The synaptic network was connected to an Arduino Due μC on a PCB for the training and testing of the synaptic network. To operate the network, the PRE spike sequence was first stored in the internal memory of the μC, and then the sequence was launched while monitoring the synaptic weights G and internal potential Vint at each cycle. The spike and fire potential and input currents were also monitored by a LeCroy WaveRunner oscilloscope with 600-MHz bandwidth and maximum 4 G sample/s sampling rate. Note that the μC is only necessary in providing spiking information (including the teacher signal) during the training and test stage, whereas all learning processes, that is, the adjustment of synaptic, were all achieved by the network of hybrid CMOS-neurons/resistive switching synapses in real time. For best accuracy in our PCB system, we adopted an axon potential decay constant τ = 8 ms, that is, longer than the biological action potential of about 1 ms. To match with the biological timescale (τ = 0.5 ms), the experimental timescale in Fig. 6 (C to F) was scaled down by a factor of 16. The time delay between the synaptic current detection and the update spike for the TE of the 1T1R synapse is about 12 μs to enable in situ learning.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.