Dendritic Hierarchical Scheduling
Compartment-level scheduling for tree-structured neuron equations on GPUs.
Read the architecture guide →THE PROJECT
About
DeepDendrite connects detailed neuron modeling with GPU computing.
Detailed neural models at practical compute scales
DeepDendrite was developed to reduce the computational cost of biophysically detailed neuron models while preserving the dependency structure of their numerical solution.
Its Dendritic Hierarchical Scheduling method and GPU framework were introduced in Nature Communications in 2023.
Cross-disciplinary research
The work involved researchers from Peking University, BAAI, KTH, Karolinska Institute, and collaborating institutions.
View publication details →Apache-2.0
DeepDendrite source code and example workloads are available on GitHub.
View source ↗