š About Me ā Scott, Cognitive Model Developer
Hey there! Iām Scott ā part researcher, part developer, part neural network tinkerer.
I spent my graduate education focused in cognitive psychology, trying to understand how the brain works, and how to use software to build artificial neural network models.
I pursued software development after grad school, and I used my skills in writing code to improve my passion project: building an artificial brain system that can control an autonomous agent.
Over the years, I've built a toy cognitive model that controls both a virtual agent and a humanoid robot.
I'm using this site to show my key takeaways from my Research in cognitive science as well as my highlights from my Development projects.
Hope you enjoy!
š§ Cognitive Neuroscience Research
I focused my graduate studies around how decision-making emerges in the prefrontal cortex, amygdala, and basal ganglia. I tried to tie together leading theories of goal-driven executive function, reward learning, and conflict detection into a basic model that could be implemented in code.
Research Highlights:
- Simple neural networks can be used to represent individual modules in the prefrontal cortex (PFC)
- The basal ganglia is looking for matches of representations from PFC and perception to initiate motor plans
- Emotions are states of the system detecting key points of progress in the effort of obtaining a goal
- The "Still-Face" experiment in children shows a great cascade of emotion states
- Intelligent systems need more than just RL algorithms and more than just deep learning nets
- The PFC has middle (medial) and outside (lateral) regions that can represent goal "progress" states and "targets" in the world, respectively
Check out the research: Human Emotion Model
š¤ Cognitive Model Controlling Humanoid Robot
With a toy model design of a brain system, I wrote a Python application that collects sensory inputs and drives behavior in virtual and physical systems.
Project Highlights:
- Goal-driven behavior requires sensory input and body need state input
- A goal tracking (emotional) system allows an agent to find new approaches to goals when obstacles appear
- Having learned from its world, the model-driven virtual agent "figured out" how to use a tool
- The emotion model-driven humanoid robot demonstrates fundamental social interaction skills
Check out the robot: Human Emotion Model in Humanoid Robot