SlideShare a Scribd company logo
Multi-layer feedforward network
                                                                                          1
                                                                                                            1



              Artificial Intelligence                                                I1
                                                                                                       H1


                                                                                                                          O1
                                                                                     I2
             Backpropagation training of                                                                H2
      multi-layer feedforward neural networks                                                                   Ii: Input unit i
                                                                                     I3
                                                                                                                Oi: Output unit i
                                                                                                                Hi: Hidden unit i
      CS 4633/6633 Artificial Intelligence                                     CS 4633/6633 Artificial Intelligence




      Training multi-layer networks                                             Backpropagation: The idea
q   Supervised learning: A teacher presents input patterns to            q   The key idea is that the error in the output/activation of a
    the network, compares the resulting outputs with the                     neuron in a hidden layer can be defined as the weighted
    desired outputs, and adjusts the network weights to reduce               sum of the error of its successor neurons.
    the error                                                            q   Learning involves changing the weights of the network
q   For perceptron networks, adjusting weights to reduce error           q   Weights are changed based partly on the derivative of the
    is straightforward because there is only one weight                      activation function of a neuron. The sigmoid function has a
    between each input and output unit                                       simple derivative: if y = 1/(1 + e-x), then dy/dx = y(1-y).
q   For multi-layer networks, adjusting weights to reduce error
    is more complicated because there are many weights
    between each input unit and output unit, and each of these
    weights contributes to more than one output unit

      CS 4633/6633 Artificial Intelligence                                     CS 4633/6633 Artificial Intelligence




                                                                                          Backpropagation is
    Backpropagation: The Algorithm
Initialize network weights to random numbers between -0.5 and 0.5
                                                                                          hill-climbing search
Repeat until error is sufficiently small (each repetition is an epoch)   q   Backprogagation is search in weight space, where the
                                                                             number of dimensions of the search space is equal to the
     For each training example:
                                                                             number of weights.
         Propagate input through network to compute output
                                                                         q   It is possible for backpropagation to converge to a local
         Propagate errors backward through network, as follows:
                                                                             minimum, although it rarely happens in practice
           For each output unit i, calculate the error:
                                                                         q   α is learning rate. If too low, learning is unnecessarly
                 Errorj ← Oi × (1 - Oi) × (Ti - Oi)
                                                                             slow. If too high, it does not converge.
           For each hidden unit i, calculate the error:
                                                                         q   Momentum: can improve learning time by making the
                Errori ← Oi × (1 - Oi) × ∑i (Wi,j × Errorj)
                                                                             weight adjustment proportional to the amount of the
           Update each network weight:
                                                                             previous weight change
                Wi,j ← Wi,j + α × Ii × Errorj

      CS 4633/6633 Artificial Intelligence                                     CS 4633/6633 Artificial Intelligence
Cross-validation training                                                        Generalization
q   A statistical approach for avoiding overfitting
q   The data set is randomly partitioned into a training set and
    a test set. The training set is further partitioned into two
    subsets, one for training and one for validation.
                                                                                                                             Training set
q   Performance will always keep improving on the training                  %
                                                                          correct
    set, due to overfitting. When performance stops improving
                                                                                                                            Validation set
    on the validation set, training is stopped. Then the network
    is tested on the test set.
     – train = training set
     – test1 = validation set for determining convergence
                                                                                              Training time (# of epochs)
     – test2 = test set

      CS 4633/6633 Artificial Intelligence                                   CS 4633/6633 Artificial Intelligence




                          Image files                                                    Network structure
q   Each file contains a 32 x 30 image, with each pixel an ASCII      q   All networks have three layers (one hidden layer)
    character that represents a greyscale intensity value between 0   q   For Glickman recognition problem
    (black) and 255 (white)
                                                                           – 960 input units, one for each pixel in image
q   attribute-value pairs                                                  – 1 output unit, yes or no
     – userid (20 possible values)                                         – 4 hidden units
     – pose (straight, left, right, up)                               q   For sunglasses recognition problem
     – expression (neutral, happy, sad, angry)                             – same
     – eyes (open, sunglasses)                                        q   For 1-of-20 face recognition problem
q   Problem: learn function from image to attribute-value pair(s)          – 20 output units
     – map image to yes if Glickman, otherwise no                          – suggestion: 20 hidden units
     – map image to yes if wearing sunglasses, otherwise no           q   For pose recognition problem
     – map image to userid                                                 – 4 output units
     – map image to pose                                                   – suggestion: 6 hidden units
      CS 4633/6633 Artificial Intelligence                                   CS 4633/6633 Artificial Intelligence




