Going deeper with morphologically detailed neural networks by simulation-based gradient propagation

Gan He, Kai Du, Tiejun Huang · 2026

Read Going deeper with morphologically detailed neural networks by simulation-based gradient propagation by Gan He, Kai Du, Tiejun Huang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13601319.txt.

Description

Gan He, Kai Du, and Tiejun Huang push brain-like networks with real dendrites deeper than before in a 2026 paper. Detailed neuron models, with branching compartments, compute richly but cost heavily to simulate, and training many layers of them had stalled on both algorithms and infrastructure. Their extension of the DeepDendrite framework changes that.

Layers are modular, and gradients travel by simulation: mirror neurons in a feedback path compute each layer's gradient while the forward pass runs. The trick amounts to backpropagation enacted inside the simulator itself. Fully connected and convolutional versions matched ordinary networks on classic image tasks. The team also probes how attack transfers differ among plain, detailed, and single-compartment networks. The framework opens large-scale study of what dendrites buy a learning system.

Catalog record

CREATORGan He, Kai Du, Tiejun Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13601319

Browse the full library · About the Tower of Babel