![Supervised Learning (Notation)
Unknown function:
Hypothesis:
f(x) =y
h(x) =y
h :ℝm → ℝ
h :ℝm → 𝒴, 𝒴 ={1,...,k}
Classification Regression
Training set: 𝒟 ={⟨x[i],y[i]⟩,i = 1,…,m},
"training examples"](https://image.slidesharecdn.com/week1-240228165744-cf62d14f/85/Week_1-Machine-Learning-introduction-pptx-30-320.jpg)
![Cont
Feature vector
X =
x1
x2
⋮
xm
x =
x1
x2
⋮
xn
X =
x[1] x[1]
1 2
x[1]
n
x[2] x[2] x[2]
x[m] x[m]
1 2
⋯
1 2 ⋯ n
⋮ ⋮ ⋱ ⋮
⋯ x[m]
n
Design Matrix Design Matrix
[i]T](https://image.slidesharecdn.com/week1-240228165744-cf62d14f/85/Week_1-Machine-Learning-introduction-pptx-32-320.jpg)
![Algorithm Categorization Schemes
• Discriminative vs generative
• Generative models (classically) describe methods that model the joint distribution P (X, Y ) =
P (Y )P (X|Y ) = P (X)P (Y|X) for training pairs < x[i], y[i] >.
• Discriminative models are taking a more “direct” approach for modeling P (Y|X) directly.
• While generative models provide typically more insights and allow sampling from the joint
distribution, discriminative models are typically easier to compute and produce more
accurate predictions.
• Discriminative modeling is like trying to extract information from text in a foreign language
without learning that language.
• Generative modeling is like generating text in a foreign language.](https://image.slidesharecdn.com/week1-240228165744-cf62d14f/85/Week_1-Machine-Learning-introduction-pptx-37-320.jpg)
Week_1 Machine Learning introduction.pptx
- 2. COURSE INFORMATION Course Number and Title: EC-452 Machine Learning Credits: 3-0 Instructor(s)-in-charge: Dr. Ahmad Rauf Subhani (Assistant Prof) Course type: Lecture Required or Elective: Elective Course pre-requisites Math-361 Probability and Statistics (Preferred) Degree and Semester DE-42 (Electrical), Semester 7 Month and Year Fall 2023
- 3. Assessment Course Assessment Exam: 1 Midterm and 1 Final Examination Assignment: ------- Quiz: 6 Quizes Grading: Quiz: 10-15% Assignments: 5-10% Mid Semester Exam: 30-35% Project 0-10% End Semester Exam: 40-50%
- 4. Topics covered in the Course
- 5. Introduction to Machine Learning • Machine learning is the field of study that gives computers the ability to learn without being explicitly programmed. — Arthur L. Samuel, AI pioneer, 1959 • A breakthrough in machine learning would be worth ten Microsofts. — Bill Gates, Microsoft Co-Founder
- 6. Introduction to Machine Learning
- 7. Introduction to Machine Learning
- 8. Introduction to Machine Learning • Machine Learning • Deep Learning • Artificial Intelligence
- 9. Introduction to Machine Learning
- 10. Introduction to Machine Learning • Machine Learning is a tool. • Like any other tool, it is important to read and understand its user manual. • What are some other daily life tools? • Do we need a user manual of a pen or a tyre???
- 11. Introduction to Machine Learning • Do we need a user manual of a pen or a tyre???
- 13. Applications of Machine Learning • Email spam detection • Face detection and matching (e.g., iPhone X) • Web search (e.g., DuckDuckGo, Bing, Google) • Sports predictions • Post office (e.g., sorting letters by zip codes) • ATMs (e.g., reading checks) • Credit card fraud • Drug design • Medical diagnoses • Smart assistants (Apple Siri, Amazon Alexa, . . . ) • Product recommendations (e.g., Netflix, Amazon) • Self-driving cars (e.g., Uber, Tesla) • Language translation (Google translate) • Sentiment analysis • Chat GPT and Google Bard • The list goes on…
- 14. Exercise • While we proceed in the class, it is a good exercise to think about how machine learning could be applied in these problem areas or tasks listed above: What is the desired outcome? What could the dataset look like? Is this a supervised or unsupervised problem, and what algorithms would you use? (Something to revisit later in this semester.) How would you measure success? What are potential challenges or pitfalls?
- 15. Common Understanding • Feature: • A measurable property of the object (data) you're trying to analyze. • In datasets, features appear as columns • Feature variable, attribute, measurements, dimension • Examples/ Samples: • Entries in features columns • In datasets, examples/samples, instances, observations appear as row • Target, synonymous to • outcome, ground truth, response variable, dependent variable, (class) label (in classification) • Output / Prediction • Use this to distinguish from targets; here, means output from the model
- 16. Common Understanding • Classification • A process of categorizing a given set of data (feature or example?) into classes. • The classes are often referred to as target, label or categories. • Regression • A technique for investigating the relationship between independent variables or features and a dependent variable or outcome. It's used as a method for predictive modelling in machine learning, in which an algorithm is used to predict continuous outcomes. y x x2 x1
- 17. Categories of Machine Learning Supervised Learning Unsupervised Learning Reinforcement Learning Labelled data Direct feedback Predict outcome/future No labels/target No feedback Find hidden structure in data Decision process Reward system Learn series of actions Source: Raschka & Mirjalili: Python Machine Learning, 2nd Ed.
- 19. Supervised Learning • Learning from labeled training data • Inputs that also contain the desired outputs or targets; basically, “examples” of what we want to predict. x2 x1 Illustration of a binary classification problem (plus, minus) and two feature variable (x1 and x2). (Source: Raschka & Mirjalili: Python Machine Learning, 2nd Ed.)
