Brain Inspired Algorithms Dr. David Fagan

Slides:



Advertisements
Similar presentations
CSC321: 2011 Introduction to Neural Networks and Machine Learning Lecture 7: Learning in recurrent networks Geoffrey Hinton.
Advertisements

Machine Learning Neural Networks
Neural NetworksNN 11 Neural Networks Teacher: Elena Marchiori R4.47 Assistant: Kees Jong S2.22
Rutgers CS440, Fall 2003 Neural networks Reading: Ch. 20, Sec. 5, AIMA 2 nd Ed.
Carla P. Gomes CS4700 CS 4700: Foundations of Artificial Intelligence Prof. Carla P. Gomes Module: Intro Neural Networks (Reading:
Neural Networks.
MSE 2400 EaLiCaRA Spring 2015 Dr. Tom Way
Artificial Neural Network Theory and Application Ashish Venugopal Sriram Gollapalli Ulas Bardak.
Artificial Neural Nets and AI Connectionism Sub symbolic reasoning.
IE 585 Introduction to Neural Networks. 2 Modeling Continuum Unarticulated Wisdom Articulated Qualitative Models Theoretic (First Principles) Models Empirical.
 The most intelligent device - “Human Brain”.  The machine that revolutionized the whole world – “computer”.  Inefficiencies of the computer has lead.
NEURAL NETWORKS FOR DATA MINING
LINEAR CLASSIFICATION. Biological inspirations  Some numbers…  The human brain contains about 10 billion nerve cells ( neurons )  Each neuron is connected.
From Machine Learning to Deep Learning. Topics that I will Cover (subject to some minor adjustment) Week 2: Introduction to Deep Learning Week 3: Logistic.
M Machine Learning F# and Accord.net. Alena Dzenisenka Software architect at Luxoft Poland Member of F# Software Foundation Board of Trustees Researcher.
Neural Networks in Computer Science n CS/PY 231 Lab Presentation # 1 n January 14, 2005 n Mount Union College.
Neural Networks Presented by M. Abbasi Course lecturer: Dr.Tohidkhah.
Neural Networks Teacher: Elena Marchiori R4.47 Assistant: Kees Jong S2.22
Computer Architecture Lecture 26 Past and Future Ralph Grishman November 2015 NYU.
Object Recognizing. Deep Learning Success in 2012 DeepNet and speech processing.
Lecture 12. Outline of Rule-Based Classification 1. Overview of ANN 2. Basic Feedforward ANN 3. Linear Perceptron Algorithm 4. Nonlinear and Multilayer.
Machine Learning Artificial Neural Networks MPλ ∀ Stergiou Theodoros 1.
Business Intelligence and Decision Support Systems (9 th Ed., Prentice Hall) Chapter 6: Artificial Neural Networks for Data Mining.
Comparing TensorFlow Deep Learning Performance Using CPUs, GPUs, Local PCs and Cloud Pace University, Research Day, May 5, 2017 John Lawrence, Jonas Malmsten,
TensorFlow The Deep Learning Library You Should Be Using.
Machine Learning Supervised Learning Classification and Regression
2/13/2018 4:38 AM © Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN.
Big data classification using neural network
TensorFlow– A system for large-scale machine learning
CS 6501: 3D Reconstruction and Understanding Convolutional Neural Networks Connelly Barnes.
Machine Learning for Big Data
Deep Learning Amin Sobhani.
Status Report on Machine Learning
Deep Learning Insights and Open-ended Questions
AlphaGo with Deep RL Alpha GO.
Status Report on Machine Learning
ICS 491 Big Data Analytics Fall 2017 Deep Learning
What is an ANN ? The inventor of the first neuro computer, Dr. Robert defines a neural network as,A human brain like system consisting of a large number.
Deep Learning Fundamentals online Training at GoLogica
Overview of TensorFlow
A VERY Brief Introduction to Convolutional Neural Network using TensorFlow 李 弘
Artificial Intelligence – Deep Learning and its Applications
Machine Learning: The Connectionist
CSE P573 Applications of Artificial Intelligence Neural Networks
Introduction to Deep Learning for neuronal data analyses
FUNDAMENTALS OF MACHINE LEARNING AND DEEP LEARNING
Intelligent Leaning -- A Brief Introduction to Artificial Neural Networks Chiung-Yao Fang.
of the Artificial Neural Networks.
Intelligent Leaning -- A Brief Introduction to Artificial Neural Networks Chiung-Yao Fang.
CSE 573 Introduction to Artificial Intelligence Neural Networks
Object Classes Most recent work is at the object level We perceive the world in terms of objects, belonging to different classes. What are the differences.
CSC 578 Neural Networks and Deep Learning
Long Short Term Memory within Recurrent Neural Networks
Neural Networks Geoff Hulten.
Lecture Notes for Chapter 4 Artificial Neural Networks
CS246: Search-Engine Scale
Fundamentals of Neural Networks Dr. Satinder Bal Gupta
Artificial Neural Networks
Artificial Neural Networks
COSC 4335: Part2: Other Classification Techniques
TensorFlow: A System for Large-Scale Machine Learning
Deep Learning Authors: Yann LeCun, Yoshua Bengio, Geoffrey Hinton
David Kauchak CS51A Spring 2019
Machine learning CS 229 / stats 229
PYTHON Deep Learning Prof. Muhammad Saeed.
CSC 578 Neural Networks and Deep Learning
CSC 578 Neural Networks and Deep Learning
Prabhas Chongstitvatana Chulalongkorn University
Parallel Systems to Compute
Patterson: Chap 1 A Review of Machine Learning
Presentation transcript:

