Portale di Ateneo - En.Unibs.it

*Introduction
-Goals of the course
-Resources available
-Applications for Machine Learning
-What is Data Mining?
-Data Mining Tasks
-Issues in Machine Learning

*Concept Learning
-Notation
-Learning task
-Representing hypotheses
-Performance measure
-Concept Learning as search
-Find-S Algorithm
-List-Then-Eliminate Algorithm
-Candidate Elimination Algorithm
-Biased and Unbiased Learners

*Classification: Basic Concepts and Decision Trees
-Classification Techniques
-Decision Trees
-Appropriate Problems for Decision Trees
-Decision Tree Induction
-Splitting criteria
-Measures of Node Impurity
-Stopping Criteria for Tree Induction
-Inductive Bias in DT Learning
-Some Practical Issues in DT learning

*Overfitting in Decision Trees
-Learning a DT with Noisy Data
-Overfitting Example: Noisy Data
-Avoiding Overfitting
-Reduced-Error Pruning
-Remarks on Reduced-Error Pruning
-Converting A Tree to Rules

*Evaluating Hypotheses and Classification Algorithms
-Statistical methods for estimating accuracy
-Two Definitions of Error
-Sample Error as Estimator of True Error
-Confidence Intervals
-Comparing two hypotheses
-Evaluation of a learning algorithm
-Evaluation of a learning algorithm
-How can we evaluate the performance of a learning algorithm?
-K-Fold Cross Validation
-ROC curve
-Confusion Matrix

*Artificial Neural Networks
-Connectionist Models
-Perceptron
-Decision Surface of a Perceptron
-Linear Unit (unthresholded Perceptron)
-Gradient Descent
-Batch Gradient Descent
-Incremental (Stochastic) Gradient Descent
-Gradient Descent for Thresholded Unit
-Multilayer Networks (Feed-forward Networks)
-Sigmoid Unit
-Backpropagation Algorithm
-Derivation of the Backpropagation Rule
-Convergence and Local Minima
-What kind of functions can be learned?
-Overfitting in ANNs
-Recurrent Networks

*Bayesian Learning
-Basic Formulas for Probabilities
-Bayes Theorem
-Brute Force MAP Concept Learning
-Learning Algorithms vs. MAP Learners
-Bayes Optimal Classifier
-Gibbs Classifier
-Naive Bayes Classifier
-Learning to Classify Text
-Examples

*Natural Language Processing
-How to deal with text?
-One-hot vector
-Bag of Words
-N-Grams
-Term Frequency, Inverse Document Frequency
-Word Embedding
-Embedding Representation
-Sentiment Analysis

*Data Analysis
-Types of Attributes
-Types of data sets
-Data Quality
-Outliers
-Missing Values
-Data Preprocessing
-Sampling
-Curse of Dimensionality
-Feature Subset Selection
-Attribute Transformation
-Examples

*Cluster Analysis
-Types of Clusterings
-Types of Clusters
-Clustering Algorithms
-K-means Clustering
-Bisecting K-means
-Hierarchical Clustering
-DBSCAN
-Cluster Validity
-Measures of Cluster Validity
-Examples

*Ensemble Methods
-Combining multiple models
-Bias-variance decomposition
-Bagging
-Random Forests
-Boosting
-Stacking
-Examples

*Support Vector Machines
-Key Ideas
-Hyperplanes as decision surfaces
-Maximizing the Margin
-Support Vectors
-Non-Separating Classes
-Support Vector Classifier
-Kernel Functions
-Kernels
-OVO: One versus One
-OVA: One versus All
-Examples

*Regression
-Classification and Regression
-Linear regression
-Logistic regression
-Maximum Likelihood
-Examples

*Machine Learning and Data Mining with Weka
-Getting started with Weka
-Exploring the Explorer
-Building a classifier
-Using a filter
-Visualizing your data
-Examples

*Machine Learning and Data Mining with R
-RStudio
-R: session management
-Basic data types
-Factors, Vectors and Matrices
-Lists and data frames
-Functions
-R graphics
-dplyr
-Connecting to Databases
-Examples