An official website of the United States government
This content will become publicly available on April 1, 2026
Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn to Explore for Real-World RL
- Award ID(s):
- 2308979
- PAR ID:
- 10580432
- Editor(s):
- Globerson, A; Mackey, L; Belgrave, D; Fan, A; Paquet, U; Tomczak, J; Zhang, C
- Publisher / Repository:
- Curran Associates, Inc.
- Date Published:
- Issue:
- 37
- Page Range / eLocation ID:
- 78715-78765
- Format(s):
- Medium: X
- Location:
- Advances in Neural Information Processing Systems 37 (NeurIPS 2024)
- Sponsoring Org:
- National Science Foundation
More Like this
-
Program optimization is the process of modifying software to execute more efficiently. Superoptimizers attempt to find the optimal program by employing significantly more expensive search and constraint solving techniques. Generally, these methods do not scale well to programs in real development scenarios, and as a result superoptimization has largely been confined to small-scale, domain-specific, and/or synthetic program benchmarks. In this paper, we propose a framework to learn to superoptimize real-world programs by using neural sequence-to-sequence models. We created a dataset consisting of over 25K real-world x86-64 assembly functions mined from open-source projects and propose an approach, Self Imitation Learning for Optimization (SILO) that is easy to implement and outperforms a standard policy gradient learning approach on our dataset. Our method, SILO, superoptimizes 5.9% of our test set when compared with the gcc version 10.3 compiler’s aggressive optimization level -O3. We also report that SILO’s rate of superoptimization on our test set is over five times that of a standard policy gradient approach and a model pre-trained on compiler optimization demonstration.more » « less
-
Existing models used in real-time scheduling are inadequate to take advantage of simultaneous multithreading (SMT), which has been shown to improve performance in many areas of computing, but has seen little application to real-time systems. The SMART task model, which allows for combining SMT and real time by accounting for the variable task execution costs caused by SMT, is introduced, along with methods and conditions for scheduling SMT tasks under global earliest-deadline-first scheduling. The benefits of using SMT are demonstrated through a large-scale schedulability study in which we show that task systems with utilizations 30% larger than what would be schedulable without SMT can be correctly scheduled. This artifact includes benchmark experiments used to compare execution times with and without SMT and code to duplicate the reported schedulability experiments.more » « less
-
The quantum-electrodynamical time-dependent density functional theory equations are solved by time propagating the wave function on a tensor product of a Fock-space and real-space grid. Applications for molecules in cavities show the accuracy of the approach. Examples include the coupling strength and light frequency dependence of the energies, wave functions, optical absorption spectra, and Rabi splitting magnitudes in cavities, as well as a description of high harmonic generation in cavities.more » « less
-
As the development of extended reality technologies bring us closer to what some call the metaverse, it is valuable to investigate how our perception of color translates from physical, reflective objects to emissive and transparent virtual renderings. Colorimetry quantifies color stimuli and color differences, and color appearance models account for adaptation and illuminance level. However, these tools do not extent satisfactorily to the novel viewing experiences of extended reality. Ongoing research aims to understand the perception of layered virtual stimuli in optical see-through augmented reality with the goal of improving or extending color appearance models. This will help ensure robust, predictable color reproduction in extended reality experiences.more » « less