Brain Inspired Algorithms Dr. David Fagan Neurocomputing Brain Inspired Algorithms Dr. David Fagan MIS40980 Natural Computing & Applications

Basic Unit of the Computer In computers we use transistors to perform computation Once enough voltage is applied to the base, the connection between collector and emitter is established. Foundation of all processors in computers. MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Basic Unit of the Brain MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Human Brain Extremely complex computational device Estimated 80 – 100 billion nerve cells in human brain It takes 82,944 processors running a network of 1.73 billion cells with 10.4 trillion synapses 40 minutes to perform 1 second of brain processing time. Using 1PB (1,000,000 GB) of memory!!!!! More here MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Neural Networks Neural Networks cover a wide variety of algorithms that take inspiration from the brain. Key to all these networks is a general model of a Neuron (Perceptron) MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Basic Neural Network A neural network is essentially: A composition of Perceptrons Connected in different ways Operating on different activation functions. Glosser.ca MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Neural Networks Neurons Bias Training Network Types MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Neurons Many different activation functions Each one has different characteristics http://engineeronadisk.com/ MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Bias Bias is a constant that is added to a layer, the value of which can be adjusted during training Bias allows the network to shift the activation function, potentially allowing for better fitting of the neuron to the data, which may be critical for successful learning. In most cases bias is helpful MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Bias http://stackoverflow.com/questions/2480650/role-of-bias-in-neural-networks MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Bias http://stackoverflow.com/questions/2480650/role-of-bias-in-neural-networks MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Training Getting a network to learn can be extremely difficult and slow. Several approaches to training neural nets exist. The most prevalent is Back propagation of Errors This is used with some form of optimization method like gradient decent MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Training In Matlab they offer the following: Levenberg-Marquardt BFGS Quasi-Newton Resilient Backpropagation Scaled Conjugate Gradient Conjugate Gradient with Powell/Beale Restarts Fletcher-Powell Conjugate Gradient Polak-Ribiére Conjugate Gradient One Step Secant Variable Learning Rate Backpropagation Click here for more info MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Learning Methods Supervised Learning Pattern Recognition, Classification), Regression, Speech and Gesture recognition. Learning with a Teacher Unsupervised Learning Clustering, Estimation of Statistical Distributions, Compression and Filtering. Some data is given and a cost function is used to drive the learning. Reinforcement Learning Control Problems, Games and Sequential Decision Making Tasks. Learn by minimising / maximising gain with each interaction with environment MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Network Types Feed Forward NN Auto Encoder Recurrent Neural Nets Restricted Boltzmann Machine Many more…........... Google is your friend here!!!!!!! or click here All based on similar fundamental MIS40980 Natural Computing & Applications

Software for Neural Nets Many Libraries Exist For beginners I generally recommend PyBrain Python based, easy to install and loads of documentation Many more to pick from Theano, Caffe, TensorFlow, PyTorch............ Give TensorFlow or PyTorch a go!!!! MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Lets Play!!!!!! The following is an interactive web based neural network that helps to show how networks behave during training Tensor Flow Playground MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Deep Neural Networks In the past it was difficult to train large networks (time, computation, hardware) Recent advances (GPU acceleration, large computational clusters etc.) have allowed for easier training of networks Deep Networks simply refers to any network that has many layers (5+ layers) Modern Deep Nets have many layers and take months/years to train Deep Nets and Deep Learning are two different things!!!!!! MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Deep Nets Source MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Deep Nets Source MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Deep Nets Source MIS40980 Natural Computing & Applications

Deep Neural Networks in the News Google Deepmind – AlphaGo Go is a very difficult game for computer In January 2016 both Google and Facebook released results for there respective GO AI projects, indicating both were reaching levels never reached by previous AI’s Google’s results where from Oct/Nov 2015 In March 2016 AlphaGo played 5 games against top human player Lee Sedol AlphaGo only lost one game after it made a mistake but took several moves to realise this mistake at which point the game was beyond saving Google had been training the AI against itself from Oct/Nov till March Recently beat he world #1 ranked go player!!!!!!!! MIS40980 Natural Computing & Applications

Deep Learning Examples Check out this page that contains Deep ConvNets that run in your browser ConvNetsJS MIS40980 Natural Computing & Applications

Machine learning and Deep Neural Nets Still confused? Check out this video Loads of courses and instructional videos available on Youtube Same goes for online courses, just ask Google MIS40980 Natural Computing & Applications

MIS40980 Natural Computing & Applications Evolution + NN NEAT Evolution of Neural Networks using a GA to decide structure and weights etc. Several version with most recent being HyperNEAT MIS40980 Natural Computing & Applications