2019 Fiscal Year Research-status Report
Development of Human-timescale Neural Circuits using Emerging Neuromorphic Devices
Project/Area Number |
19K21083
|
Allocation Type | Multi-year Fund |
Research Institution | National Institute of Advanced Industrial Science and Technology |
Principal Investigator |
Stoliar Pablo 国立研究開発法人産業技術総合研究所, エレクトロニクス・製造領域, 主任研究員 (40824545)
|
Project Period (FY) |
2019-04-01 – 2021-03-31
|
Keywords | Neuromorphic systems / Emerging devices / Recurrent neural net. / Homeostasis mechanisms / Ferroelec. tunnel junct. |
Outline of Annual Research Achievements |
During FY31, I did 4 types of research (3 papers published, 1 in 2nd round of review): 1) I developed type II neurons compatible with emerging device physics. Type I neurons produce a fixed frequency for a fixed input. In type II, the frequency reduces in time. 2) I did a proof-of-concept for stabilisation; it is a system (11 neurons) that mimics how the brain combines signals from ears. The stabilization is to compensate for drifts. 3) I synchronize the firing of FTJ-based neurons, determined that a dynamic range > 10 is required in RS synapses, and developed VLSI compatible firing and output circuits. 4) I developed an ADQ system to study neurons. I used a 32-bit ARM core microcontroller (programmed in C++) connected to a PC (programmed in LabVIEW). I developed an ad-hoc analog frontend.
|
Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
I proposed 4 types of research, namely: 1) Development and study of Type II neurons. 2) Learning stabilisation in neuromorphic systems. 3) Study i) synchronisation of neurons, ii) dynamic range requirements for the synapses, iii) Alternative firing circuits for the neurons, iv) Alternative circuits to generate the output pulses of the neurons. 4) Implement a hardware interface to test the neurons. I completed 100% of these 4 types of research.
|
Strategy for Future Research Activity |
For the promotion of the result, I am preparing the following set of publications based on the results: 1) Long time-scale neurons based on emerging research devices (completed 75%), 2) neuromorphic system mimicking auditory system (completed 25%), 3) implementation of neurons in VLSI (completed 75%), and 4) neuron response to random signals (completed 25%). Also, from a collaboration with UCSD, I recently obtained VO2 devices. I will try to use it to develop neurons using the concepts of this project.
|
Causes of Carryover |
This money is for publication expenses of the last paper. The review process was delayed. The editor struggled to find a 2nd reviewer. To the time of writing this report (5/27/2020), the publication was already accepted and the payment was made (USD2950.00). I will use the remaining money (yen 27190) to buy components to build a VO2-based neuron.
|