Deep Learning: Model Design and Application

This course addresses the foundational concepts and components of Artificial Neural Networks (ANN), highlighting their capabilities, strengths, and weaknesses as a machine learning algorithm. Students taking this course will develop ANN models from scratch in Python as a basis for understanding their design as well as the underlying mechanics and calculations that shape their behavior. Key topics such as forward-backward propagation, loss function characteristics and optimization will be considered in relation to model design and computational efficiency as well as to problems such as exploding and vanishing gradients. Training strategies (e.g., dropout, initialization, batch normalization) will further enable students to assess trade-offs in model bias & variance. Coupled with hands-on assignments, these building blocks provide the knowledge and skills required to effectively design and implement ANN models that are ethically and technically sound. As well as foreground important architectures such as Convolutional ANNs, Recurrent ANNs, LSTMS, and Transformers as well as their applicability to modern problems. Student learning and proficiency will be assessed based on a combination of quizzes, coding assignments, exams, and a culminating project.