List view
The wiki needs to define the general vision, some of dirty details, and have instructions for building new simulations such that people who don't want to dig into the dirty stuff can still enjoy watching semi-intelligent agents do things.
No due date•0/2 issues closedThe deep neural network gaussian mixture model experiential model works by layering a DNN on top of a GMM. The DNN weights and biases act as the trainable parameters, and the GMM coefficients are all a function of this network. The mixture model allows for the capturing of bimodal data and a concise representation of model that can be evaluated quickly by a spider in the decision process.
No due date•0/2 issues closedThis project strives to do a number of things, among them: build an environment in which learning can take place, build agents that can interact with that environment, build different ways in which that agent can learn, etc. An important goal, then, is to have the many pieces of the puzzle interact modularly. For example, the learning environment should be separate from the agents, and it should be easy to run the same, or different agents, in a variety of learning environments, without modifying agent code. Zooming in, a modular system for experiential modeling means that it should be easy to switch the model type, or to add new model types, without modifying the agent code. Each component then needs a reliable sort of API that can be consistently referred to by other elements, and that suitably hides what doesn't concern other components. As pieces and components are added, their interactions will change slightly, but modular interaction should be maintained and updated throughout these additions. Thus, this milestone has no due date, and should remain active throughout the project.
No due date•0/1 issues closed