Connectome Inspired Neural Network

Drosophila

• Connectome-constrained deep mechanistic networks predict neural responses across the fly visual system at single-neuron resolution
  Srinivas C. Turaga
  !!!

• Previous work: A Connectome Based Hexagonal Lattice Convolutional Network Model of the Drosophila Visual System

• One-to-one mapping between deep network units and real neurons uncovers a visual population code for social behavior
  Jonathan W. Pillow, Mala Murthy

• The connectome predicts resting-state functional connectivity across the Drosophila brain

• A Drosophila computational brain model reveals sensorimotor processing Here we create a leaky integrate-and-fire computational model of the entire Drosophila brain, on the basis of neural connectivity and neurotransmitter identity.
  Limitation: We assume each neuron is either exclusively inhibitory or excitatory. We ignore neural morphology and receptor dynamics. The underlying synapses or neurotransmitter predictions may not be fully accurate. Gap junctions cannot be identified in the electron microscopy dataset, so we ignore their possibility.

• From Synapses to Dynamics: Obtaining Function from Structure in a Connectome Constrained Model of the Head Direction Circuit Ila Fiete, 2025

$$\tau \frac{d x_j^A(t)}{d t}=-\ell x_j^A(t)+\sigma\left(\sum_{B \in \mathcal{C}} \sum_k w_0\left(1+Z^{A B}\right) s g n^B C_{j k} x_k^B(t)+b^A+u_j(t)\right)$$

reduces the number of optimized parameters from $439^{2} +439+1 = 193, 161$ to just $7^{2} +7+1 = 57$ parameters Total Loss: Linear Consistency Loss, Stability Loss, Minimum Speed Loss, Entropy Loss, L1 and L2 Regularization

• Differential temporal filtering in the fly optic lobe Alexander Borst Taking advantage of the known connectome I simulate a network of five adjacent optical columns each comprising 65 different cell types. Each neuron is modeled as an electrically compact single compartment, conductance-based element that receives input from other neurons within its column and from its neighboring columns according to the intra- and inter-columnar connectivity matrix.

Human

• Learning Dynamic Graph Representation of Brain Connectome with Spatio-Temporal Attention
  Byung-Hoon Kim, Jong Chul Ye, Jae-Jin Kim
  NeurIPS 2021 Poster

• A Prefrontal Cortex-inspired Architecture for Planning in Large Language Models
  Ida Momennejad

• BrainGNN: Interpretable Brain Graph Neural Network for fMRI Analysis

• sinergia connectomics summer school

• Connectome-Based Attractor Dynamics Underlie Brain Activity in Rest, Task, and Disease

• Enhancer-driven cell type comparison reveals similarities between the mammalian and bird pallium

Perturbation

• Mapping effective connectivity by virtually perturbing a surrogate brain

• Effective Brain Connectome: the whole-brain effective connectivity from neural perturbational inference
  Quanying Liu

• Mapping dysfunctional circuits in the frontal cortex using deep brain stimulation
  Andreas Horn

• Individualized perturbation of the human connectome reveals reproducible biomarkers of network dynamics relevant to cognition
  TMS EEG

monkey

• Compact deep neural network models of visual cortex
  Jonathan W. Pillow, Matthew A. Smith

Rodent

• Inferring brain-wide interactions using data-constrained recurrent neural network models
  Kanaka Rajan, Mount Sinai
  RNN
  • CURBD allows the total activity of each region to be decomposed into a set of source currents from all other regions

C. Elegans

• Elegans-AI: How the connectome of a living organism could model artificial neural networks
  Francesco Bardozzo, Andrea Terlizzi, Claudio Simoncini, Pietro Lió, Roberto Tagliaferri

• Deep connectomics networks: Results from neural network architectures inspired from network neuroscience
  Nicholas Roberts, Vinay Uday Prabhu
  ICML Deep Phenomena 2019

