A ‘Simple’ Neural-Network Model

My previous blog post summarized the main points of the Watts 2002 paper. At the very end of the post I mentioned ways we can extend the Watts model with neural-networks. Here I outline the steps of the neural-network simulation. Generally the model goes though an initiation step, a simulation step, and a cleanup step. In the initiation step, the agents are created and connected in a network. Then selected agents are seeded with a value before continuing onto the simulation step.

Expanding the Watts Model

I wrote a blog post and gave a seminar talk about the Watts 2002 paper. The next step is thinking about expanding the binary decisions with externalities model to a psychological plausible decisions with externalities model. This has been the goal of the multi-agent neural-network (MANN) project. Context from the Watts model From an information diffusion perspective, a cascade is the spread of information from an initial set (seed) of individuals (nodes).

A Simple Model of Global Cascades on Random Networks

Duncan J Watts wrote a paper that was published in 2002 titled “A simple model of global cascades on random networks” in the Proceedings of the National Academy of Sciences (PNAS). It’s a seminal paper in my current work on information diffusion in (social) networks. Watts shows how the interactions between local dependencies, fractional threshold, and heterogeneity relate to information cascades in networks. My work builds on these ideas, so it’s important to have a strong understanding of the terms and model specifications outlined in the paper.