MSAI 531 - Neural Networks and Deep Learning

Credit(s): 3 hours
This course explores the principles of neural networks, starting from the basic building blocks and progressing to complex architectures. The course topics include: activation functions, feed-forward networks, backpropagation, gradient descent optimization, regularization techniques, and model evaluation. In addition, this course covers deep learning, convolutional neural networks (CNNs) for image analysis and recognition, recurrent neural networks (RNNs) for sequential data modeling, and attention mechanisms for tasks like language processing and machine translation.

Learning Objectives:

•Develop an understanding of the core concepts underlying neural networks, including neuron functioning, activation functions, and forward and backward propagation principles.

•Demonstrate the ability to design and implement neural network architectures, ranging from basic feedforward networks to advanced models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs).

•Learn various optimization methods, including gradient descent and its variants, and become skilled in selecting and applying appropriate optimization strategies to improve training efficiency and model performance.•Explore practical applications of neural networks in image analysis, natural language processing, and sequence modeling, and understand how to tailor network architectures to different problem domains.

•Develop an awareness of ethical challenges related to bias, fairness, and interpretability in deep learning models, and learn strategies to mitigate these issues when designing and deploying neural networks.

•Be exposed to advances and trends in neural network research, including: attention mechanisms, generative models like GANs and VAEs, and emerging techniques such as neural architecture search and explainable AI