                   Changes to code                                                                  Pointers
                                                                      q   How well a network learns depends on number of hidden
q   Sunglasses recognition:
                                                                          units, learning rate, etc. You are given these.
       Change the target output by changing the function
                                                                      q   Backpropagation algorithm may not train a net to achieve
       “load_target” in imagenet.c, as follows.
                                                                          perfect classification accuracy
                if (!strcmp(eyes, “sunglasses”)) {
q   1-of-20 face recognition and pose recognition:                    q   To test network w/o training it, set command-line
                                                                          parameter for epochs to zero
        Change number of output and hidden units by changing
       arguments to bpnn_create in facetrain.c. Change the            q   Output lines: epoch number, sum of error over all units, for
       evaluation routines “performance_on_imagelist()” and               training set: percentage correct and average error, for test
       “evaluate_performance” in facetrain.c to reflect the new           set1: percentage correct and average error, for test set2:
       network structure. Also change the target outputs by               percentage correct and average error
       changing the function “load_target” in imagenet.c.             q   Not required to do questions 11 and 12. But consider
                                                                          drawing learning curves to show results of training.
      CS 4633/6633 Artificial Intelligence                                   CS 4633/6633 Artificial Intelligence

More Related Content

What's hot

Lecture 06 marco aurelio ranzato - deep learning
Lecture 06   marco aurelio ranzato - deep learningLecture 06   marco aurelio ranzato - deep learning
Lecture 06 marco aurelio ranzato - deep learning
mustafa sarac
 
Learning stochastic neural networks with Chainer
Learning stochastic neural networks with ChainerLearning stochastic neural networks with Chainer
Learning stochastic neural networks with Chainer
Seiya Tokui
 
A multithreaded method for network alignment
A multithreaded method for network alignmentA multithreaded method for network alignment
A multithreaded method for network alignment
David Gleich
 
2.5 backpropagation
2.5 backpropagation2.5 backpropagation
2.5 backpropagation
Krish_ver2
 
14 Machine Learning Single Layer Perceptron
14 Machine Learning Single Layer Perceptron14 Machine Learning Single Layer Perceptron
14 Machine Learning Single Layer Perceptron
Andres Mendez-Vazquez
 
CUDA and Caffe for deep learning
CUDA and Caffe for deep learningCUDA and Caffe for deep learning
CUDA and Caffe for deep learning
Amgad Muhammad
 
Principles of soft computing-Associative memory networks
Principles of soft computing-Associative memory networksPrinciples of soft computing-Associative memory networks
Principles of soft computing-Associative memory networks
Sivagowry Shathesh
 
SOIAM (SOINN-AM)
SOIAM (SOINN-AM)SOIAM (SOINN-AM)
SOIAM (SOINN-AM)
SOINN Inc.
 
Extreme learning machine:Theory and applications
Extreme learning machine:Theory and applicationsExtreme learning machine:Theory and applications
Extreme learning machine:Theory and applications
James Chou
 
Efficient BP Algorithms for General Feedforward Neural Networks
Efficient BP Algorithms for General Feedforward Neural NetworksEfficient BP Algorithms for General Feedforward Neural Networks
Efficient BP Algorithms for General Feedforward Neural Networks
Francisco Zamora-Martinez
 
Predicting organic reaction outcomes with weisfeiler lehman network
Predicting organic reaction outcomes with weisfeiler lehman networkPredicting organic reaction outcomes with weisfeiler lehman network
Predicting organic reaction outcomes with weisfeiler lehman network
Kazuki Fujikawa
 
Back propagation
Back propagation Back propagation
Back propagation
DrBaljitSinghKhehra
 
Introduction to Chainer 11 may,2018
Introduction to Chainer 11 may,2018Introduction to Chainer 11 may,2018
Introduction to Chainer 11 may,2018
Preferred Networks
 
"Deep Learning" Chap.6 Convolutional Neural Net
"Deep Learning" Chap.6 Convolutional Neural Net"Deep Learning" Chap.6 Convolutional Neural Net
"Deep Learning" Chap.6 Convolutional Neural Net
Ken'ichi Matsui
 
