Deep Learning-Based Artificial Intelligence

Credits: 
3
Hours: 
36
Area: 
Big Data Mining
Teachers: 
Vincenzo Lomonaco
Academic Year: 
2023-2024
Description: 

The module presents the methodological aspects, technologies and systems for designing predictive systems of Artificial Intelligence through machine learning and deep neural networks. The emphasis is placed on the analysis of application problems using examples and case studies, with practical exercises.

Prerequisites: Python & Data Mining & Machine Learning

Notions: 
  • Lecture 1: Fundamentals of Machine Learning for AI
    • Introduction to the course
    • Machine learning and AI
    • Machine learning paradigms
    • Model Selection
    • Hands-on Session for data processing with Numpy / Scikit-learn
  • Lecture 2: Fundamentals of Neural Networks
    • Biological and Artificial Neuron
    • Logistic regression as a neural network with 1 Neuron
    • Hands-on Session with Numpy
  • Lecture 3: Neural Network Training
    • Optimization algorithms
    • Stochastic-Gradient Descent (SGD) and Backpropagation
    • Tricks and Tips for NNs training
    • Hands-on Session with Keras (training) and Tensorboard
  • Lecture 4: Multi-layer Perceptron (4 hours)
    • From shallow networks to deep learning
    • MLP and Deep Feedforward Networks
    • Solving an image classification problem with Keras
  • Lecture 5: Convolutional Neural Networks (CNN) (4 Hours)
    • Visual processing with neural networks
    • Convolutions and CNN building blocks
    • Hands-on Session with Keras and CNNs
  • Lecture 6: Recurrent Neural Networks (RNN) (4 Hours)
    • Vanilla Recurrent Networks
    • Gated Recurrent Models (LSTM)
    • Hands-on Session with Keras for Sequence Classification
  • Lecture 7: Neural Networks Applications (4 hours)
    • Deep neural networks tools and libraries (Tensorflow and Pytorch)
    • NNs for Computer Vision Applications (Recognition, Detection and Segmentation)
    • NNs for Time-series (forecasting and classification)
    • Use of pre-trained models in Keras
  • Lecture 8: Advanced Deep Learning topics (4 Hours)
    • Autoencoders
    • Generative models
    • Continual Learning
    • Recent applications
  • Lecture 9: Application of AI-based Deep Learning Methods 
     
Technics and tools: 
  • Tensorflow
  • Keras
  • Pytorch
  • Scikit-learn
  • Numpy
  • Pandas
  • Matplotlib
  • Seaborn
Competences: 

The student will acquire knowledge and skills on the main technologies for machine learning through deep neural networks. He will also have references to the application problems of Artificial Intelligence and basic knowledge for the application of these methodologies to new problems.

Partners