- 20. Supervised Learning y x Illustration of a linear regression model with one feature (predictor) variable (x1) and the target (response) variable y. The dashed-line indicates the functional form of the linear regression model. (Source: Raschka & Mirjalili: Python Machine Learning, 2nd Ed.)
- 21. Unsupervised learning • Unsupervised learning is concerned with unlabelled data • Common tasks in unsupervised learning are clustering analysis (assigning group memberships) and dimensionality reduction (compressing data onto a lower- dimensional subspace or manifold) x2 x1 Illustration of clustering, dashed lines indicate potential group membership assignments of unlabeled data points. (Source: Raschka & Mirjalili: Python Machine Learning, 2nd Ed.)
- 22. Unsupervised learning • Dimensionality reduction
- 23. Reinforcement learning • The process of learning from rewards while performing a series of actions • We do not tell the learner, for example, a (ro)bot, which action to take • But merely assign a reward to each action and/or the overall outcome. • Instead of having “correct/false” label for each step, the learner must discover or learn a behavior that maximizes the reward for a series of actions. • Not a supervised setting and somewhat related to unsupervised learning Illustration of reinforcement learning (Source: Raschka & Mirjalili: Python Machine Learning, 2nd Ed.)
- 24. Common Understanding (Jargons) • Feature: • A measurable property of the object (data) you're trying to analyze. • In datasets, features appear as columns • Predictor, variable, independent variable, input, attribute, covariate • Examples/ Samples (of training and testing): • Entries in features columns • In datasets, examples/samples appear as row • Observation, training record, training instance, training sample (in some contexts, sample refers to a collection of training examples) • Target, synonymous to • outcome, ground truth, output, response variable, dependent variable, (class) label (in classification) • Output / Prediction, use this to distinguish from targets; here, means output from the model
- 25. Common Understanding (Jargons) • Identify features and examples in the following data?
- 26. Common Understanding (Jargons) • Supervised learning: • Learn function to map input x (features) to output y (targets) • Structured data: • Databases, spreadsheets/csv files • Unstructured data: • Features like image pixels, audio signals, text sentences (before DL, extensive feature engineering required)
- 27. Common Understanding (Jargons) • Unstructured data
- 29. A Roadmap for Building Machine Learning Systems Label s Ra w Dat a Training Dataset Test Dataset Label s Learnin g Algorith m Preprocessi ng Learnin g Evaluatio n New Data Labels Predic tion Final Model Feature Extraction and Scaling Feature Selecti on i Dimensionality Reduction Sampling Model Selection Cross-Validation Performance Metrics Hyperparameter Optimization Mostly not needed in DL
- 30. Supervised Learning (Notation) Unknown function: Hypothesis: f(x) =y h(x) =y h :ℝm → ℝ h :ℝm → 𝒴, 𝒴 ={1,...,k} Classification Regression Training set: 𝒟 ={⟨x[i],y[i]⟩,i = 1,…,m}, "training examples"
- 31. Data Representation x = x1 x2 ⋮ xn Feature vector
- 32. Cont Feature vector X = x1 x2 ⋮ xm x = x1 x2 ⋮ xn X = x[1] x[1] 1 2 x[1] n x[2] x[2] x[2] x[m] x[m] 1 2 ⋯ 1 2 ⋯ n ⋮ ⋮ ⋱ ⋮ ⋯ x[m] n Design Matrix Design Matrix [i]T
- 33. m = n = Data Representation (structured data) 33
- 34. Entire hypothesis space Hypothesis space a particular learning algorithm category has access to Hypothesis space a particular learning algorithm can sample Particular hypothesis (i.e., a model/classifier) Hypothesis Space
- 35. Classes of Machine Learning Algorithms Below are some classes of algorithms that we are going to discuss in this class: • Generalized linear models (e.g., logistic regression) • Support vector machines (e.g., linear SVM, RBF-kernel SVM) • Artificial neural networks (e.g., multi-layer perceptrons) • Tree- or rule-based models (e.g., decision trees) • Graphical models (e.g., Bayesian networks) • Ensembles (e.g., Random Forest) • Instance-based learners (e.g., K-nearest neighbors
- 36. Algorithm Categorization Schemes • Eager vs lazy learners • Eager learners process training data immediately • lazy learners defer the processing step until the prediction, e.g., the nearest neighbor algorithm. • Batch vs online learning • In batch learning, the model is learned on the entire set of training examples. • Online learners, in contrast, learn from one training example at the time. • It is common, in practical applications, to learn a model via batch learning and then update it later using online learning. • Parametric vs nonparametric models • Parametric models are “fixed” models, where we assume a certain functional form for f (x) = y. For example, linear regression with h(x) = w1x1 + ... + wmxm + b. • Nonparametric models are more “flexible” and do not have a prespecfied number of parameters. In fact, the number of parameters grows typically with the size of the training set. For example, a decision tree would be an example of a nonparametric model, where each decision node (e.g., a binary “True/False” assertion) can be regarded as a parameter.
- 37. Algorithm Categorization Schemes • Discriminative vs generative • Generative models (classically) describe methods that model the joint distribution P (X, Y ) = P (Y )P (X|Y ) = P (X)P (Y|X) for training pairs < x[i], y[i] >. • Discriminative models are taking a more “direct” approach for modeling P (Y|X) directly. • While generative models provide typically more insights and allow sampling from the joint distribution, discriminative models are typically easier to compute and produce more accurate predictions. • Discriminative modeling is like trying to extract information from text in a foreign language without learning that language. • Generative modeling is like generating text in a foreign language.