On-Line Learning of Two-Layered Neural Network with Randomly Diluted Connections
スポンサーリンク
概要
- 論文の詳細を見る
We investigate an on-line learning of two-layered feed-forward neural networks with randomly diluted connections by using a gradient descent algorithm. We derive coupled first-order differential equations for order parameters which describe learning process in the thermodynamic limit by assuming the self-averaging within the framework of the statistical physics. We clarify that the learning time for asymmetric dilution of connections in a teacher network is shorter than the learning time for symmetric dilution of the connections in the teacher network. But we find that the learning does not converge when the teacher network is too diluted. We obtain a phase diagram in a learning rate versus rate of dilution plane for the diluted teacher network. It turns out that the learning converges imperfectly when a student network is diluted too much, because the average value of connections for the student network is essentially smaller than the value of connection for the teacher network.
- 社団法人日本物理学会の論文
- 2002-02-15
著者
-
Katayama Katsuki
Department Of Computer And Mathematical Sciences Graduate School Of Information Sciences Tohoku Univ
-
Katayama Katsuki
Department Of Computer And Mathematical Sciences Graduate School Of Information Sciences Tohoku Univ
-
Horiguchi Tsuyoshi
Department Of Applied Science Faculty Of Engineering Tohoku University
関連論文
- Critical Properties of a Spin_3/2Ising Model on a Square Lattice
- Spin Models for Packet Routing Control in Computer Networks
- Complex Networks by Non-growing Model with Preferential Rewiring Process(Cross-disciplinary Physics and Related Areas of Science and Technology)
- Sequence Processing Neural Network with Q-States Monotonic Transfer Function
- Ordering of a Frustrated Spin S Ising Model as a Roughening Transition in SOS Model
- Numerical Simulations for Two-Dimensional Traffic-Flow Problem
- Calculation of the Electronic Density of States and the Electric Conductivity of Random Alloys in the Coherent potential Approximation
- Pseudo-Temperature Monte Carlo Simulation for Ground State of Antiferromagnetic Spin-S Ising Model on Triangular Lattice
- Spin-3/2 Ising Model and Ashkin-Teller Model
- Dynamical Neural Network Model of Hippocampus with Excitatory and Inhibitory Neurons (Cross-disciplinary Physics and Related Areas of Science and Technology)
- Noise Effect on Memory Recall in Dynamical Neural Network Model of Hippocampus
- Coherent-Anomaly Analysis with Cluster Variation Method for Two-Dimensional Ising Model with Nearest-Neighbor and Next-Nearest-Neighbor Interactions
- Low-Temperature Behavior of Antiferromagnetic Ising Model on Triangular Lattice
- Coherent-Anomaly Analysis with Cluster Variation Method for Spin-Pair Correlation Function of Ising Model on Square Lattice
- Coherent-Anomaly Method for the Wave-Number Dependence of the Susceptibility
- Low Temperature Properties of One-Dimensional Classical XY Model with Frustration
- Synchronous Phenomena of Neural Network Models Using Hindmarsh–Rose Equation
- Numerical Simulations for Traffic Flow in Two-Dimensional Network with Obstacles
- Unsupervised Image Segmentation Based on Bethe Approximation
- Generalization Ability of Hopfield Neural Network with Spin-S Ising Neurons : General Physics
- Low-Temperature Behaviors of Antiferromagnetic Ising Model of Infinite-Spin on Triangular Lattice
- Finite Temperature Properties of One-Dimensional Chiral XY-model
- Phase Diagrams of Spin-3/2 Ising Model on a Square Lattice in Terms of Corner Transfer Matrix Renormalization Group Method
- Low-Temperature Properties of Infinite-Spin Ising Model in One-Dimensional Lattice
- Tetradics Formulation of the Two-Time Green's Function Method and Its Application to the Heisenberg Ferromagnet
- Ising Model in Correlated Random Fields on Kagome Lattice
- Sequence Processing Neural Network with a Non-Monotonic Transfer Function : General Physics
- An Extension of Perron-Frobenius Theorem for Positive Symmetric Matrices
- Ising Model with Even-Bonds Plaquettes on the Square Lattice
- Ground State Properties of One-Dimensional Quantum Chiral XY -model
- Self-Organization in Four-Direction Traffic-Flow Model : Cross-Disciplinary Physics
- Temperature Dependence of the Spin-Pair Correlation Functions for the Heisenberg Antiferromagnet and the NMR Line Width of F^ in RbMnF_3 in the Paramagnetic Region
- Formulas for the Lattice Green's Functions for the Cubic Lattices in Terms of the Complete Elliptic Integral
- Phase Transition in a Four-Direction Traffic Flow Model on a Two-Dimensional Network
- Restoration Process for Gray-scale Patterns by Using Path-integral
- Models of MT and MST areas using wake-sleep algorithm
- Da Costa-Nobre-Yokoi Model with Spin S
- On-Line Learning of Two-Layered Neural Network with Randomly Diluted Connections
- Ghatak-Sherrington Model with Spin S
- Chaotic Behavior for Asymmetrically Diluted Hopfield Neural Network with a Non-Monotonic Transfer Function(General)
- First Order Green Function Theory of Ferromagnetism
- Color Image Compression Algorithm Using Self-Organizing Feature Map
- Dynamical Properties of the Diluted Heisenberg and XY Magnets at Infinite Temperature. I : Spin Diffusion Constant
- Low-and High-Temperature Behaviors of Ferromagnetic Ising Model of Infinite-Spin
- Structure of Ground States of Antiferromagnetic Ising Model with General-Spin on Triangular Lattice
- Optimal Network Structure for Packet Flow in Shortest-Path Routing Control Model
- Universality Class for Spin Model With Generalized Ising Spin on One-Dimensional Lattice : Progress Letters
- Note on the Susceptibility of the Two-Dimensional Heisenberg Ferromagnet
- Probabilistic, Iterated and Quantum-Iterated Computational Methods in Gray-Level Image Restoration
- Lattice Green's Function for the Simple Cubic Lattice
- Optimal Network Structure for Packet Flow in Shortest-Path Routing Control Model
- Spin Models for Packet Routing Control in Computer Networks
- Fully Frustrated Ising Model on a Square Lattice