Machine

Learning
last update 12-03-
2022 •Dr. Ghulam Gilanie Janjua
•PhD (Artificial Intelligence)

Building machines that can make

decisions based on common sense
is no easy feat. A machine must be
able to do more than merely find
patterns in data; it also needs a
way of interpreting the intentions
and beliefs behind people’s choices
 • Classification-
uses predefined classes in
which objects are assigned
or it is the task of
What Machine Learning
Machine predicting a discrete class
can solve
label
Learning • Clustering identifies
•Computational intelligence
(evolutionary algorithms,
similarities between etc.)
objects, which •Computer Vision (CV)
it groups according to
those characteristics in •Natural Language Processing
(NLP)
common and which
differentiate them from •Recommender Systems
other groups of objects. •Reinforcement Learning
These groups are known as •Graphical Models
"clusters. •And more…
• Regression is the task of
predicting a continuous
 1. Collecting Data
Machine
2. Preparing the Data
Learning 3. Choosing a Model
Application
Steps 4. Training the Model
5. Evaluating the
Model
6. Parameter Tuning
7. Making Predictions
 Common Algorithms
Types
•Linear Regression
•Supervised (the goal is to •Logistic Regression
generate formula based
Machine on input and output
•Decision Tree
•SVM
Learning values)
•Naive Bayes
•Un-supervised (we find •kNN
an association between •K-Means
input values and group •Random Forest
them)
•Dimensionality Reduction
•Reinforcement or semi- Algorithms
supervised Learning (an •Gradient Boosting
agent learn through algorithms
delayed feedback by • GBM
interacting with the • XGBoost
environment) • LightGBM
 Machine Learning Algorithms
Regression Algorithms Bayesian Algorithms Association Rule Learning Algorithms
• Ordinary Least Squares Regression • Naive Bayes • Apriori algorithm
• Linear Regression • Gaussian Naive Bayes
• Eclat algorithm
• Logistic Regression • Multinomial Naive Bayes
• Stepwise Regression • Averaged One-Dependence Estimators Deep Learning Algorithms
• Multivariate Adaptive Regression Splines • Bayesian Belief Network (BBN) • Convolutional Neural Network (CNN)
(MARS) • Bayesian Network (BN) • Recurrent Neural Networks (RNNs)
• Locally Estimated Scatterplot Smoothing Clustering Algorithms • Long Short-Term Memory Networks
(LOESS) • k-Means
•Instance-based Algorithms • k-Medians (LSTMs)
• k-Nearest Neighbor (kNN) • Expectation Maximisation (EM) • Stacked Auto-Encoders
• Learning Vector Quantization (LVQ) • Hierarchical Clustering • Deep Boltzmann Machine (DBM)
• Self-Organizing Map (SOM) Artificial Neural Network Algorithms • Deep Belief Networks (DBN)
• Locally Weighted Learning (LWL) • Perceptron
• Support Vector Machines (SVM) Dimensionality Reduction Algorithms
• Multilayer Perceptrons (MLP)
•Regularization Algorithms • Back-Propagation • Principal Component Analysis (PCA)
• Ridge Regression • Stochastic Gradient Descent • Principal Component Regression (PCR)
• Least Absolute Shrinkage and Selection Operator • Hopfield Network • Partial Least Squares Regression (PLSR)
• Elastic Net • Radial Basis Function Network (RBFN)
• Least-Angle Regression (LARS)
• Sammon Mapping
Ensemble Algorithms
•Decision Tree Algorithms • Boosting
• Multidimensional Scaling (MDS)
• Classification and Regression Tree (CART) • Bootstrapped Aggregation • Projection Pursuit
• Iterative Dichotomiser 3 (ID3) (Bagging) • Linear Discriminant Analysis (LDA)
• C4.5 and C5.0
• Chi-squared Automatic Interaction Detection
• AdaBoost • Mixture Discriminant Analysis (MDA)
• Weighted Average (Blending) • Quadratic Discriminant Analysis (QDA)
• Decision Stump
• Stacked Generalization
• M5
(Stacking)
• Flexible Discriminant Analysis (FDA)
• Conditional Decision Trees • Independent Component analysis (ICA)
• Gradient Boosting Machines
 Description
Purpose

Classification Apply pretrained networks directly to classification

problems. To classify a new image, use classify.

Feature Use a pretrained network as a feature extractor by

Common Extraction using the layer activations as features. You can use
these activations as features to train another
Approac machine learning model, such as a support vector
hes machine (SVM).

