[2009.01269] Can Single Neurons Solve MNIST? The Computational Power of Biological Dendritic Trees

Physiological experiments have highlighted how the dendrites of biological neurons can nonlinearly process distributed synaptic inputs. This is in stark contrast to units in artificial neural networks that are generally linear apart from an output nonlinearity. If dendritic trees can be nonlinear, biological neurons may have far more computational power than their artificial counterparts. Here we use a simple model where the dendrite is implemented as a sequence of thresholded linear units. We find that such dendrites can readily solve machine learning problems, such as MNIST or CIFAR-10, and that they benefit from having the same input onto several branches of the dendritic tree. This dendrite model is a special case of sparse network. This work suggests that popular neuron models may severely underestimate the computational power enabled by the biological fact of nonlinear dendrites and multiple synapses per pair of neurons. The next generation of artificial neural networks may significantly benefit from these biologically inspired dendritic architectures.

Can Single Neurons Solve MNIST? The Computational Power of Biological Dendritic arXiv:2009.01269v1 [q-bio.NC] 2 Sep 2020 Trees Ilenna Simone Jones1 and Konrad Kording2 1 Department of Neuroscience, University of Pennsylvania 2 Departments of…

related reading

• Towards deep learning with segregated dendrites | eLifeelifesciences.org
• Toy Models of Superpositiontransformer-circuits.pub
• Neural networks and deep learningneuralnetworksanddeeplearning.com
• How Computationally Complex Is a Single Neuron? | Quanta Magazinequantamagazine.org
• Neural networks and deep learningneuralnetworksanddeeplearning.com
• Conductance-based dendrites perform Bayes-optimal cue integrationpmc.ncbi.nlm.nih.gov
• Dimensionality reduction in neural data analysisxcorr.net
• Neuroscientists have vastly underestimated brain cells’ computing powerscientificamerican.com
• A Million Spiking-Neuron Integrated Circuit with a Scalable Communication Network and Interfacepaulmerolla.com
• A deep learning framework for neuroscience - PMCncbi.nlm.nih.gov
• Mechanistic Interpretability: A Challenge Common to Both Artificial and Biological Intelligencekempnerinstitute.harvard.edu
• Simulation of networks of spiking neurons: A review of tools and strategiespmc.ncbi.nlm.nih.gov