These characteristics are (i) the spiking behavior, integration of the incoming stimuli, detection of the activation threshold, refractory period, and generation of excitatory or inhibitory stimuli and (ii) properties related to the synaptic plasticity that include the long-term potentiation and depression, which increases or decreases the synaptic weights w s. Typically, spiking neurons (SNs) model the key characteristics of the neural cells, which are used for information processing and adaptability.
In addition to the benefit of higher biological plausibility, a hardware-based SNN offers very low energy consumption and efficient decoding of the data generated by event-based sensors. The SNN is potentially best suited for hardware implementation because it is based on parallel operation using a significant number of high complexity neurons with the simple “integrate-and-fire” feature. This is because an SNN operates based on discrete events (i.e., precise timing spikes), which make it sensitive to time-varying functions and random occurrence of events. The SNN benefits from the increased computation power and accuracy compared with other types of NNs, including the traditional ANNs. Within this context, spiking neural networks (SNNs) have emerged as the most successful approach to model the behavior and learning features of biological neural networks (NNs), and to represent and integrate information in time, space, frequency, and phase domains. Artificial neural networks (ANNs), which mimic the human brain’s process of acquisition and processing of sensory information, have received a great deal of attention for a range of applications. However, in dynamic and evolving environments, most reported models require retraining of the learned algorithms, which is not desirable. In recent years, several algorithms for machine learning-based stream learning have been reported in the literature. Furthermore, we show the axon delay as a function of the illuminance using a fitted model for which the root mean square error (RMS) similarity is 0.95. For the proposed optical axon, we have demonstrated line of sight transmission over a maximum link length of 190 cm with a delay of 8 μs. We show that the optical intensity fluctuations and link’s misalignment result in delay in activation of the synapses. The synaptic weights are encoded by the energy of the stimuli, which are then optically transmitted independently. To address this, in this paper we propose and experimentally verify optical axons with synapses activated optically using digital signals.
For systems with several neuromorphic sensors, which are connected optically to the main unit, the use of optical synapses is not an advantage. The performance of spiking neural networks has been improved using optical synapses, which offer parallel communications between the distanced neural areas but are sensitive to the intensity variations of the optical signal. In bio-inspired systems these sensors are connected to the main neural unit to perform post-processing of the sensor data. Recently, neuromorphic sensors, which convert analogue signals to spiking frequencies, have been reported for neurorobotics.