rtus
- North America > United States (0.28)
- North America > Canada > Alberta (0.14)
- Europe > Portugal > Braga > Braga (0.04)
- Research Report > Experimental Study (0.93)
- Research Report > New Finding (0.92)
Real-Time Recurrent Learning using Trace Units in Reinforcement Learning
Elelimy, Esraa, White, Adam, Bowling, Michael, White, Martha
Recurrent Neural Networks (RNNs) are used to learn representations in partially observable environments. For agents that learn online and continually interact with the environment, it is desirable to train RNNs with real-time recurrent learning (RTRL); unfortunately, RTRL is prohibitively expensive for standard RNNs. A promising direction is to use linear recurrent architectures (LRUs), where dense recurrent weights are replaced with a complex-valued diagonal, making RTRL efficient. In this work, we build on these insights to provide a lightweight but effective approach for training RNNs in online RL. We introduce Recurrent Trace Units (RTUs), a small modification on LRUs that we nonetheless find to have significant performance benefits over LRUs when trained with RTRL. We find RTUs significantly outperform other recurrent architectures across several partially observable environments while using significantly less computation.
- North America > United States (0.28)
- North America > Canada > Alberta (0.14)
- Europe > Portugal > Braga > Braga (0.04)
- Information Technology > Artificial Intelligence > Machine Learning > Reinforcement Learning (1.00)
- Information Technology > Artificial Intelligence > Machine Learning > Neural Networks > Deep Learning (0.86)
- Information Technology > Artificial Intelligence > Machine Learning > Learning Graphical Models > Undirected Networks > Markov Models (0.46)
Encycle Joins Distech Controls' Digital Partner Program
Encycle Corporation, a software technology company focused on helping commercial enterprise-level utility customers dramatically improve the efficiency of their HVAC systems using IoT-enabled services, announced that it has joined Distech Controls' Digital Partner Program (DPP). Participation in the program will help link Encycle's Swarm Logic energy-saving software with commercial and industrial customers looking to reduce HVAC-related energy consumption, costs, and emissions through trusted, best-in-class solutions. The DPP brings together a network of world-class digital companies that share their expertise, technologies, and best practices to help make buildings more efficient, connected, and attractive. Distech Controls selected Encycle as a DPP partner based on its IoT-centered technology, complementarity with other program partners, technological openness (Swarm Logic is available on Tridium's Niagara Framework), and collaborative practices. "We are proud to partner with Distech Controls to tackle difficult HVAC energy challenges and mutually deliver proven energy and decarbonization results through our patented, Energy-as-a-Service approach to energy management," stated Steve Alexander, Encycle President and CEO.
- Information Technology (1.00)
- Energy (1.00)
- Construction & Engineering > HVAC (1.00)
- Information Technology > Communications > Web (0.40)
- Information Technology > Artificial Intelligence (0.34)