Pattern representation and recognition with accelerated analog neuromorphic systems

Autor: D. Stockel, Ilja Bytschok, Dan Husmann, Mihai A. Petrovici, Johannes Bill, Johannes Schemmel, Maurice Güttler, Karlheinz Meier, Oliver Breitwieser, Wolfgang Maass, Robert Legenstein, Eric Müller, Johannes Partzsch, Anand Subramoney, Christian Mayr, Andreas Hartel, Anna Schroeder, Stephan Hartmann, Andreas Grübl, Alexander Kononov, Stefan Schiefer, Kai Husmann, Guillaume Bellec, Mitja Kleider, Johann Klähn, Stefan Scholze, Vitali Karasenko, Sebastian Schmitt, V. Thanasoulis, Christian Mauch, Paul Müller, Rene Schuffny, Bernhard Vogginger, Sebastian Jeltsch, Christoph Koke, Thomas Pfeil
Rok vydání: 2017
Předmět:
0301 basic medicine
FOS: Computer and information sciences
Computer science
Distributed computing
Central nervous system
Machine Learning (stat.ML)
02 engineering and technology
03 medical and health sciences
Robustness (computer science)
Statistics - Machine Learning
0202 electrical engineering
electronic engineering
information engineering
medicine
Neural and Evolutionary Computing (cs.NE)
Representation (mathematics)
610 Medicine & health
Artificial neural network
business.industry
020208 electrical & electronic engineering
Computer Science - Neural and Evolutionary Computing
030104 developmental biology
medicine.anatomical_structure
Neuromorphic engineering
Quantitative Biology - Neurons and Cognition
FOS: Biological sciences
Key (cryptography)
Neurons and Cognition (q-bio.NC)
Artificial intelligence
business
Zdroj: ISCAS
DOI: 10.48550/arxiv.1703.06043
Popis: Despite being originally inspired by the central nervous system, artificial neural networks have diverged from their biological archetypes as they have been remodeled to fit particular tasks. In this paper, we review several possibilites to reverse map these architectures to biologically more realistic spiking networks with the aim of emulating them on fast, low-power neuromorphic hardware. Since many of these devices employ analog components, which cannot be perfectly controlled, finding ways to compensate for the resulting effects represents a key challenge. Here, we discuss three different strategies to address this problem: the addition of auxiliary network components for stabilizing activity, the utilization of inherently robust architectures and a training method for hardware-emulated networks that functions without perfect knowledge of the system's dynamics and parameters. For all three scenarios, we corroborate our theoretical considerations with experimental results on accelerated analog neuromorphic platforms.
Comment: accepted at ISCAS 2017
Databáze: OpenAIRE
Externí odkaz: