Neural Networks Basics

This course is designed to provide a comprehensive introduction to neural networks and their applications in the Transportation & Logistics industries. Participants will learn the fundamental concepts, architectures, and training techniques of neural networks, enabling them to develop and deploy neural network models for various tasks relevant to transportation and logistics, such as demand forecasting, route optimization, and predictive maintenance.

• Understand the biological inspiration and mathematical foundations of neural networks

• Implement and train feedforward neural networks using Python and deep learning frameworks

• Apply backpropagation and gradient-based optimization techniques for neural network training

• Design and tune neural network architectures for specific Transportation & Logistics use cases

• Evaluate and interpret the performance of neural network models using appropriate metrics and techniques

1. Introduction to Neural Networks

• Historical context and biological inspiration behind neural networks

• Artificial neurons and activation functions

• Single-layer perceptrons and the XOR problem

• Hands-on exercises: Implementing a single-layer perceptron in Python

2. Feedforward Neural Networks

• Multi-layer perceptrons (MLPs) and their architecture

• Forward propagation and the universal approximation theorem

• Activation functions (sigmoid, tanh, ReLU) and their properties

• Hands-on exercises: Building and training MLPs using Python and NumPy

3. Training Neural Networks

• Cost functions and optimization objectives

• Backpropagation algorithm and gradient descent

• Stochastic gradient descent and mini-batch training

• Regularization techniques (L1/L2 regularization, dropout, early stopping)

• Hands-on exercises: Implementing backpropagation and training MLPs on transportation and logistics datasets

4. Neural Network Architectures and Hyperparameter Tuning

• Deeper architectures and their advantages

• Convolutional Neural Networks (CNNs) for spatial data (e.g., satellite imagery, traffic maps)

• Recurrent Neural Networks (RNNs) for sequential data (e.g., time series, GPS trajectories)

• Hyperparameter tuning and model selection techniques

• Hands-on exercises: Designing and tuning neural network architectures for Transportation & Logistics tasks

5. Applications and Advanced Topics

• Case studies of neural networks in the Transportation & Logistics industries (e.g., demand forecasting, route optimization, predictive maintenance)

• Unsupervised learning with neural networks (autoencoders, self-organizing maps)

• Introduction to deep learning frameworks (TensorFlow, PyTorch)

• Hands-on exercises: Developing a neural network model for a specific Transportation & Logistics use case

• Strong understanding of linear algebra, calculus, and probability theory

• Proficiency in programming with Python and libraries such as NumPy and Matplotlib

• Familiarity with basic machine learning concepts and techniques