DFT: From Neural Principles to Autonomous Neural Dynamic Agents
How can we build an autonomous agent that is adaptive, interpretable, and grounded in neural principles in 90 minutes?
This tutorial introduces Dynamic Field Theory (DFT), a mathematical framework for modeling cognition, perception, and action as the continuous evolution of neural activation patterns across time and space. Unlike static or purely data-driven approaches, DFT supports the development of autonomous agents that generate goal-directed behavior in real time, integrating perception, memory, and prior knowledge. The intrinsic stability properties of all neural representations in such DFT agents lead to scalability and enable online learning and adaptation. DFT enables neural process models that are transparent, interpretable, and that provide mechanistic explanations grounded in neural principles.
Participants will gain hands-on experience with DFT by collaboratively building a simple autonomous neural dynamic agent. The tutorial also introduces the latest applications of DFT in cognitive science, robotics, and hybrid approaches that integrate DFT with deep neural networks, combining the strengths of both methods.
Building a Neural Dynamic Agent (90 min)
Participants will be guided through the step-by-step construction of a simple autonomous neural dynamic agent in DFT. This agent will be developed live and continuously updated to solve a series of progressively more complex toy tasks in simplified environments. Each modeling step is carefully tied back to DFT's core theoretical concepts, allowing participants to gain both hands-on modeling insight and deeper conceptual understanding.
No prior experience with DFT is required.
Break and Networking (30 min)
An informal break with space for questions, networking, and informal discussions.
Latest DFT Applications (30 min)
DFT researchers will present short, focused flash talks (approximately 5 minutes each) on the latest DFT applications, illustrating the versatility of the framework across domains.
Open Discussion and Q&A (30 min)
We invite all participants to join an open discussion. Topics may include practical questions about applying DFT, conceptual debates around neural process modeling, or reflections on the role of neural dynamics in modern AI.