Artificial Intelligence with Bayesian Networks and BayesiaLab
Training Overview
 Teaching objectives: Comprehensive understanding of the Bayesian network paradigm plus practical skills for realworld research applications
 Length: 3 days
 Required Level: The course is taught at a beginner level, so no prior knowledge of Bayesian networks is necessary. However, undergraduatelevel familiarity with probability theory and statistics is recommended.
 Teaching methods: Tutorials with practical exercises using BayesiaLab plus plenty of oneonone coaching
 Trainer: Dr. Lionel Jouffe, CEO, Bayesia SAS.
 Training materials: A printed tutorial (approx. 300 slides), plus a memory stick containing numerous exercises and white papers
 Bayesian network Software: Bayesia provides all trainees with an unrestricted 90day license of BayesiaLab Professional Edition, so they can participate in all exercises on their own laptops
 Cost: Between 2,100 and 2,500 Euros/Trainee, depending on the location of the training session. Discounts are available for groups of trainees of the same company. A special academic discount of 50% is also available for students and teachers of accredited educational institutions.
Here is a link to a Prezi presentation that describes the entire learning journey, from the short introduction to Bayesian networks to the 3rd day of the Advanced course: https://prezi.com/view/VeZ1oBtGKQtLZHwwkVhn/
The registration is complete upon payment of the fee by Bank Transfer, or Credit Cards. Visit the BayesiaLab Store to get the prices corresponding to the type of your organization and number of seats your are interested in.
Training Program
Day 1: Theoretical Introduction
 Bayesian Networks: Artificial Intelligence for Decision Support under Uncertainty
 Probabilistic Expert System
 The Modeling World
 Bayesian Networks and Cognitive Science
 Unstructured and Structured Particles Describing the Domain
 Expert Based Modeling and/or Machine Learning
 Predictive (association) versus Explicative (causation) Models
 Application Examples: Medical Expert Systems, Stock Market Analysis, Microarray Analysis, Consumer Segmentation, Drivers Analysis and Product Optimization

 Probabilistic Axioms
 Perception of the Particles
 Joint Probability Distribution (JPD)
 Probabilistic Expert System for Decision Support: Types of Requests
 Leveraging Independence Properties
 Product/Chain Rule for Compact Representation of JPD

 Qualitative Part: Directed Acyclic Graph
 Graph Terminology
 Graphical Properties
 DSeparation
 Markov Blanket
 Quantitative Part: Marginal and Conditional Probability Distributions
 Exact and Approximate Inference in Bayesian networks
 Example of Probabilistic Inference: Alarm System

 ExpertBased Modeling via Brainstorming
 Why ExpertBased Modeling?
 Value of ExpertBased Modeling
 Structural Modeling: BottomUp and TopDown Approaches
 Parametric Modeling
 Cognitive Biases
 BEKEE: Bayesia Expert Knowledge Elicitation Environment


Day 2: Machine Learning  Part 1
 Maximum Likelihood Estimation
 Bayesian Parameter Estimation with Dirichlet Priors
 Smooth Probability Estimation (Laplacian Correction)

 Information is a Measurable Quantity: LogLoss
 Expected LogLoss
 Entropy
 Conditional Entropy
 Mutual Information
 Symmetric Relative Mutual Information
 KullbackLeibler Divergence

 Entropy Optimization
 Minimum Description Length (MDL) Score
 Structural Coefficient
 Minimum Size of Data Set
 Search Spaces
 Search Strategies
 Learning Algorithms
 Maximum Weight Spanning Tree
 Taboo Search
 EQ
 TabooEQ
 SopLEQ
 Taboo Order
 Data Perturbation
 Example: Exploring the relationships in body dimensions
 Data Import (Typing, Discretization)
 Definition of Classes
 Exclusion of a Node
 Heuristic Search Algorithms
 Data Perturbation (Learning, Bootstrap)
 Choice of the Structural Coefficient
 Console
 Symmetric Layout
 Analysis of the Model (Arc Force, Node Force, Pearson Coefficient)
 Dictionary of Node Positions
 Association of an Image in the Background

 Learning Algorithms
 Naive
 Augmented Naive
 Manual Augmented Naive
 TreeAugmented Naive
 Sons & Spouses
 Markov Blanket
 Augmented Markov Blanket
 Minimal Augmented Markov Blanket
 Variable selection with Markov Blanket
 Example: Predictions based on body dimensions
 Data Import (Data Type, Supervised Discretization)
 Heuristic Search Algorithms
 Target Evaluation (InSample, OutofSample: KFold, Test Set)
 Smoothed Probability Estimation
 Analysis of the Model (Monitors, Mapping, Target Report, Target Posterior Probabilities, Target Interpretation Tree)
 Evidence Scenario File
 Automatic EvidenceSetting
 Adaptive Questionnaire
 Batch Labeling


Day 3: Machine Learning  Part 2
 Algorithms
 Example: S&P 500 Analysis
 Variable Clustering
 Changing the number of Clusters
 Dynamic Dendrogram
 Dynamic Mapping
 Manual Modification of Clusters
 Manual Creation of Clusters
 SemiSupervised Learning
 Search Tool (Nodes, Arcs, Monitors, Actions)
 Sticky Notes

 Synthesis of a Latent Variable
 ExpectationMaximization Algorithm
 Ordered Numerical Values
 Cluster Purity
 Cluster Mapping
 LogLoss and Entropy of the Data
 Contingency Table Fit
 Hypercube Cells Per State
 Example: Segmentation of men based on body dimensions
 Data Clustering (Equal frequency discretization, MetaClustering)
 Quality Metrics (Purity, LogLoss, Contingency Table Fit)
 Posterior Mean Analysis (Mean, Deltameans, Radar charts)
 Mapping
 Cluster Interpretation with Target Dynamic Profile
 Cluster Interpretation with Target Optimization Tree
 Projection of the Cluster on other Variables

 PSEM Workflow
 Unsupervised Structural Learning
 Variable Clustering
 Multiple Clustering for Creating a Factor Variable (via data Clustering) per Cluster of Manifest Variables
 Unsupervised Learning for Representing the Relationships between the Factors and the Target Variables
 Example: The French Market of Perfumes
 Crossvalidation of the Clusters of Variables
 Displayed Classes
 Total Effects
 Direct Effects
 Direct Effect Contributions
 Tornado Analysis
 Taboo, EQ, TabooEQ, and Arc Constraints
 MultiQuadrants
 Export Variations
 Target Optimization with Dynamic Profile
 Target Optimization with Tree