NIPS2017 Few-shot Learning and Graph Convolution
NIPS2017 Few-shot Learning and Graph ConvolutionNIPS2017 Few-shot Learning and Graph Convolution
NIPS2017 Few-shot Learning and Graph Convolution
Kazuki Fujikawa
 
elm
elmelm
ELM: Extreme Learning Machine: Learning without iterative tuning
ELM: Extreme Learning Machine: Learning without iterative tuningELM: Extreme Learning Machine: Learning without iterative tuning
ELM: Extreme Learning Machine: Learning without iterative tuning
zukun
 
Sparse autoencoder
Sparse autoencoderSparse autoencoder
Sparse autoencoder
Devashish Patel
 
Lecture 9 Perceptron
Lecture 9 PerceptronLecture 9 Perceptron
Lecture 9 Perceptron
Marina Santini
 
Deep Learning with PyTorch
Deep Learning with PyTorchDeep Learning with PyTorch
Deep Learning with PyTorch
Mayur Bhangale
 

What's hot (20)

Lecture 06 marco aurelio ranzato - deep learning
Lecture 06   marco aurelio ranzato - deep learningLecture 06   marco aurelio ranzato - deep learning
Lecture 06 marco aurelio ranzato - deep learning
 
Learning stochastic neural networks with Chainer
Learning stochastic neural networks with ChainerLearning stochastic neural networks with Chainer
Learning stochastic neural networks with Chainer
 
A multithreaded method for network alignment
A multithreaded method for network alignmentA multithreaded method for network alignment
A multithreaded method for network alignment
 
2.5 backpropagation
2.5 backpropagation2.5 backpropagation
2.5 backpropagation
 
14 Machine Learning Single Layer Perceptron
14 Machine Learning Single Layer Perceptron14 Machine Learning Single Layer Perceptron
14 Machine Learning Single Layer Perceptron
 
CUDA and Caffe for deep learning
CUDA and Caffe for deep learningCUDA and Caffe for deep learning
CUDA and Caffe for deep learning
 
Principles of soft computing-Associative memory networks
Principles of soft computing-Associative memory networksPrinciples of soft computing-Associative memory networks
Principles of soft computing-Associative memory networks
 
SOIAM (SOINN-AM)
SOIAM (SOINN-AM)SOIAM (SOINN-AM)
SOIAM (SOINN-AM)
 
Extreme learning machine:Theory and applications
Extreme learning machine:Theory and applicationsExtreme learning machine:Theory and applications
Extreme learning machine:Theory and applications
 
Efficient BP Algorithms for General Feedforward Neural Networks
Efficient BP Algorithms for General Feedforward Neural NetworksEfficient BP Algorithms for General Feedforward Neural Networks
Efficient BP Algorithms for General Feedforward Neural Networks
 
Predicting organic reaction outcomes with weisfeiler lehman network
Predicting organic reaction outcomes with weisfeiler lehman networkPredicting organic reaction outcomes with weisfeiler lehman network
Predicting organic reaction outcomes with weisfeiler lehman network
 
Back propagation
Back propagation Back propagation
Back propagation
 
Introduction to Chainer 11 may,2018
Introduction to Chainer 11 may,2018Introduction to Chainer 11 may,2018
Introduction to Chainer 11 may,2018
 
"Deep Learning" Chap.6 Convolutional Neural Net
"Deep Learning" Chap.6 Convolutional Neural Net"Deep Learning" Chap.6 Convolutional Neural Net
"Deep Learning" Chap.6 Convolutional Neural Net
 
NIPS2017 Few-shot Learning and Graph Convolution
NIPS2017 Few-shot Learning and Graph ConvolutionNIPS2017 Few-shot Learning and Graph Convolution
NIPS2017 Few-shot Learning and Graph Convolution
 
elm
elmelm
elm
 
ELM: Extreme Learning Machine: Learning without iterative tuning
ELM: Extreme Learning Machine: Learning without iterative tuningELM: Extreme Learning Machine: Learning without iterative tuning
ELM: Extreme Learning Machine: Learning without iterative tuning
 
Sparse autoencoder
Sparse autoencoderSparse autoencoder
Sparse autoencoder
 
Lecture 9 Perceptron
Lecture 9 PerceptronLecture 9 Perceptron
Lecture 9 Perceptron
 
Deep Learning with PyTorch
Deep Learning with PyTorchDeep Learning with PyTorch
Deep Learning with PyTorch
 

Viewers also liked

Artificial Intelligence Notes- Set6
Artificial Intelligence Notes- Set6Artificial Intelligence Notes- Set6
Artificial Intelligence Notes- Set6
Gurpreet singh
 
