Liu, Eric | Engineering Papers

DOE OSTI2023

High-Quality Surface Plasmon Polaritons in Large-Area Sodium Nanostructures

Sodium (Na) is predicted to be an ideal plasmonic material with ultralow optical loss across visible to near-infrared (NIR). However, there has been limited research on Na plasmonics. We develop a scalable fabrication method for Na nanostructures by combining phase-shift photolithography and a thermo-assisted spin-coating process. Using this method, we fabricated Na nanopit arrays with varying periodicities (300-600 nm) and with tunable surface plasmon polariton (SPP) modes spanning visible to NIR. We achieved SPP resonances as narrow as 9.3 nm. In addition, Na nanostructures showed line width narrowing from visible toward NIR, showing their prospect operating in the NIR. To address the challenges associated with the high reactivity of Na, we designed a simple encapsulation strategy and stabilized the Na nanostructures in ambient conditions for more than two months. As a low-cost and low-loss plasmonic material, Na offers a competitive option for nanophotonic devices and plasmon-enhanced applications.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

DOE OSTI2020

Hindsight logging for model training

In modern Machine Learning, model training is an iterative, experimental process that can consume enormous computation resources and developer time. To aid in that process, experienced model developers log and visualize program variables during training runs. Exhaustive logging of all variables is infeasible, so developers are left to choose between slowing down training via extensive conservative logging, or letting training run fast via minimalist optimistic logging that may omit key information. As a compromise, optimistic logging can be accompanied by program checkpoints; this allows developers to add log statements post-hoc, and "replay" desired log statements from checkpoint---a process we refer to as hindsight logging. Unfortunately, hindsight logging raises tricky problems in data management and software engineering. Done poorly, hindsight logging can waste resources and generate technical debt embodied in multiple variants of training code. In this paper, we present methodologies for efficient and effective logging practices for model training, with a focus on techniques for hindsight logging. Our goal is for experienced model developers to learn and adopt these practices. To make this easier, we provide an open-source suite of tools for Fast Low-Overhead Recovery (flor) that embodies our design across three tasks: (i) efficient background logging in Python, (ii) adaptive periodic checkpointing, and (iii) an instrumentation library that codifies hindsight logging for efficient and automatic record-replay of model-training. Model developers can use each flor tool separately as they see fit, or they can use flor in hands-free mode, entrusting it to instrument their code end-to-end for efficient record-replay. Our solutions leverage techniques from physiological transaction logs and recovery in database systems. Evaluations on modern ML benchmarks demonstrate that flor can produce fast checkpointing with small user-specifiable overheads (e.g. 7%), and still provide hindsight log replay times orders of magnitude faster than restarting training from scratch.

Computer Science↗

NASA NTRS1987

Deconvolution of sea state parameters from altimeter waveforms

The paper reports on an on-going effort at the JPL to estimate the accuracy of ocean state parameters which have been obtained from the specular point probability density function (pdf) of the ocean surface. This pdf is obtained by the deconvolution of the return waveform of oceanographic altimeters such as Seasat, Geosat, or Topex.

Rodriguez, Ernesto↗