The workshop will be held virtually on Sun, Jul 19. It is free. Please sign up here.
Recent advances in machine learning algorithms and hardware acceleration have revolutionized the field of artificial intelligence (AI) and have led to machines outperforming humans in cognitive tasks where humans have prevailed so far. However, there still remains a huge gap between humans and machines in terms of energy-efficiency, robustness, and lifelong learning capabilities. Neuromorphic computing, an interdisciplinary field at the intersection of neuroscience, computer science, computer engineering, and material science, has the potential for further breakthroughs in machine intelligence. For example, new findings in neuro- and cognitive science can inspire novel supervised/unsupervised learning algorithms or novel spiking neural models with improved accuracy and robustness. Meanwhile, breakthroughs at the level of nanoelectronic devices and materials can lead to advanced neuro-mimetic features, better performance, and improved energy efficiency of such systems.
The objective of this workshop is to bring together researchers from multiple disciplines, ranging from physical to biological sciences, to discuss the most promising approaches and overarching goals of neuromorphic computing technologies and paradigms that have the potential to drastically improve conventional approaches. These may include, self-organizing neuromorphic systems for robust and general artificial intelligence, novel architectures and systems with systemic theories justifying spike-based computing, and emerging devices and biological substrates with complex dynamics that can provide orders of magnitude higher energy efficiency. Exemplary research efforts from both academia and industry will be presented. The neuromorphic computing workshop aims to establish a forum to discuss
- the current practices;
- future research needs; and
- discussions for investigating new principles and tradeoffs across the entire neuromorphic information processing stack with the goal to apply them holistically to future machine learning systems.
