Quick Start

Build a dendritic schedule in the browser, then continue with the simulation and learning examples.

Why scheduling matters

The Hines matrix follows the topology of a neuron. During triangularization, a compartment can run only after its children. During back-substitution, it can run only after its parent.

DHS groups independent compartments into execution steps, exposing the maximum safe parallelism without changing the result.

Tree morphology

Threads per step: 4

DENDRITIC TREEclick a node to inspect

EXECUTION SCHEDULEready

NODES0

STEPS0

UTILIZATION0%

Choose a morphology and thread count, then run DHS to generate dependency-safe execution levels.

Run DHS in DeepDendrite

STEP 1

Prepare the model

Use the modified NEURON frontend to prepare morphologies, mechanisms, and network inputs.

cd src/nrn_modify
./configure --prefix ../../install \
  --without-iv --with-paranrn \
  --with-nrnpython=`which python`
make -j8
make install

Installation guide →

STEP 2

Select DHS

Run the GPU engine with permutation type 3 and choose the thread count for each cell.

./install/bin/deepdendrite \
  --datpath /path/to/coredat --gpu \
  --cell-permute 3 --cell-nthread 16

Runtime guide →

STEP 3

Open an example

Start with equation solving, full-spine simulation, or HPC-Net training.

cd Figure5/run_samples
python3 run.py

View examples ↓

Examples

EQUATION SOLVING

Equation-solver benchmark

Compare DHS thread configurations and solver performance across morphologies.

Open on GitHub ↗

FULL-SPINE SIMULATION

Spine-level human neuron

Run clustered and distributed input experiments with explicit dendritic spines.

Open on GitHub ↗

DETAILED-NEURON LEARNING

HPC-Net classification

Train or test the Human Pyramidal Cell Network on MNIST image classification.

Open on GitHub ↗