Artificial Intelligence Notes- Set1
Artificial Intelligence Notes- Set1Artificial Intelligence Notes- Set1
Artificial Intelligence Notes- Set1
Gurpreet singh
 
Artificial Intelligence Notes- Set5
Artificial Intelligence Notes- Set5Artificial Intelligence Notes- Set5
Artificial Intelligence Notes- Set5
Gurpreet singh
 
Artificial Intelligence Notes- Set3
Artificial Intelligence Notes- Set3Artificial Intelligence Notes- Set3
Artificial Intelligence Notes- Set3
Gurpreet singh
 
Artificial Intelligence Notes- Set4
Artificial Intelligence Notes- Set4Artificial Intelligence Notes- Set4
Artificial Intelligence Notes- Set4
Gurpreet singh
 
Artificial Intelligence Notes- Set2
Artificial Intelligence Notes- Set2Artificial Intelligence Notes- Set2
Artificial Intelligence Notes- Set2
Gurpreet singh
 

Viewers also liked (6)

Artificial Intelligence Notes- Set6
Artificial Intelligence Notes- Set6Artificial Intelligence Notes- Set6
Artificial Intelligence Notes- Set6
 
Artificial Intelligence Notes- Set1
Artificial Intelligence Notes- Set1Artificial Intelligence Notes- Set1
Artificial Intelligence Notes- Set1
 
Artificial Intelligence Notes- Set5
Artificial Intelligence Notes- Set5Artificial Intelligence Notes- Set5
Artificial Intelligence Notes- Set5
 
Artificial Intelligence Notes- Set3
Artificial Intelligence Notes- Set3Artificial Intelligence Notes- Set3
Artificial Intelligence Notes- Set3
 
Artificial Intelligence Notes- Set4
Artificial Intelligence Notes- Set4Artificial Intelligence Notes- Set4
Artificial Intelligence Notes- Set4
 
Artificial Intelligence Notes- Set2
Artificial Intelligence Notes- Set2Artificial Intelligence Notes- Set2
Artificial Intelligence Notes- Set2
 

Similar to Multilayer Slides

backpropagation in neural networks
backpropagation in neural networksbackpropagation in neural networks
backpropagation in neural networks
Akash Goel
 
6
66
AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...
AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...
AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...
AILABS Academy
 
nural network ER. Abhishek k. upadhyay
nural network ER. Abhishek  k. upadhyaynural network ER. Abhishek  k. upadhyay
nural network ER. Abhishek k. upadhyay
abhishek upadhyay
 
(Artificial) Neural Network
(Artificial) Neural Network(Artificial) Neural Network
(Artificial) Neural Network
Putri Wikie
 
soft computing
soft computingsoft computing
soft computing
AMIT KUMAR
 
Simulation of Single and Multilayer of Artificial Neural Network using Verilog
Simulation of Single and Multilayer of Artificial Neural Network using VerilogSimulation of Single and Multilayer of Artificial Neural Network using Verilog
Simulation of Single and Multilayer of Artificial Neural Network using Verilog
ijsrd.com
 
Lec 6-bp
Lec 6-bpLec 6-bp
Lec 6-bp
Taymoor Nazmy
 
Classification by back propagation, multi layered feed forward neural network...
Classification by back propagation, multi layered feed forward neural network...Classification by back propagation, multi layered feed forward neural network...
Classification by back propagation, multi layered feed forward neural network...
bihira aggrey
 
Artificial Neural Network for machine learning
Artificial Neural Network for machine learningArtificial Neural Network for machine learning
Artificial Neural Network for machine learning
2303oyxxxjdeepak
 
Artificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning ModelsArtificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning Models
DrBaljitSinghKhehra
 
Artificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning ModelsArtificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning Models
DrBaljitSinghKhehra
 
Artificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning ModelsArtificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning Models
DrBaljitSinghKhehra
 
Artificial neural network
Artificial neural networkArtificial neural network
Artificial neural network
mustafa aadel
 
Character Recognition using Artificial Neural Networks
Character Recognition using Artificial Neural NetworksCharacter Recognition using Artificial Neural Networks
Character Recognition using Artificial Neural Networks
Jaison Sabu
 
Basics of Artificial Neural Network
Basics of Artificial Neural Network Basics of Artificial Neural Network
Basics of Artificial Neural Network
Subham Preetam
 
