Elements of Applied Artificial Intelligence

Course Description
This course offers a pragmatic and applied approach to modern artificial intelligence. The originality of this subject lies in the joint use of Orange Data Mining, a "no-code/low-code" visual programming platform, and the Python language. This synergy allows students to quickly master the data pipeline (preprocessing, modeling, evaluation) through an intuitive graphical approach, while developing rigorous technical programming skills for the customization of complex algorithms.

Targeted Competencies

  • Identify and model AI opportunities in complex engineering problems.

  • Design complete Machine Learning pipelines via the Orange visual interface.

  • Master ethical foundations and algorithmic biases in data usage.

  • Ability to switch from a visual prototype (Orange) to a software implementation (Python).

Objectives

  • Master workflows in applied AI.

  • Gain an introduction to fundamental concepts (Machine Learning & Deep Learning) through visual experimentation.

  • Use Orange for exploratory visualization of multidimensional data.

  • Deepen Python knowledge for scripting custom widgets and using specialized libraries.

Prerequisites

  • Basics of algorithmics and Python programming.

Necessary Materials

  • Orange Data Mining software (with extensions: Image Analytics, Text Mining, Time Series).

  • Python environment (Anaconda or Google Colab) with libraries: NumPy, Pandas, Scikit-learn, Matplotlib, TensorFlow/PyTorch.

Course Content
Chapter 1: AI Foundations and Environments (01 week)

  • Definitions and fields of application in engineering sciences.

  • Overview of tools: From visual programming (Orange) to coding (Python).

  • Introduction to major domains: Supervised, unsupervised learning, and Deep Learning.

Chapter 2: Mathematics and Data Analysis (01 week)

  • Review of linear algebra and statistics for AI.

  • Visual data exploration (Distributions, Box Plots in Orange).

  • Simple linear regression: Formulation and implementation (Linear Regression widget vs. Scikit-learn).

  • Exercises: Matrix manipulation with NumPy and correlation visualization in Orange.

Chapter 3: Machine Learning Pipeline (03 weeks)

  • Key concepts: Data, features, labels, generalization, and overfitting.

  • Pipeline phases: Loading, Cleaning (Preprocess), Training, Validation.

  • Workflow architecture in Orange: Interaction between data and model widgets.

Chapter 4: Supervised Classification (03 weeks)

  • Principles of decision models.

  • Algorithms: SVM (Support Vector Machine), Decision Trees, and Random Forest.

  • Performance evaluation: Confusion matrix, precision, recall, F1-score.

  • Exercises: Real-time model comparison in Orange and Python scripting for specific metrics extraction.

Chapter 5: Unsupervised Learning and Clustering (02 weeks)

  • Logic of clustering: Partitioning data without labels.

  • Algorithms: K-means, DBSCAN, and Hierarchical Clustering.

  • 2D/3D visualization (MDS, PCA) and cluster interpretation.

  • Exercises: Segmentation of an industrial dataset and visualization of cluster boundaries.

Chapter 6: Neural Networks and Deep Learning (03 weeks)

  • Architecture: Artificial neuron (Perceptron), hidden layers, weights, bias, and activation functions (ReLU, Sigmoid).

  • Introduction to Deep Learning: Deep neural networks and Convolutional Neural Networks (CNN).

  • Transfer Learning: Using pre-trained models in Orange (Image Analytics extension).

  • Exercises: Image classification by drag-and-drop (Orange) and introduction to code with Keras/TensorFlow.

Chapter 7: Integrative Mini-Project (Supervised personal work)
Design of a complete AI system (Preprocessing -> Model -> Visualization) choosing from:

  • Predictive maintenance: Detection of sound or vibration anomalies on machines.

  • Computer vision: Character recognition or classification of manufacturing defects.

  • NLP: Sentiment analysis or company FAQ-oriented Chatbot.

  • Prediction: Time series analysis for natural disasters.

Practical Work (TP)

  • TP 01: Getting started with Orange (Basic Workflows) and Python environment.

  • TP 02: Exploratory analysis and Regression (Orange vs. Matplotlib/Seaborn).

  • TP 03: Advanced data preprocessing: Handling missing values and normalization.

  • TP 04: Supervised classification: Optimization of hyperparameters (SVM and Trees).

  • TP 05: Applied clustering: Analysis of data profiles via K-means and PCA.

  • TP 06: Deep Learning: Image classification (MNIST) and simple text analysis.

Evaluation Mode

  • Final exam: 60%

  • Continuous Assessment (Mini-project + TP): 40%