Practicum for Deep Neural Networks

In the last decade, Neural Networks (NN) have attracted a lot of research due to their immense application potential. Breakthroughs of Deep Learning in image classification, speech recognition, and other challenging areas have provided the best solutions to many problems and significantly advanced “state of the art” AI machines in their ability to learn from data.

This course will introduce students to fundamental concepts of Deep Neural Network (DNN) development. It will cover important algorithms and modelling used in the development of these networks. Utilizing rich sets of available NN models, we will perform detailed analysis of leading ML approaches and popular NN like CNN, FCNN, ImageNet, ResNet, and RNN. TensorFlow framework exercises will develop practical skills, allowing students to gain confidence in understanding what ML concepts mean in neural network spaces. By the end of this course you will have learned important deep learning concepts and skills to create and train neural networks for specific practical problems.

The goal of this course is to open career opportunities in the highly sought-after fields of machine learning and AI. The key topics covered will provide insights into how AI can revolutionize your career. Students will also gain a better understanding of the advantages and limitations of Generative AI.
 

Learning Outcomes:

• Understand the architecture of deep neural networks (DNN), including different families of DNN models and the taxonomy of DNN architectures.
• Build and train a multilayer perceptron, adapting it to include different layer architectures and comparing performance.
• Apply convolutional neural networks (CNN) in image classification and segmentation tasks.
• Utilize DNNs in natural language processing applications.
• Introduce the concept of Generative AI, focusing on autoregressive model architecture.
• Explore Transformers architecture with a practical walkthrough for building a GPT model for text generation.

Course typically offered: Online in Summer and Winter quarters

Prerequisites: CSE-41287: Linear Algebra for Machine Learning and CSE-41305: Probability and Statistics for Deep Learning or equivalent knowledge

Next Steps: Upon completion, consider additional coursework in our specialized certificate in Machine Learning Methods to continue learning

More Information: For more information about this course, please contact unex-techdata@ucsd.edu