{"id":1878,"date":"2020-03-25T19:11:47","date_gmt":"2020-03-25T19:11:47","guid":{"rendered":"https:\/\/www.rtc.us.es\/?page_id=1878"},"modified":"2020-04-08T16:05:50","modified_gmt":"2020-04-08T16:05:50","slug":"small","status":"publish","type":"page","link":"https:\/\/grupo.us.es\/rtclab\/neuromorphic-engineering\/small\/","title":{"rendered":"SMALL"},"content":{"rendered":"\r\n

Spiking Memristive Architectures for Learning to Learn<\/h2>\r\n\r\n\r\n\r\n

Contemporary AI applications often rely on deep learning, which implies heavy computational\u00a0loads with current technology. However, there is a growing demand for low-power autonomously learning AI systems that are employed \u201cin the field\u201d. We will investigate in this project options for learning in low-power unconventional hardware that is based on spiking neural networks (SNNs) implemented in analog neuromorphic hardware combined with nano-scale memristive synaptic devices. Hence, the envisioned computational paradigm combines the three most promising avenues for minimizing energy consumption in hardware:<\/p>\r\n\r\n\r\n\r\n

    \r\n
  1. \u00a0 analog neuromorphic computation,\u00a0<\/li>\r\n
  2. \u00a0 spike-based communication, and\u00a0<\/li>\r\n
  3. \u00a0 memristive analog memory.\u00a0<\/li>\r\n<\/ol>\r\n\r\n\r\n\r\n

    Experts in each of these fields will collaborate on the development of a functional prototype system. We will in particular consider recurrent SNNs (RSNNs) as their internal recurrent dynamics render them more suitable for real-world AI applications that have temporal input and demand some form of short-term memory. We will adapt a recently developed training algorithm such that it can be used to optimize SNNs in neuromorphic hardware with memristive synapses. \u201cIn the field\u201d applications often demand online adaptation of such systems, which often necessitates hardware-averse training procedure.<\/p>\r\n\r\n\r\n\r\n

    To overcome this problem, we will investigate the applicability of \u201clearning to learn\u201d (L2L) to spiking memristive neuromorphic hardware. In an initial optimization, the hardware is trained to become a good learner for the target application. Here, arbitrarily complex learning algorithms can be used on a host system with the hardware \u201cin the loop\u201d. In the application itself, simpler algorithms \u2013 that can be easily implemented in neuromorphic hardware \u2013 provide adaptation of the hardware RSNNs.\u00a0<\/p>\r\n\r\n\r\n\r\n

    In summary, the goal of this project is to build versatile and adaptive low-power small size neuromorphic AI machinery based on SNNs with memristive synapses using L2L. We will deliver an experimental system in a real-world robotics environment to provide a proof of concept. The ED-Scorbot is being used as test-bed (http:\/\/www.rtc.us.es\/ed-scorbot\/<\/a>).<\/p>\r\n\r\n\r\n\r\n

    \r\n
    \"\"<\/figure>\r\n<\/div>\r\n\r\n\r\n\r\n

    Consortium:<\/strong><\/p>\r\n\r\n\r\n\r\n