CS767_Lecture_05.pptx
CS767_Lecture_05.pptxCS767_Lecture_05.pptx
CS767_Lecture_05.pptx
ShujatHussainGadi
 
Neural networks
Neural networksNeural networks
Neural networks
HarshitGupta367
 
Batch normalization presentation
Batch normalization presentationBatch normalization presentation
Batch normalization presentation
Owin Will
 
MNN
MNNMNN

Similar to Multilayer Slides (20)

backpropagation in neural networks
backpropagation in neural networksbackpropagation in neural networks
backpropagation in neural networks
 
6
66
6
 
AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...
AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...
AILABS - Lecture Series - Is AI the New Electricity? Topic:- Classification a...
 
nural network ER. Abhishek k. upadhyay
nural network ER. Abhishek  k. upadhyaynural network ER. Abhishek  k. upadhyay
nural network ER. Abhishek k. upadhyay
 
(Artificial) Neural Network
(Artificial) Neural Network(Artificial) Neural Network
(Artificial) Neural Network
 
soft computing
soft computingsoft computing
soft computing
 
Simulation of Single and Multilayer of Artificial Neural Network using Verilog
Simulation of Single and Multilayer of Artificial Neural Network using VerilogSimulation of Single and Multilayer of Artificial Neural Network using Verilog
Simulation of Single and Multilayer of Artificial Neural Network using Verilog
 
Lec 6-bp
Lec 6-bpLec 6-bp
Lec 6-bp
 
Classification by back propagation, multi layered feed forward neural network...
Classification by back propagation, multi layered feed forward neural network...Classification by back propagation, multi layered feed forward neural network...
Classification by back propagation, multi layered feed forward neural network...
 
Artificial Neural Network for machine learning
Artificial Neural Network for machine learningArtificial Neural Network for machine learning
Artificial Neural Network for machine learning
 
Artificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning ModelsArtificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning Models
 
Artificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning ModelsArtificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning Models
 
Artificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning ModelsArtificial Neural Networks-Supervised Learning Models
Artificial Neural Networks-Supervised Learning Models
 
Artificial neural network
Artificial neural networkArtificial neural network
Artificial neural network
 
Character Recognition using Artificial Neural Networks
Character Recognition using Artificial Neural NetworksCharacter Recognition using Artificial Neural Networks
Character Recognition using Artificial Neural Networks
 
Basics of Artificial Neural Network
Basics of Artificial Neural Network Basics of Artificial Neural Network
Basics of Artificial Neural Network
 
CS767_Lecture_05.pptx
CS767_Lecture_05.pptxCS767_Lecture_05.pptx
CS767_Lecture_05.pptx
 
Neural networks
Neural networksNeural networks
Neural networks
 
Batch normalization presentation
Batch normalization presentationBatch normalization presentation
Batch normalization presentation
 
MNN
MNNMNN
MNN
 

More from ESCOM

redes neuronales tipo Som
redes neuronales tipo Somredes neuronales tipo Som
redes neuronales tipo Som
ESCOM
 
redes neuronales Som
redes neuronales Somredes neuronales Som
redes neuronales Som
ESCOM
 
redes neuronales Som Slides
redes neuronales Som Slidesredes neuronales Som Slides
redes neuronales Som Slides
ESCOM
 
red neuronal Som Net
red neuronal Som Netred neuronal Som Net
red neuronal Som Net
ESCOM
 
Self Organinising neural networks
Self Organinising  neural networksSelf Organinising  neural networks
Self Organinising neural networks
ESCOM
 
redes neuronales Kohonen
redes neuronales Kohonenredes neuronales Kohonen
redes neuronales Kohonen
ESCOM
 
Teoria Resonancia Adaptativa
Teoria Resonancia AdaptativaTeoria Resonancia Adaptativa
Teoria Resonancia Adaptativa
ESCOM
 
ejemplo red neuronal Art1
ejemplo red neuronal Art1ejemplo red neuronal Art1
ejemplo red neuronal Art1
ESCOM
 
redes neuronales tipo Art3
redes neuronales tipo Art3redes neuronales tipo Art3
redes neuronales tipo Art3
ESCOM
 
Art2
Art2Art2
Art2
ESCOM
 
Redes neuronales tipo Art
Redes neuronales tipo ArtRedes neuronales tipo Art
Redes neuronales tipo Art
ESCOM
 
