NEURAL NETWORKS & DEEP LEARNING
NEURAL NETWORKS & DEEP LEARNING
This course focuses on understanding and working with advanced deep learning models used in tasks such as image recognition, natural language processing, and generating new data. It covers key neural network models, including Convolutional Neural Networks, Recurrent Neural Networks, Transformers, and Generative Models like Autoencoders and Generative Adversarial Networks. The course aims to provide both the theoretical knowledge behind these models and practical skills to implement them.
The course starts by introducing the basics of deep feedforward networks, providing a foundation for understanding more complex models. It then moves on to CNNs, which are particularly useful for processing image and time-series data. You will learn how CNNs capture patterns such as edges and textures and how they reduce the need for large amounts of data by sharing weights across different parts of the input.
Next, the course covers Recurrent and Recursive Neural Networks. These models are designed to handle sequential data, such as time-series or text, by learning patterns from past inputs. The challenges of training these models, such as handling long-term dependencies and dealing with problems like vanishing gradients, will also be discussed.
The course also explores Transformers, which have become crucial in natural language processing. You will learn about the self-attention mechanism that allows Transformers to handle long-range dependencies more effectively than RNNs. This part of the course will help you understand why Transformers are now widely used in many machine learning tasks.
Additionally, the course includes an introduction to Generative Models like Autoencoders and GANs. Autoencoders are useful for tasks like data compression and anomaly detection, while GANs are used to generate new data that mimics real data. Variational Autoencoders (VAEs) will also be introduced, providing a probabilistic approach to data generation.
Alongside the models themselves, the course will cover important techniques for training and optimizing deep learning models. You will learn about optimization methods like gradient descent and regularization techniques that help prevent overfitting and improve model performance on new data.
By the end of the course, you will have a strong understanding of deep learning models and their applications. You will also gain practical experience in implementing these models, training them, and applying them to real-world tasks like image classification and data generation. This course will provide you with the knowledge and skills needed to work on complex machine learning problems in your future work.
- In collaboration with: Carlo Metta (CNR-ISTI)
- Estimated time: ≈ 6h