and event-scheduling bottlenecks in mainstream simulators, we introduce three core techniques: compact data structures, lazily bound event queues, and lock-free memory pools—significantly reducing both memory footprint and synchronization overhead. Our framework enables fine-grained, packet-level simulation of data center and HPC networks with over one million endpoints on a single commodity laptop, achieving a throughput of 10 million packets per second—three orders of magnitude higher than state-of-the-art shared-memory simulators. The open-source framework supports rapid prototyping and validation of novel interconnect protocols, providing a reproducible, high-fidelity foundation for path diversity analysis and performance optimization in ultra-large-scale networks.

3 citationsRead paper

Fast Risk Assessment in Power Grids through Novel Gaussian Process and Active Learning

Aug 15, 2023

To address the lack of formal performance guarantees for machine learning in safety-critical power system applications, this paper proposes a graph-structured Gaussian process (GP) method for voltage-constraint risk assessment. We introduce a novel vertex-degree kernel (VDK) that explicitly encodes topological dependencies between voltages and loads on the power grid, and design an active learning strategy aligned with the additive structure of VDK. Theoretically, we prove that the risk estimation error of the VDK-GP matches that of the AC power flow model and establish, for the first time, a statistically grounded probabilistic error bound for graph-based stochastic modeling. Evaluations on 500- and 1354-bus systems demonstrate over a twofold reduction in sample complexity, more than 15× speedup in computation versus Monte Carlo simulation, and risk estimation errors at the 10⁻⁴ level.

2 citationsRead paper

Quantum Circuit Pre-Synthesis: Learning Local Edits to Reduce $T$-count

Jan 27, 2026

This work addresses the challenge of high T-count in Clifford+T circuits, which significantly increases resource overhead in fault-tolerant quantum computing. Existing local synthesis methods are constrained by circuit representations and struggle to achieve optimal T-count and depth. To overcome this limitation, the paper introduces Q-PreSyn, the first approach to integrate reinforcement learning into the pre-synthesis phase of quantum circuit compilation. By training an agent to learn sequences of function-preserving local editing operations, Q-PreSyn produces circuit representations that are more amenable to downstream synthesis. Without introducing any approximation error, the method achieves up to a 20% reduction in T-count compared to state-of-the-art techniques on circuits with up to 25 qubits, substantially improving synthesis efficiency.

1 citationsRead paper
Recent publications

Latest Papers