AI-ordlista
Den kompletta ordlistan över AI
Dynamic Vision Sensor (DVS)
Neuromorphic sensor that asynchronously captures intensity changes, generating spatio-temporal events instead of full-frame images at fixed frame rates.
Asynchronous Events
Discrete signals generated by DVS when pixels detect an intensity change exceeding a threshold, each containing spatial coordinates, a timestamp, and a polarity.
Leaky Integrate-and-Fire Neuron
Fundamental neuronal model that integrates weighted inputs into its membrane potential with temporal leakage, generating a spike when the threshold is reached.
Spike-Timing-Dependent Plasticity (STDP)
Local learning mechanism where synaptic strength is modulated based on the temporal offset between pre- and post-synaptic spikes, implementing temporal Hebbian learning.
Neuromorphic Computing
Hardware and software design approach inspired by brain architecture, optimizing energy consumption and latency for real-time sensory data processing.
Event-Based Vision
Computer vision paradigm based on processing asynchronous event streams rather than traditional images, offering sub-millisecond latency and superior energy efficiency.
Membrane Potential
Internal electrical state of a spiking neuron that evolves according to the integration of synaptic inputs and passive leakage, determining spike generation timing.
Event-Based Sampling
Data acquisition method where measurements are triggered only when significant changes occur, drastically reducing redundancy and data volume.
Spatio-Temporal Demultiplexing
Process of reconstructing coherent visual information from disordered event streams by exploiting spatial and temporal correlations inherent in DVS data.
Spike-Frame Conversion
Technique for transforming asynchronous spike streams into synchronous matrix representations for integration with traditional CNN architectures or batch analysis.
Asynchronous Parallel Architecture
Computational structure where processing units operate independently and communicate via asynchronous events, maximizing parallelism and minimizing latency.
Spatio-Temporal Hierarchy
Multilayer organization of spiking neurons where lower layers capture fast and local patterns while higher layers integrate more global and slower information.
First-Spike Coding
Representation strategy where information is primarily contained in the arrival time of the first spike after a stimulus, offering optimal latency for rapid recognition.
Neuronal Resonance
Phenomenon where spiking neurons synchronize their activities to specific frequencies of the input signal, facilitating the detection of periodic patterns in event streams.