NVIS: An Interactive Visualization Tool for Neural Networks Matt Streeter Prof. Matthew O. Ward by Prof. Sergio A. Alvarez advised by and.

Slides:



Advertisements
Similar presentations
FUNCTION FITTING Student’s name: Ruba Eyal Salman Supervisor:
Advertisements

A Brief Overview of Neural Networks By Rohit Dua, Samuel A. Mulder, Steve E. Watkins, and Donald C. Wunsch.
Perceptron Learning Rule
Intelligent Environments1 Computer Science and Engineering University of Texas at Arlington.
Visualization of hidden node activity in a feed forward neural network Adam Arvay.
Artificial Neural Networks
Producing Artificial Neural Networks using a Simple Embryogeny Chris Bowers School of Computer Science, University of Birmingham White.
Prénom Nom Document Analysis: Artificial Neural Networks Prof. Rolf Ingold, University of Fribourg Master course, spring semester 2008.
Connectionist models. Connectionist Models Motivated by Brain rather than Mind –A large number of very simple processing elements –A large number of weighted.
1. Elements of the Genetic Algorithm  Genome: A finite dynamical system model as a set of d polynomials over  2 (finite field of 2 elements)  Fitness.
Evolutionary Feature Extraction for SAR Air to Ground Moving Target Recognition – a Statistical Approach Evolving Hardware Dr. Janusz Starzyk Ohio University.
1 Chapter 11 Neural Networks. 2 Chapter 11 Contents (1) l Biological Neurons l Artificial Neurons l Perceptrons l Multilayer Neural Networks l Backpropagation.
Artificial Neural Networks Artificial Neural Networks are (among other things) another technique for supervised learning k-Nearest Neighbor Decision Tree.
Pattern Recognition using Hebbian Learning and Floating-Gates Certain pattern recognition problems have been shown to be easily solved by Artificial neural.
Prénom Nom Document Analysis: Artificial Neural Networks Prof. Rolf Ingold, University of Fribourg Master course, spring semester 2008.
AN INTERACTIVE TOOL FOR THE STOCK MARKET RESEARCH USING RECURSIVE NEURAL NETWORKS Master Thesis Michal Trna
Multi Layer Perceptrons (MLP) Course website: The back-propagation algorithm Following Hertz chapter 6.
The Performance of Evolutionary Artificial Neural Networks in Ambiguous and Unambiguous Learning Situations Melissa K. Carroll October, 2004.
Neural Networks. Background - Neural Networks can be : Biological - Biological models Artificial - Artificial models - Desire to produce artificial systems.
Hazırlayan NEURAL NETWORKS Radial Basis Function Networks II PROF. DR. YUSUF OYSAL.
Ranga Rodrigo April 5, 2014 Most of the sides are from the Matlab tutorial. 1.
Traffic Sign Recognition Using Artificial Neural Network Radi Bekker
CHAPTER 12 ADVANCED INTELLIGENT SYSTEMS © 2005 Prentice Hall, Decision Support Systems and Intelligent Systems, 7th Edition, Turban, Aronson, and Liang.
Chapter 11: Artificial Intelligence
Evolving a Sigma-Pi Network as a Network Simulator by Justin Basilico.
Multiple-Layer Networks and Backpropagation Algorithms
Slides are based on Negnevitsky, Pearson Education, Lecture 12 Hybrid intelligent systems: Evolutionary neural networks and fuzzy evolutionary systems.
Integrating Neural Network and Genetic Algorithm to Solve Function Approximation Combined with Optimization Problem Term presentation for CSC7333 Machine.
A Comparison of Nature Inspired Intelligent Optimization Methods in Aerial Spray Deposition Management Lei Wu Master’s Thesis Artificial Intelligence Center.
Soft Computing Lecture 18 Foundations of genetic algorithms (GA). Using of GA.
An Introduction to Artificial Intelligence and Knowledge Engineering N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering,
Chapter 9 Neural Network.
Appendix B: An Example of Back-propagation algorithm
2004, 9/1 1 Optimal Content-Based Video Decomposition for Interactive Video Navigation Anastasios D. Doulamis, Member, IEEE and Nikolaos D. Doulamis, Member,
NEURAL NETWORKS FOR DATA MINING
Neural Networks for Protein Structure Prediction Brown, JMB 1999 CS 466 Saurabh Sinha.
COMPARISON OF IMAGE ANALYSIS FOR THAI HANDWRITTEN CHARACTER RECOGNITION Olarik Surinta, chatklaw Jareanpon Department of Management Information System.
1 Chapter 11 Neural Networks. 2 Chapter 11 Contents (1) l Biological Neurons l Artificial Neurons l Perceptrons l Multilayer Neural Networks l Backpropagation.
Laboratory for Computational Intelligence, University of British Columbia Belief & Decision Networks Stochastic Local Search Neural NetworksGraph Searching.
Artificial Neural Networks An Introduction. What is a Neural Network? A human Brain A porpoise brain The brain in a living creature A computer program.
Neural and Evolutionary Computing - Lecture 9 1 Evolutionary Neural Networks Design  Motivation  Evolutionary training  Evolutionary design of the architecture.
Handwritten Recognition with Neural Network Chatklaw Jareanpon, Olarik Surinta Mahasarakham University.
Multi-Layer Perceptron
Introduction to Neural Networks and Example Applications in HCI Nick Gentile.
 Based on observed functioning of human brain.  (Artificial Neural Networks (ANN)  Our view of neural networks is very simplistic.  We view a neural.
CITS7212: Computational Intelligence An Overview of Core CI Technologies Lyndon While.
Introduction to Neural Networks Introduction to Neural Networks Applied to OCR and Speech Recognition An actual neuron A crude model of a neuron Computational.
Hazırlayan NEURAL NETWORKS Backpropagation Network PROF. DR. YUSUF OYSAL.
C - IT Acumens. COMIT Acumens. COM. To demonstrate the use of Neural Networks in the field of Character and Pattern Recognition by simulating a neural.
Neural Networks. Background - Neural Networks can be : Biological - Biological models Artificial - Artificial models - Desire to produce artificial systems.
Pattern Recognition. What is Pattern Recognition? Pattern recognition is a sub-topic of machine learning. PR is the science that concerns the description.
Neural Networks Lecture 4 out of 4. Practical Considerations Input Architecture Output.
An Evolutionary Algorithm for Neural Network Learning using Direct Encoding Paul Batchis Department of Computer Science Rutgers University.
A PID Neural Network Controller
George Yauneridge.  Machine learning basics  Types of learning algorithms  Genetic algorithm basics  Applications and the future of genetic algorithms.
Ghent University Backpropagation for Population-Temporal Coded Spiking Neural Networks July WCCI/IJCNN 2006 Benjamin Schrauwen and Jan Van Campenhout.
 Negnevitsky, Pearson Education, Lecture 12 Hybrid intelligent systems: Evolutionary neural networks and fuzzy evolutionary systems n Introduction.
Evolutionary Computation Evolving Neural Network Topologies.
Neural networks.
Evolutionary Algorithms Jim Whitehead
Introduction to Neural Networks And Their Applications
Prof. Carolina Ruiz Department of Computer Science
Emre O. Neftci  iScience  Volume 5, Pages (July 2018) DOI: /j.isci
Creating Data Representations
Neural Networks II Chen Gao Virginia Tech ECE-5424G / CS-5824
Gautam Dey, Tobias Meyer  Cell Systems 
Review NNs Processing Principles in Neuron / Unit
Neural Networks II Chen Gao Virginia Tech ECE-5424G / CS-5824
Prof. Carolina Ruiz Department of Computer Science
Presentation transcript:

