STRATA
Skip to study
XLVII

VI.14 / BIOLOGICAL

Bio-inspired engineering and neuromorphic computation

At a common task accuracy, how do energy use, latency and performance under noise compare across neuromorphic and conventional implementations?

A source-led introduction1 source references1 related experiments

Start with the question.

Bio-inspired engineering extracts a specific principle from a living system and tests its usefulness in a designed one. Mead's neuromorphic electronics provides a historical example of organizing electronic computation around features of neural systems. The comparison becomes meaningful through a defined task and resource measure. Similar spikes or circuit shapes alone do not establish that two systems process information in the same way.

Compare event-driven and regularly sampled representations.

Use a slowly changing signal interrupted by a few rapid transitions. Count the samples in a fixed-rate representation and the events in a threshold-triggered representation. Inspect what each loses when the threshold or sampling rate changes, then compare the computation needed to recover a task-relevant feature.

What would you inspect?

  • Event rate, temporal precision, energy or operation count, and task accuracy.
  • The biological principle transferred and the engineering constraints that modify it.

Take the next step.

Choose one neuromorphic task and compare a documented event-based method with a conventional baseline using the same input and success criteria.

Put part of the question to work.

These models isolate a mechanism you can vary. Their assumptions define the part of this topic they can illuminate.

Try sampling and aliasing.

Change one parameter, compare the numerical output, and export the result. The experiment loads when you open it.

Open its full page

Watch the quantities change.

12 seconds · Original model visualization · Silent

Sampling and aliasing

See how different continuous signals can pass through the same samples.

Read the explanation and method

The cream curve is an 18 hertz sinusoid. Orange dots show measurements taken at the displayed sampling rate.

The orange curve is a signed lower-band candidate that passes through the same sample values. Agreement at the dots cannot establish what happened between them.

The sampling rate rises above 36 samples per second. With the stated bandlimit, sampling strictly above twice the signal frequency removes this ambiguity.

Method. For a sampling rate fₛ, adding an integer multiple of fₛ to a sinusoid’s frequency leaves its samples unchanged. Fold the signed frequency into [−fₛ/2, fₛ/2) and evaluate both curves at identical sample times.

Read the supporting material.

  1. Neuromorphic electronic systems

    Carver Mead / Caltech · Analog electronic systems inspired by neural organization and computation.

Continue through the versioned study records.

Follow the comparisons.

Read the neighboring questions.