Neuromorphic computing is one proposed computing architecture for the beyond Moore's Law computing landscape. Neuromorphic computers are software/hardware systems that have implementations with features inspired by biological brains. Key research questions associated with neuromorphic computing are (1) how to program or train these architectures to perform tasks and (2) what supporting software is required in order to integrate these architectures into real systems and make them accessible to novice users. The goal of this project is to develop algorithms for training neuromorphic computers and supporting software for neuromorphic computers, including development environments, visualization tools, and hardware simulators.
Filter Projects
Area of Research
There has been a recent surge of success in utilizing Deep Learning (DL) in imaging and speech applications for its relatively automatic feature generation and, in particular for convolutional neural networks (CNNs), high accuracy classification abilities. While these models learn their parameters through data-driven methods, model selection (as architecture construction) through hyper-parameter choices remains a tedious and highly intuition driven task. To address this, Multi-node Evolutionary Neural Networks for Deep Learning (MENNDL) is proposed as a method for automating network selection on computational clusters through hyper-parameter optimization performed via genetic algorithms.
EDEN is a visual analytics tool for exploratory analysis of multivariate data sets. Based on an interactive variant of parallel coordinates, EDEN includes statistical analytics that guide the user to significant associations in complex data sets without information loss.
While the term ‘innovation ecosystem’ is often utilized, the concept is rarely quantified. Oak Ridge National Lab conducted a ground-breaking application of natural language processing, link analysis and other computational techniques to transform text and numerical data into metrics on clean energy innovation activity and geography for the U.S. Department of Energy. The project demonstrates that a machine-assisted methodology gives the user a replicable method to rapidly identify, quantify and characterize clean energy innovation ecosystems.