This is a project based course, focusing on the science connecting the field of neural network with human brain mechanism, as well as advancements which are at the front of the field. We start by introducing a few neural network architectures with their learning paradigms, including deep feedforward and recurrent networks, Hopfield’s memory, and Kohonen’s self organizing maps. We continue by building on the top of these: clique structures and brain-like updatable architectures, explainable AI, combining symbolic with subsymbolic into one strong AI, robust AI, deep fake applications, and up to lifelong learning. The course is project based – with grades based on active class participation, presenting non-trivial topics, research project that will be done during the semester, and research paper written based on this work. I’ll assist students to get their research submitted to publications, if their work will reach high level. Students in 591NR will be able to run simpler projects. COMPARISON BETWEEN COMPSCI 682 AND 591NR/691NR: 682 teaches the engineering techniques necessary to train current neural networks. 591NR/691NR provides far larger view of the field, focusing on the science – both relation with human and natural intelligence, and on the advancements which are in the forefront of the neural networks field. 3 credits. This course counts as a CS Elective toward the CS major (BA or BS).
Undergraduate Prerequisites: COMPSCI 311.
Class Hours: Tuesday/Thursday 10:00AM – 11:15AM