Eight challenges in developing theory of intelligence

Introduction to dynamical mean-field theory of generic random neural networks

Emergence of hierarchical modes from deep learning

Spectrum of non-Hermitian deep-Hebbian neural networks

Statistical mechanics of continual learning: variational principle and mean-field potential

Associative memory model with arbitrary Hebbian length

Eigenvalue spectrum of neural networks with arbitrary Hebbian length

Ensemble perspective for understanding temporal credit assignment

Equivalence between belief propagation instability and transition to replica symmetry breaking in perceptron learning systems

Data-driven effective model shows a liquid-like deep learning

Learning credit assignment

Relationship between manifold smoothness and adversarial vulnerability in deep learning with local errors

Variational mean-field theory for training restricted Boltzmann machines with binary synapses

Weakly-correlated synapses promote dimension reduction in deep neural networks

Minimal model of permutation symmetry in unsupervised learning

Statistical physics of unsupervised learning with prior knowledge in neural networks

H. Huang and A. Goudarzi, Random active path model of deep neural networks with diluted binary synapses, to appear in Phys Rev E (2018)

H. Huang, Role of zero synapses in unsupervised feature learning, arXiv: 1703.07943, J. Phys. A: Math. Theor. 51 08LT01 (2018) zero synapses act as indicators of concept formation, Published as a Letter.

H. Huang and T. Toyoizumi, Reinforced stochastic gradient descent for deep neural network learning, arXiv: 1701.07974 (2017)

H. Huang, Mechanisms of dimensionality reduction and decorrelation in deep neural networks, arXiv:1710.01467 (2017)

H. Huang, Statistical mechanics of unsupervised feature learning in a restricted Boltzmann machine with binary synapses, arXiv:1612.01717 (2016), J. Stat. Mech. (2017) 053302, recommended in Quora

H. Huang, Theory of population coupling and applications to describe high order correlations in large populations of interacting neurons, arXiv: 1602.08299, J. Stat. Mech. (2017) 033501.

H. Huang and T. Toyoizumi, Clustering of neural codewords revealed by a first-order phase transition, Phys. Rev. E 93, 062416 (2016) arXiv: 1602.06630. selected as one of the most interesting and intriguing arXiv papers from the past week by MIT Technology Review (www.technologyreview.com/s/600956/the-best-of-the-physics-arxiv-week-ending-march-5-2016/).

H. Huang and T. Toyoizumi, Unsupervised feature learning from finite data by message passing: discontinuous versus continuous phase transition, Phys. Rev. E 94, 062310 (2016), arXiv:1608.03714 (2016) discovered spontaneous symmetry breaking in unsupervised learning

H. Huang and T. Toyoizumi, Advanced mean field theory of the restricted Boltzmann machine, Phys. Rev. E 91, 050101(Rapid Communication) (2015).

H. Huang, Code optimization, frozen glassy phase and improved decoding algorithms for low-density parity-check codes, Commun. Theor. Phys. 63, 115 (2015).

H. Huang, Effects of hidden nodes on network structure inference, J. Phys. A: Math. Theor. 48 355002 (2015).

T. Toyoizumi and H. Huang, Structure of attractors in randomly connected networks, Phys. Rev. E 91, 032802 (2015).

H. Huang and Y. Kabashima, Dynamics of asymmetric kinetic Ising systems revisited. J. Stat. Mech.: Theory Exp. P05020 (2014). Proposed back-action field concept in asymmetric kinetic Ising system