UPLOADING / CONNECTOME WORKSHOP
MaleCNS.
Your task.
Build custom input-to-action loops on a real MaleCNS graph.
Run inference locally. Define your own inputs and readouts.
Fit a task readout. Save it. Use it in your own application.
BUILD WITH CNSKIT
A small API.
Your own task loop.
Start with an example, then bring your observations and targets. Run locally with Python; no cloud account required.
Read the technical documentation ↗Repository access currently requires an invitation. Open-source licensing is being prepared.
01 Run your first loop
python -m pip install -e .
python examples/local_task.pyA synthetic example to verify installation before downloading the anatomical graph.
02 Fit a readout
from cns_tinker.readout import RidgeReadout
model = RidgeReadout.fit(
train_activity, train_targets,
feature_names=channels,
output_names=["steer"], alpha=1.0)
model.save("steering.json")Local ridge regression. The graph stays frozen. You provide labels and separate evaluation episodes.
Training guide ↗03 Use your mapping
model = RidgeReadout.load("steering.json")
action = model.predict(
activity, feature_names=channels)Named channels prevent silent reordering. Your application turns continuous outputs into actions.
Readout fitting is implemented. Connectome fine-tuning, automatic task training and GPU execution are not.
TASK SPECIFICATION
Define your integration.
A task specification you can take with you.
Exports configuration; does not execute or train a task.
Specify observations
and control outputs.
01 / YOUR TASK CONTRACTInspect generated JSON + SDK handoff ↗
01 / RUNNABLE EXAMPLES
Your task.
An explicit input-to-action loop.
Turn observations into actions.
Run a synthetic recipe locally, then change its adapters.
Developer reference / existing sensory recordings Open traces ↗
A shift
in the field.
An expanding input.
A changing internal state.
Inspect the underlying signals Input · state · readout ↗
Recorded synthetic dynamics, not biological spikes. Real MaleCNS wiring is not loaded in this preview.
02 / A SMALL SURFACE AREA
Install locally.
Configure a recipe.
Read the outputs.
Keep the world you built. Use the Python SDK to configure a scenario, run the local scaffold and export its evidence.
Encode your inputsExplicit sensory adapters, defined by your task.
Run and inspectFollow state and decoded actions through a run.
Take the evidence with youConfiguration, traces and a truth ledger.
Local execution AVAILABLE · training not implemented
python -m pip install -e .03 / BRING A WORLD
Example tasks.
Explicit limitations.
Start with a recipe. Make the mapping explicit.
Inspect the implementation.
BUILT TO BE INSPECTED
Every signal has a story.
Keep the receipts.
Versioned inputs. Explicit dynamics. Traceable outputs.
Connectome-based experiments deserve evidence you can examine.