MARCS Monday Meeting - 30 March 2019 - Dr Moritz Milde and Dr Ying Xu

Join us at the next MMM with invited speaker Dr Moritz Milde as he discusses The Role of Time in Neural Computation, and also hear from Dr Ying Xu in her talk about Event-based Feature Extraction and Classification using Neuromorphic Cochlea.

NOTE: MMM talks will take place via Zoom, however it will also be streamed to Room 3.G.55.

Topic: The Role of Time in Neural Computation
Speaker: Dr Moritz Milde

Nothing we experience is static. As time progresses, our senses continuously capture the rich features of our environment which our brain transforms into abstract concepts. The continuous nature of time leads to causal relations between stimuli and thus might be the key to construct a coherent impression of reality.

In this talk, we focus on the role of time in sensory processing. Our research suggests that precise spike timing can be used to extract relative motion and learn sequences of spatio-temporal features. We will further show that precise spike-timing allows performing temporal predictions, a fundamental property to implement predictive coding.

Dr Moritz Milde is a new member of the ICNS team at the Werrington South Campus. Dr Milde joined ICNS from the University of Zurich where he completed his PhD in Event-based computation & algorithm development. Dr Milde is interested in transferring insights of biological information processing into start-of-the-art software and hardware.

Topic: Event based Feature Extraction and Classification using Neuromorphic Cochlea
Speaker: Dr Ying Xu

This work presents a biologically inspired digital implementation of an event-based binaural cochlear system using a widely available Field Programmable Gate Array (FPGA). It consists of a pair of Cascade of Asymmetric Resonators with Fast Acting Compression (CAR FAC) cochlear models and a stochastic Leaky Integrate and Fire (LIF) neuron model. We have implemented a 70×2×9 channel cochlear system at 44.1 kHz sampling rate on an Altera Cyclone V FPGA board using time multiplexing and pipeline parallelising techniques. Additionally, we propose and investigate an auditory Event-based Feature Extraction Using Adaptive Selection Thresholds (FEAST) approach to extract acoustic features embedded in the auditory spike streams. The feature extraction method is tested on TIDIGTIS benchmarking dataset. 

Dr Ying Xu was awarded her PhD by Western Sydney University in 2019 for A Digital Neuromorphic Auditory Pathway. Ying is currently a postdoctoral fellow working with the International Centre for Neuromorphic Systems team. Her research focus is on auditory neuromorphic systems, application-specified integrated circuit/field- programmable gate array design, and machine learning.

Please note, Room 3.G.55 will be set up for those wanting to attend in person.

The zoom ID is: 627 146 998. Link to zoom https://uws.zoom.us/j/627146998