Organizers: | |||
|---|---|---|---|
Abstract: | In this session, we will discuss theoretical frameworks for building principled neuromorphic algorithms and the features that make an ideal framework for Loihi. We will discuss efforts to develop a robust and flexible system that can go beyond networks that perform just a single task. As well, we will cover how proposed frameworks might synthesize into a high-level abstraction paradigm for building general purpose algorithms. | Speakers | . How can these frameworks contribute to a language to formalize and optimize neuromorphic computing algorithms? How can neuromorphic algorithms be systematized, modularized, combined and reused? Due to time, please only ask understanding questions during and between talks, and save questions for the general discussion at the end. |
Speakers | 8:00-8:08: Friedrich Sommer: Taxonomy of conceptual frameworks and neuromorphic VSA 8:08-8:16: Paxon Frady: Frameworks for efficient coding and computation on neuromorphic hardware 8:16-8:24: Gregor Schöner: Dynamic Field Theory as a theoretical framework for neuromorphic embodied cognition 8:24-8:32: Christian Tetzlaff: The usage of on-chip learning for universal computing 8:32-8:40: Hyeryung Jang: Information-Theoretic Principles for Neuromorphic Computing: Information Bottleneck and Bayesian Learning 8:40-8:48: Gregor Lenz: Efficient ANN-SNN conversion 8:48-9:00: General Discussion and Questions | ||
Reference material | https://www.pnas.org/content/pnas/116/36/18050.full.pdf https://redwood.berkeley.edu/wp-content/uploads/2020/08/resonator1.pdf | ||
Recording | Not yet available | ||
Link to Presentation | Not yet available |
...