This simulation, like the one above, shows how a network, world, and gauge can be used together to study the way agents represent their environments. The network was trained using backpropagation to discriminate the distributed patterns of input activity produced by the three objects. The points in the gauge correspond to the three objects perceived from different distances. The middle point corresponds to perceiving no object. (To view, open workspace: bp > 3-2-3_single.sim, and open a gauge).

This simulation uses competitive learning to distinguish specific types of flowers and cheese, as well as specific combinations of flowers and cheese. This simulation also shows that multiuple subnetworks can be linked together (To view, open workspace: sims > competitive > combinationDetector.sim).

This is just a big network of nodes which can be used to see how propagation of activity is visualized in Simbrain. (To view, open network: networks> localconns > biglocal.net).

This is a network of Integrate and Fire neurons, which produce discrete spiking events (shown as yellow lines). The this screenshot, two neurons are currently firing. (To view, open network: networks> intFire > 4sx6intFire.net).

This screenshot shows a hopfield subnetwork which is about to be trained on its current pattern. The grey weights indicate that all weights are currently at 0. The screenshot also demonstrates how users interact with subnetworks, via tabs and their pop-up menus. Such a network can be created by right-clicking in a network window and selecting New Network > Hopfield.

This is a close up of a gauge component. This gauge is using PCA to project patterns of activity in a 9-node network down to two dimensions. The red dot corresponds to the current state of the network being represented. The yellow inset shows a tooltip, which shows which high dimensional point corresponds to a given point in the gauge projection.

This is a neuron dialog box, which is called up whenever one or more selected neurons are double clicked. Available parameters change as neuron type is changed.

This is a synapse dialog box, which is called up whenever one or more selected synapses are double clicked. Available parameters change as synapse type is changed. This and the screenshot above give a sense of how many different synapse and neuron types are avaialable.

This is the interface used to train a backpropoagation network.

Various types of "subnetwork" can be created in Simbrain. Shown here is a creation dialog which appears when one creates a new Competitive network. Relevant parameters can also be changed after the subnetwork has been created.