GSoC 2020 project idea 16: Contextual Neurodevelopmental Dynamics

Building on our organization’s (Orthogonal Research and Education Laboratory) past two years as a Google Summer of Code participant, we seek to combine the work on two Artificial Intelligence paradigms: Developmental Braitenberg Vehicles and Contextual Geometric Structures. In the case of the former (Developmental Braitenberg Vehicles), we learned lessons on how to create computational analogues of nervous systems that emerge during the developmental period. Through a variety of implementations, these embodied nervous systems can perform behaviors such as spatial navigation, multisensory integration, and coordinated emergence. In the case of the latter (Contextual Geometric Structures), we were able to instantiate a series of mathematical models of brain function and cultural classifications into code.

This year’s project will involve integrating elements of these two initiatives into something called Contextual Neurodevelopmental Dynamics. We are interested in building artificial nervous systems that can expand their number of neurons and connections (nodes and links) over time, with each node possessing a representation that provides context to the information flow across the network. These representations will be based on a Contextual Geometric Structure (CGS) architecture, descriptions of which can be found in the CGS project repository. A CGS is a kernel that classifies a range of sensory inputs according to a set of beliefs specific to that kernel. Code-level implementation of both the Developmental Braitenberg Vehicle and CGS approaches currently exist in Python and Kotlin, while simulations have involved the use of genetic algorithms, physics-based and agent-based models, and Hebbian learning. Contextual Connection Machines are an instance of a meta-brain model, a hybrid model that captures multiple aspects of intelligent behavior and scales of the brain.

As a student, you will join our Representational Brains and Phenotypes group, as well as participate in our Saturday Morning NeuroSim meeting. You will also contribute to building upon emerging intra-organizational topics such as Meta-Brain Models and Cybernetics. The successful student should approach this project with both technical skill and intellectual curiosity. A background in programming languages such as Python, Kotlin, or Julia are essential, and an interest in techniques such as genetic algorithms or agent-based models are preferred.

Mentor: Bradly Alicea (bradly.alicea@outlook.com), Orthogonal Research and Education Laboratory (https://orthogonal-research.weebly.com/).

Hello All!
I am Ankit Gupta, a senior year engineering undergrad from Indian Institute of Technology (IIT), Kharagpur.

After our recent weekly meetup last Saturday and discussion about this project in the meeting, I went through the details of the project and realized this can be a good project to continue and build upon my existing work on Braitenberg Vehicles for the Orthogonal Research and Education Laboratory.

I have been a contributor to the Representational Brains and Phenotypes group and been an active member of the Saturday Morning NeuroSim meetings. And it should be a year of association with @b.alicea by now. I have always been fascinated by the idea behind Braitenberg Vehicles and I feel I would continue to enjoy my time by being associated with this project as well.

I thank Bradly for always being there to guide and encourage my ideas and hoping to be guided on this project as well.

1 Like

Hello! I’m Ishita, an undergraduate studying computer science and neuroscience in the Bay Area. I have experience with C++, Python, Swift and I’ve worked with frameworks like Tensorflow and OpenCV alongside doing neuroscience research. I hope to work on this project and once again, it’s great to meet all of you!

Hello again,

I actually joined Orthogonal Lab last year as part of their GSoC proposal period and have been a part of the Braitenberg Vehicle project, along with Ankit above and some others. (See our Preprint here). I’ve been a part of the RB&P group, Saturday Morning NeuroSim , and also the adjacent DevoWorm / DevoWorm Machine Learning groups that Bradly is associated with via OpenWorm. I’ve presented a few times on neuronal dynamics and other cognition-related concepts, and will be continuing research along those lines going forward. Outside of previous work within Orthogonal Lab, I have experience with agent-based modeling, complexity and computation, and a strong interest in examining the mechanics and infrastructure that affords cognition.

I think combining the recent BV work with the prior contextual geometric structures will be great project to work on and fit with my own personal research agenda. Some of that is found in my FrontierMap project, embodied cognition research, and other collaborations within the lab.

I look forward to continuing our work and meeting new coders and researchers who may come to work with us along the way!

Feel free to connect on twitter at @JesParent or my website jesparent.com

Hello Again!
I submitted my final proposal a couple of days ago. Thanks again to @b.alicea, for providing some valuable feedback on my proposal idea.
Looking forward to working on the project this summer!

MANAGED BY INCF