Neuromorphic Computing
- Biological inspiration: spiking neurons, synaptic plasticity, temporal coding
- Spiking neural networks (SNNs): integrate-and-fire models (LIF, IF), spike timing
- Learning in SNNs: STDP (spike-timing-dependent plasticity), surrogate gradient methods, conversion from ANNs
- Neuromorphic hardware: Intel Loihi 2, IBM TrueNorth, SpiNNaker, BrainScaleS
- Event-driven computation: asynchronous processing, energy efficiency
- Event cameras (DVS): neuromorphic vision sensors, sparse temporal data
- Applications: low-power edge inference, robotics, always-on sensing
- Comparison with conventional deep learning: latency, power, accuracy tradeoffs