• Deep Connectomics Networks: Neural Network Architectures Inspired by Neuronal Networks
  Nicholas Roberts, Dian Ang Yap, Vinay Uday Prabhu
  Real Neurons & Hidden Units @ NeurIPS 2019 Poster
  C. Elegans and the mouse visual cortex

• Biological connectomes as a representation for the architecture of artificial neural networks
  Samuel Schmidgall, Catherine Schuman, Maryam Parsa
  ICLR 2023 Conference Withdrawn Submission

• A machine learning toolbox for the analysis of sharp-wave ripples reveals common waveform features across species
  Analysis toolbox

• Learning dynamic representations of the functional connectome in neurobiological networks

• Connectome-constrained Latent Variable Model of Whole-Brain Neural Activity
  Lu Mi, .., Srinivas C Turaga

• An integrative data-driven model simulating C. elegans brain, body and environment interactions Nature Computational Science, 2024 Neuron models (Neurons were modeled by morphologically derived multicompartmental models with somatic Hodgkin–Huxley dynamics and passive neurites) + Graded synapse and gap junction models:

Dataset

• Neural signal propagation atlas of Caenorhabditis elegans

Theory-Based

• Learning to Learn with Feedback and Local Plasticity
  Real Neurons & Hidden Units @ NeurIPS 2019 Oral
  Jack Lindsey, Columbia University

• The Simplest Neural Model and a Hypothesis for Language
  Daniel Mitropolsky, Columbia University

• Prediction of neural activity in connectome-constrained recurrent networks
  Manuel Beiran, Ashok Litwin-Kumar

  • Is the connectome insufficient to constrain the dynamics?

• Connectivity Structure and Dynamics of Nonlinear Recurrent Neural Networks

• Spatially embedded recurrent neural networks reveal widespread links between structural and functional neuroscience findings

• Geometric Scaling Law in Real Neuronal Networks
  Gang Yan

cognitive inspired

• TopoNets: High performing vision and language models with brain-like topography

the following arecollected by Ruizhe Zhou

• Bridging the data gap between children and large language models
  Frank, M. C. (2023)

• Cognitive science in the era of artificial intelligence: A roadmap for reverse-engineering the infant language-learner
  Dupoux, E. (2018)

• Findings of the BabyLM Challenge: Sample-Efficient Pretraining on Developmentally Plausible Corpora
  Warstadt, A. et al. (2023)

• MEWL: Few-shot multimodal word learning
  Jiang, G. et al. (2023)

• Lexicon-Level Contrastive Visual-Grounding Improves Language Modeling
  Zhuang, C. et al. (2024)

• Visual Grounding Helps Learn Word Meanings in Low-Data Regimes
  Zhuang, C. et al. (2023)

• Context Limitations Make Neural Language Models More Human-Like
  Kuribayashi, T. et al. (2022)

• Does Vision Accelerate Hierarchical Generalization in Neural Language Learners?
  Kuribayashi, T. (2023)

• Emergent Word Order Universals from Cognitively-Motivated Language Models Tatsuki Kuribayashi, Ryo Ueda, Ryo Yoshida, Yohei Oseki, Ted Briscoe, Timothy Baldwin

Others

• The structural connectome constrains fast brain dynamics

• Towards Biologically Plausible Convolutional Networks NeurIPS 2021

potential model

• Piecewise quadratic neuron model: A tool for close-to-biology spiking neuronal network simulation on dedicated hardware

Basic Architecture

• Exploring Randomly Wired Neural Networks for Image Recognition
  Saining Xie, Alexander Kirillov, Ross Girshick, Kaiming He

• Optimal structure and parameter learning of Ising models

Review

• When brain-inspired AI meets AGI
  Meta-Radiology

• Catalyzing next-generation Artificial Intelligence through NeuroAI

• Deciphering the Blueprint of the Fruit Fly’s Brain

• 人类智能如何从大脑中涌现？大脑精细模拟重塑 NeuroAI 范式

Researcher (TODO)

• Andreas Horn