NVIS: An Interactive Visualization Tool for Neural Networks Matt Streeter Prof. Matthew O. Ward by Prof. Sergio A. Alvarez advised by and

What is a Neural Network? Weighted, directed graph, organized into layers Set of neurons (nodes) and synapses (edges), with signals transmitted between neurons via synapses Valuable tool for pattern recognition and function approximation

Why create a visualization tool for neural networks? Understand how neural networks work, gain insight into problem being solved Understand how genetic algorithm evolves networks Other tools exist, but do not show neuron activations or genealogical relationships

Feedforward network visualization Synapse strength represented by length and brightness of colored bars (linear scale). Blue lines indicate positive weights; red lines indicate negative Diameter of white circles represents neuron’s output or activation Each weight acts as a slider

Compact matrix representation Purpose is to allow many networks to be displayed on the screen at once One matrix for each level of weights Row x, column y of matrix n represents weight from node y of layer n to node x of layer n+1 (same colors)

Generations & family trees Row of compact matrix for each generation, ordered by fitness User can select any network in the population history Separate window shows family tree of selected network

Interactive environment Set evolution strategy, network architecture, and training set Graph representation and family tree available for any network in population history Load/save networks Real-time fitness graph

Designing networks By dragging weights, user can design a network to solve a problem, or refine a network that has already been trained Real-time display of fitness score; easy to see importance of particular weight Not a practical way to find a network to solve a problem

Understanding genetic drift Genetic drift is tendency for members of artificial populations to all be alike Initial diversity in generations 0-2, rapidly lost in generations 3-5 Best (leftmost) network in generation 3 is parent of best network in generation 4, grandparent of best 8 in generation 5, and ancestor of all later networks (not shown)

Changing weights & local optima Error backpropagation algorithm performs gradient-based search (local optimum) Weight dragged while backprop is running will either “snap back” to original optimum, or all weights will shift to new optimum Can estimate the length of a local optimum with respect to each axis in weight-space

Extracting domain knowledge Positive weights in all but first layer; effect of input nodes therefore directly related to incident weights Higher crime rates (C) tend to reduce value of house; higher number of rooms (R) tends to increase value Analysis could be applied to problem domains where no a priori knowledge exists

Future work Graph representation does not scale well Implement a variety of evolutionary algorithms (breeding & selection schemes) Depict network architectures other than feedforward User evaluation

For more information... Visit See technical report WPI-CS-TR-00-11, available at: Questions?