Dynamic input
A dynamic input routine runs an experiment step by step: append one input block to the selected input array, deliver it, read the response, choose the next block. Test a routine on a Brian2 simulation, then switch its Stimulation target to run the same protocol on hardware.
Starting a routine
- Select the input array to extend.
- Click Design dynamic input (View toolbar or the editor's right-click menu), pick a routine and click Continue.
- Set the parameters and click Start (Start routine for STDP tuning).
If the array already holds input, choose Clear and start over or Continue after last input (hardware rheobase also offers Start in new arrays). Blocks are appended after the simulated time, never into the past, and stay in the array afterwards (undo, edit, save as usual).
One routine runs at a time: Start in another routine is refused until the running one stops. While it runs, its array stays selected and cannot be drawn on, cleared or given a pattern, and undo/redo are off.
Pause and Stop act before the next block (a running Brian2 run is not cancelled). Clear input array zeroes the array; Reset input array trims it back to the simulated time. The log records every probe and decision; live plots open in a separate window.
On a fresh model the routine first runs a short simulation to create trace
monitors. If no / v trace appears, add a StateMonitor for v in
File → Edit Model.
Stimulation targets
| Target | Routines | Amplitude means |
|---|---|---|
| Brian2 simulation | all | Input array units |
| SysCon2 (calcium imaging) | rheobase, STDP window | Fraction (0-1) of the export target power |
| BrainScaleS-2 (neuromorphic) | rheobase | Fraction (0-1) of the tab's i_offset max |
For SysCon2, Hardware settings sets power, laser, spots, recording tail (3 s), artefact removal and trace extraction (Suite2p or ROI average). With Run blocks automatically and record them with ScanImage (default) each block is run, recorded and extracted for you. Unticked, you run, record and import each block, then click Calcium data imported (rheobase) or Block done / data imported (STDP window). Use SysCon2 dt sets the array timestep to the SysCon2 grid. Details: Holographic stimulation, ScanImage acquisition, BrainScaleS-2.
Minimal input for spiking (rheobase)
Finds the smallest input that makes one neuron spike, by bisection rather than a full grid scan. Each probe is a recovery period, a pulse, and a post-pulse window in which spikes are counted. The search tests the maximum amplitude (no spike: stop, raise it), then the minimum (spike: threshold is at or below it), then halves the bracket until it is below the tolerance or Maximum iterations (25) is reached.
Sweeping strategy:
- Amplitude and duration: rheobase at the longest duration, then the chronaxie (duration at twice the rheobase) by bisecting duration.
- Fixed duration, varying amplitude: threshold at one duration.
- Fixed amplitude, varying duration: shortest spiking duration.
| Parameter | Brian2 | SysCon2 |
|---|---|---|
| Amplitude range / tolerance | 1-100 / 0.05 | 0.1-1.0 / 0.05 |
| Fixed pulse amplitude | 10 | 0.5 |
| Pulse duration range / tolerance | 10-50 ms / 1 ms | 16.7-100 ms / 1 ms |
| Recovery time | 200 ms | 1000 ms |
| Post-pulse window | 20 ms | 300 ms |
| Repeats per probe | 1 | 3 |
With several repeats a probe counts as a response when most trials respond. Brian2 counts crossings of Spike threshold (-20 mV). SysCon2 uses the Calcium readout: a ΔF/F rise of at least Response threshold (20 %), or OASIS-deconvolved spikes above Spike threshold (0.1 ΔF/F). BrainScaleS-2 starts with the hardware amplitudes and repeats and counts the chip neuron's spikes; connect from the dialog or the BrainScaleS-2 tab.
STDP timing window (LTP/LTD vs Δt)
Sweeps Δt = t_post − t_pre and measures the synaptic change at each point. Use a suprathreshold amplitude (the rheobase routine can find one).
- Readout: EPSP amplitude change (%) (five presynaptic probe pulses after each point) or Synaptic weight change (%) (reads the synapse weight directly; use it on recurrent networks).
- Sweep mode: Reset to baseline each Δt restores a Brian2 checkpoint before every point. Continuous sweep keeps one synapse and visits Δt by increasing magnitude, alternating sign. SysCon2 is always continuous, with a calcium readout.
Defaults: amplitude 100 (0.8 on SysCon2), pulse 5 ms, Δt -100 to 100 ms in 21 points (Δt = 0 excluded), 1 pairing per point, recovery 150 ms, post-pulse window 50 ms.
STDP synaptic tuning
Drives one synapse's EPSP to a target relative to its initial size (Brian2
only). Try it in File → Open example project → Network → Two Hodgkin-Huxley
Neurons with STDP with array Exc_stimulation, trace Simulation :: P / v,
presynaptic 0, postsynaptic 1.
- Five presynaptic probe pulses, 100 ms apart; their mean EPSP is the 100 % baseline.
- Outside the Tolerance window (1 %) of Target relative EPSP amplitude (120 %), append one pairing: pre before post below target, post before pre above.
- Probe again; repeat until within tolerance or after Maximum plasticity induction iterations (20). Run until the last iteration regardless of convergence keeps pairing.
Pairing strategy: Fixed pre-post pairing time (30 ms) uses the same interval each time. Adaptive (single parameter, dt increment) starts at 30 ms and shrinks the interval by dt increment (1 ms) when the remaining error exceeds the last EPSP change, otherwise grows it (between one sample and 1000 ms).
Pulses default to 100 (array units) for 5 ms. EPSPs are measured with the EPSP amplitudes plugin; the routine stops if the trace lacks the postsynaptic neuron or an EPSP is missed. It assumes excitatory EPSPs and pre-before-post potentiation, and timings snap to the array timestep.