Unveiling the Mystery of Deep Learning: Past, Present, and Future
Deep learning has
revolutionized artificial intelligence, but its journey from early
theoretical foundations to modern breakthroughs has been long and complex.
“Unveiling the Mystery of Deep Learning: Past, Present, and
Future” is a lecture series that will explore the historical evolution
of deep learning, tracing its origins from the early days of neural networks
in the 1980s to its resurgence in the 2010s and 2020s.
The series will be hosted by RCAC in conjunction with
Purdue’s Institute for Physical AI (IPAI) .
Dr. Elham Barezi, an AI Research Scientist for RCAC, continues to lead the course.
Throughout the series, participants will obtain a comprehensive understanding of what deep learning is, how it evolved, and where it is headed. The goal is to equip participants with deeper knowledge of AI’s development, enabling them to think critically about future innovations rather than just follow trends. By understanding the strengths and limitations of different deep learning techniques across time, participants will be better equipped to choose the most suitable approach for their specific problems and data.
The first session in the series focuses on the history of deep learning and AI research as a whole. Participants learnt about the beginnings of AI research in the 1960s, “AI Winters” and why deep learning remained dormant for decades, and how technological advancements have triggered its most recent rise. The session also covered how deep learning has developed over time, giving participants insight into precisely what deep learning is and how it works.
View and Download Session 1 Slides
The second session in the series focuses on fundamental discriminative deep learning models to explore the foundations of deep learning, including CNNs, RNNs, and early attention-based mechanisms before the Transformer revolution. It covers how deep networks extract and represent features, dives into autoencoders, and aims to develop an understanding of the role of deep learning in modern feature engineering and representation.
View and Download Session 2 Slides