Neocognitron
NeocognitronNeocognitron
Neocognitron
ESCOM
 
Neocognitron
NeocognitronNeocognitron
Neocognitron
ESCOM
 
Neocognitron
NeocognitronNeocognitron
Neocognitron
ESCOM
 
Fukushima Cognitron
Fukushima CognitronFukushima Cognitron
Fukushima Cognitron
ESCOM
 
Counterpropagation NETWORK
Counterpropagation NETWORKCounterpropagation NETWORK
Counterpropagation NETWORK
ESCOM
 
Counterpropagation NETWORK
Counterpropagation NETWORKCounterpropagation NETWORK
Counterpropagation NETWORK
ESCOM
 
Counterpropagation
CounterpropagationCounterpropagation
Counterpropagation
ESCOM
 
Teoría de Resonancia Adaptativa Art2 ARTMAP
Teoría de Resonancia Adaptativa Art2 ARTMAPTeoría de Resonancia Adaptativa Art2 ARTMAP
Teoría de Resonancia Adaptativa Art2 ARTMAP
ESCOM
 
Teoría de Resonancia Adaptativa ART1
Teoría de Resonancia Adaptativa ART1Teoría de Resonancia Adaptativa ART1
Teoría de Resonancia Adaptativa ART1
ESCOM
 

More from ESCOM (20)

redes neuronales tipo Som
redes neuronales tipo Somredes neuronales tipo Som
redes neuronales tipo Som
 
redes neuronales Som
redes neuronales Somredes neuronales Som
redes neuronales Som
 
redes neuronales Som Slides
redes neuronales Som Slidesredes neuronales Som Slides
redes neuronales Som Slides
 
red neuronal Som Net
red neuronal Som Netred neuronal Som Net
red neuronal Som Net
 
Self Organinising neural networks
Self Organinising  neural networksSelf Organinising  neural networks
Self Organinising neural networks
 
redes neuronales Kohonen
redes neuronales Kohonenredes neuronales Kohonen
redes neuronales Kohonen
 
Teoria Resonancia Adaptativa
Teoria Resonancia AdaptativaTeoria Resonancia Adaptativa
Teoria Resonancia Adaptativa
 
ejemplo red neuronal Art1
ejemplo red neuronal Art1ejemplo red neuronal Art1
ejemplo red neuronal Art1
 
redes neuronales tipo Art3
redes neuronales tipo Art3redes neuronales tipo Art3
redes neuronales tipo Art3
 
Art2
Art2Art2
Art2
 
Redes neuronales tipo Art
Redes neuronales tipo ArtRedes neuronales tipo Art
Redes neuronales tipo Art
 
Neocognitron
NeocognitronNeocognitron
Neocognitron
 
Neocognitron
NeocognitronNeocognitron
Neocognitron
 
Neocognitron
NeocognitronNeocognitron
Neocognitron
 
Fukushima Cognitron
Fukushima CognitronFukushima Cognitron
Fukushima Cognitron
 
Counterpropagation NETWORK
Counterpropagation NETWORKCounterpropagation NETWORK
Counterpropagation NETWORK
 
Counterpropagation NETWORK
Counterpropagation NETWORKCounterpropagation NETWORK
Counterpropagation NETWORK
 
Counterpropagation
CounterpropagationCounterpropagation
Counterpropagation
 
Teoría de Resonancia Adaptativa Art2 ARTMAP
Teoría de Resonancia Adaptativa Art2 ARTMAPTeoría de Resonancia Adaptativa Art2 ARTMAP
Teoría de Resonancia Adaptativa Art2 ARTMAP
 
Teoría de Resonancia Adaptativa ART1
Teoría de Resonancia Adaptativa ART1Teoría de Resonancia Adaptativa ART1
Teoría de Resonancia Adaptativa ART1
 

Recently uploaded

PIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf IslamabadPIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf Islamabad
AyyanKhan40
 
How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17
Celine George
 
LAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UP
LAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UPLAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UP
LAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UP
RAHUL
 
clinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdfclinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdf
Priyankaranawat4
 
S1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptxS1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptx
tarandeep35
 
Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...
Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...
Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...
National Information Standards Organization (NISO)
 
How to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold MethodHow to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold Method
Celine George
 
Hindi varnamala | hindi alphabet PPT.pdf
Hindi varnamala | hindi alphabet PPT.pdfHindi varnamala | hindi alphabet PPT.pdf
Hindi varnamala | hindi alphabet PPT.pdf
Dr. Mulla Adam Ali
 
Liberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdfLiberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdf
WaniBasim
 
Smart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICTSmart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICT
simonomuemu
 
Main Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docxMain Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docx
adhitya5119
 
The Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collectionThe Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collection
Israel Genealogy Research Association
 
Life upper-Intermediate B2 Workbook for student
Life upper-Intermediate B2 Workbook for studentLife upper-Intermediate B2 Workbook for student
Life upper-Intermediate B2 Workbook for student
NgcHiNguyn25
 
Advanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docxAdvanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docx
adhitya5119
 
How to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRMHow to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRM
Celine George
 
Cognitive Development Adolescence Psychology
Cognitive Development Adolescence PsychologyCognitive Development Adolescence Psychology
Cognitive Development Adolescence Psychology
paigestewart1632
 
Azure Interview Questions and Answers PDF By ScholarHat
Azure Interview Questions and Answers PDF By ScholarHatAzure Interview Questions and Answers PDF By ScholarHat
Azure Interview Questions and Answers PDF By ScholarHat
Scholarhat
 
How to Make a Field Mandatory in Odoo 17
How to Make a Field Mandatory in Odoo 17How to Make a Field Mandatory in Odoo 17
How to Make a Field Mandatory in Odoo 17
Celine George
 
Your Skill Boost Masterclass: Strategies for Effective Upskilling
Your Skill Boost Masterclass: Strategies for Effective UpskillingYour Skill Boost Masterclass: Strategies for Effective Upskilling
Your Skill Boost Masterclass: Strategies for Effective Upskilling
Excellence Foundation for South Sudan
 
ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...
ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...
ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...
PECB
 

Recently uploaded (20)

PIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf IslamabadPIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf Islamabad
 
How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17
 
LAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UP
LAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UPLAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UP
LAND USE LAND COVER AND NDVI OF MIRZAPUR DISTRICT, UP
 
clinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdfclinical examination of hip joint (1).pdf
clinical examination of hip joint (1).pdf
 
S1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptxS1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptx
 
Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...
Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...
Pollock and Snow "DEIA in the Scholarly Landscape, Session One: Setting Expec...
 
How to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold MethodHow to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold Method
 
Hindi varnamala | hindi alphabet PPT.pdf
Hindi varnamala | hindi alphabet PPT.pdfHindi varnamala | hindi alphabet PPT.pdf
Hindi varnamala | hindi alphabet PPT.pdf
 
Liberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdfLiberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdf
 
Smart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICTSmart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICT
 
Main Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docxMain Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docx
 
The Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collectionThe Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collection
 
Life upper-Intermediate B2 Workbook for student
Life upper-Intermediate B2 Workbook for studentLife upper-Intermediate B2 Workbook for student
Life upper-Intermediate B2 Workbook for student
 
Advanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docxAdvanced Java[Extra Concepts, Not Difficult].docx
Advanced Java[Extra Concepts, Not Difficult].docx
 
How to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRMHow to Manage Your Lost Opportunities in Odoo 17 CRM
How to Manage Your Lost Opportunities in Odoo 17 CRM
 
Cognitive Development Adolescence Psychology
Cognitive Development Adolescence PsychologyCognitive Development Adolescence Psychology
Cognitive Development Adolescence Psychology
 
Azure Interview Questions and Answers PDF By ScholarHat
Azure Interview Questions and Answers PDF By ScholarHatAzure Interview Questions and Answers PDF By ScholarHat
Azure Interview Questions and Answers PDF By ScholarHat
 
How to Make a Field Mandatory in Odoo 17
How to Make a Field Mandatory in Odoo 17How to Make a Field Mandatory in Odoo 17
How to Make a Field Mandatory in Odoo 17
 
Your Skill Boost Masterclass: Strategies for Effective Upskilling
Your Skill Boost Masterclass: Strategies for Effective UpskillingYour Skill Boost Masterclass: Strategies for Effective Upskilling
Your Skill Boost Masterclass: Strategies for Effective Upskilling
 
ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...
ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...
ISO/IEC 27001, ISO/IEC 42001, and GDPR: Best Practices for Implementation and...
 

