Google, IBM & Meta Certificates — All 10,000+ Courses at 40% Off
One annual plan covers every course and certificate on Coursera. 40% off for a limited time.
Get Full Access
Attend this plenary talk from JSALT 2025 exploring how differentiable forward modeling can revolutionize machine learning by incorporating scientific knowledge into deep learning architectures. Learn about cutting-edge research themes including differentiable modeling, optimization of Transformer attention mechanisms, and Large Audio Language Models from Professor Ramani Duraiswami of the University of Maryland. Discover how making traditional scientific forward models differentiable enables their integration with deep learning for parameter optimization, inverse problem solving, implicit neural representations, and explainable model development, particularly valuable in data-sparse domains. Explore practical applications across computer graphics, human hearing, room acoustics, signal processing, and mathematical physics inverse problems. Gain insights from a leading researcher whose work has resulted in two spin-off companies and whose audio engine technology powers millions of VR headsets, PCs, and headphones worldwide. The presentation covers the intersection of automatic differentiation, backpropagation, and scientific modeling, demonstrating how decades of scientific knowledge can be preserved and enhanced through modern machine learning techniques.