Transfer Learning Take layers from a network trained on a large data

set and fine-tune on a new data set.
 Network Depth Size Parameters (Millions) Image Input Size
squeezenet 18 5.2 MB 1.24 227-by-227
googlenet 22 27 MB 7.0 224-by-224
inceptionv3 48 89 MB 23.9 299-by-299
densenet201 201 77 MB 20.0 224-by-224
mobilenetv2 53 13 MB 3.5 224-by-224
resnet18 18 44 MB 11.7 224-by-224
resnet50 50 96 MB 25.6 224-by-224
resnet101 101 167 MB 44.6 224-by-224
xception 71 85 MB 22.9 299-by-299
inceptionresnetv2 164 209 MB 55.9 299-by-299
shufflenet 50 5.4 MB 1.4 224-by-224
nasnetmobile * 20 MB 5.3 224-by-224
nasnetlarge * 332 MB 88.9 331-by-331
darknet19 19 78 MB 20.8 256-by-256
darknet53 53 155 MB 41.6 256-by-256
efficientnetb0 82 20 MB 5.3 224-by-224
alexnet 8 227 MB 61.0 227-by-227
vgg16 16 515 MB 138 224-by-224
vgg19 19 535 MB 144 224-by-224
1. Machine Learning for Computer Vision (ANN Architecture)
❖ MNIST images Classification (multi-class classification, grayscale images) ❖ Male Female Classification (binary classification, grayscale images)
Machine learning for computer vision (CNN Architecture)
Applying techniques are:
● Data Augmentation ● Using Pre-trained model (VGG16, VGG19, Restnet18, Restnet50, Googlenet, Alexnet, etc.)
● Feature extraction with a pretrained model ● Fine-tuning a pre-trained model
❖ MNIST images Classification (multi-class classification, grayscale images) ❖ Cats vs. Dogs Classification (binary classification, RGB images)
❖skin_cancer_malignant_vs_benign (binary classification, RGB images) ❖ Sign Language Recognition (ASL) (multi-class classification, RGB
images)
❖ Male Female Classification (binary classification, grayscale images) ❖ Flowers Classification (variation of classes, Mixed dataset) (multi-class
classification, RGB images)
2. HeatMaps using a pre-trained model
❖ elephant_heatmaps_with_vgg16 ❖ heatmaps_with_InceptionV3
3. Images Segmentation
❖ Cats vs. Dogs Datasets (binary classification, RGB images)
4. Object Detection using Yolov5 Object Detection/Segmentation using MaskRCNN
❖ Create own datasets and trained model Classification and regression ❖Classifying movie reviews: A binary classification example (IMDB Datasets)
❖classifying newswires: A multiclass classification example (The Reuters dataset) ❖Predicting house prices: A regression example (The Boston housing price
dataset)
5. Deep learning for time series
❖Jena weather dataset (jena_climate_2009_2016.csv)
● Dense Network ● LSTM Network ● RNN Network Deep learning for text
❖ IMDB movie reviews data (binary classification)
● Processing words as a set:
The bag-of-words approach
● Processing words as a sequence: The sequence model approach Implementing DeepDream in Keras
imds = imageDatastore(Path','IncludeSubfolders',true,'LabelSource','foldernames');
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7,'randomized');
numTrainImages = numel(imdsTrain.Labels);

Transfer Learning using Alexnet

idx = randperm(numTrainImages,16);
net=alexnet;
analyzeNetwork(net)
inputSize = net.Layers(1).InputSize
layersTransfer = net.Layers(1:end-3);
numClasses = numel(categories(imdsTrain.Labels))
layers = [ layersTransfer
fullyConnectedLayer(numClasses,'WeightLearnRateFactor',20,'BiasLearnRateFactor',20)
softmaxLayer
classificationLayer];
pixelRange = [-30 30];
imageAugmenter = imageDataAugmenter( ... 'RandXReflection',true, ... 'RandXTranslation',pixelRange, ... 'RandYTranslation',pixelRange);
augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain, ... 'DataAugmentation',imageAugmenter);
augimdsValidation = augmentedImageDatastore(inputSize(1:2),imdsValidation);
options = trainingOptions('sgdm', ... 'MiniBatchSize',10, ... 'MaxEpochs',6, ... 'InitialLearnRate',1e-4, ... 'Shuffle','every-epoch', ...
'ValidationData',augimdsValidation, ... 'ValidationFrequency',3, ... 'Verbose',false, ... 'Plots','training-progress');
netTransfer = trainNetwork(augimdsTrain,layers,options);
[YPred,scores] = classify(netTransfer,augimdsValidation);
idx = randperm(numel(imdsValidation.Files),4);
figure
for i = 1:4
subplot(2,2,i)
I = readimage(imdsValidation,idx(i));
imshow(I)
label = YPred(idx(i));
title(string(label));
end
YValidation = imdsValidation.Labels;accuracy = mean(YPred == YValidation)
 Features Extraction from Pretrained and Classification Using SVM
imds = imageDatastore(‘Path','IncludeSubfolders',true,'LabelSource','foldernames');
[imdsTrain,imdsTest] = splitEachLabel(imds,0.7,'randomized');
numTrainImages = numel(imdsTrain.Labels);
idx = randperm(numTrainImages,16);
figure
for i = 1:16
subplot(4,4,i)
I = readimage(imdsTrain,idx(i));
imshow(I)
end
net = resnet18
inputSize = net.Layers(1).InputSize;
analyzeNetwork(net)
augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain);
augimdsTest = augmentedImageDatastore(inputSize(1:2),imdsTest);
layer = 'pool5';
featuresTrain = activations(net,augimdsTrain,layer,'OutputAs','rows');
featuresTest = activations(net,augimdsTest,layer,'OutputAs','rows');
whos featuresTrain
YTrain = imdsTrain.Labels;
YTest = imdsTest.Labels;
classifier = fitcecoc(featuresTrain,YTrain);
YPred = predict(classifier,featuresTest);
idx = [1 5 10 15];
figure
for i = 1:numel(idx)
subplot(2,2,i)
I = readimage(imdsTest,idx(i));
label = YPred(idx(i));
imshow(I)
title(char(label))
end
accuracy = mean(YPred == YTest)