Multilayer Slides

  • 1. Multi-layer feedforward network 1 1 Artificial Intelligence I1 H1 O1 I2 Backpropagation training of H2 multi-layer feedforward neural networks Ii: Input unit i I3 Oi: Output unit i Hi: Hidden unit i CS 4633/6633 Artificial Intelligence CS 4633/6633 Artificial Intelligence Training multi-layer networks Backpropagation: The idea q Supervised learning: A teacher presents input patterns to q The key idea is that the error in the output/activation of a the network, compares the resulting outputs with the neuron in a hidden layer can be defined as the weighted desired outputs, and adjusts the network weights to reduce sum of the error of its successor neurons. the error q Learning involves changing the weights of the network q For perceptron networks, adjusting weights to reduce error q Weights are changed based partly on the derivative of the is straightforward because there is only one weight activation function of a neuron. The sigmoid function has a between each input and output unit simple derivative: if y = 1/(1 + e-x), then dy/dx = y(1-y). q For multi-layer networks, adjusting weights to reduce error is more complicated because there are many weights between each input unit and output unit, and each of these weights contributes to more than one output unit CS 4633/6633 Artificial Intelligence CS 4633/6633 Artificial Intelligence Backpropagation is Backpropagation: The Algorithm Initialize network weights to random numbers between -0.5 and 0.5 hill-climbing search Repeat until error is sufficiently small (each repetition is an epoch) q Backprogagation is search in weight space, where the number of dimensions of the search space is equal to the For each training example: number of weights. Propagate input through network to compute output q It is possible for backpropagation to converge to a local Propagate errors backward through network, as follows: minimum, although it rarely happens in practice For each output unit i, calculate the error: q α is learning rate. If too low, learning is unnecessarly Errorj ← Oi × (1 - Oi) × (Ti - Oi) slow. If too high, it does not converge. For each hidden unit i, calculate the error: q Momentum: can improve learning time by making the Errori ← Oi × (1 - Oi) × ∑i (Wi,j × Errorj) weight adjustment proportional to the amount of the Update each network weight: previous weight change Wi,j ← Wi,j + α × Ii × Errorj CS 4633/6633 Artificial Intelligence CS 4633/6633 Artificial Intelligence
  • 2. Cross-validation training Generalization q A statistical approach for avoiding overfitting q The data set is randomly partitioned into a training set and a test set. The training set is further partitioned into two subsets, one for training and one for validation. Training set q Performance will always keep improving on the training % correct set, due to overfitting. When performance stops improving Validation set on the validation set, training is stopped. Then the network is tested on the test set. – train = training set – test1 = validation set for determining convergence Training time (# of epochs) – test2 = test set CS 4633/6633 Artificial Intelligence CS 4633/6633 Artificial Intelligence Image files Network structure q Each file contains a 32 x 30 image, with each pixel an ASCII q All networks have three layers (one hidden layer) character that represents a greyscale intensity value between 0 q For Glickman recognition problem (black) and 255 (white) – 960 input units, one for each pixel in image q attribute-value pairs – 1 output unit, yes or no – userid (20 possible values) – 4 hidden units – pose (straight, left, right, up) q For sunglasses recognition problem – expression (neutral, happy, sad, angry) – same – eyes (open, sunglasses) q For 1-of-20 face recognition problem q Problem: learn function from image to attribute-value pair(s) – 20 output units – map image to yes if Glickman, otherwise no – suggestion: 20 hidden units – map image to yes if wearing sunglasses, otherwise no q For pose recognition problem – map image to userid – 4 output units – map image to pose – suggestion: 6 hidden units CS 4633/6633 Artificial Intelligence CS 4633/6633 Artificial Intelligence Changes to code Pointers q How well a network learns depends on number of hidden q Sunglasses recognition: units, learning rate, etc. You are given these. Change the target output by changing the function q Backpropagation algorithm may not train a net to achieve “load_target” in imagenet.c, as follows. perfect classification accuracy if (!strcmp(eyes, “sunglasses”)) { q 1-of-20 face recognition and pose recognition: q To test network w/o training it, set command-line parameter for epochs to zero Change number of output and hidden units by changing arguments to bpnn_create in facetrain.c. Change the q Output lines: epoch number, sum of error over all units, for evaluation routines “performance_on_imagelist()” and training set: percentage correct and average error, for test “evaluate_performance” in facetrain.c to reflect the new set1: percentage correct and average error, for test set2: network structure. Also change the target outputs by percentage correct and average error changing the function “load_target” in imagenet.c. q Not required to do questions 11 and 12. But consider drawing learning curves to show results of training. CS 4633/6633 Artificial Intelligence CS 4633/6633 Artificial Intelligence