-
Notifications
You must be signed in to change notification settings - Fork 13
Brain Designer
Add & Remove neurons / layers - see how they affect the squid's ability to process his environment
Click the brain button in the bottom right of the network tab to swich into Designer mode
FROM LEFT TO RIGHT:
- -- Fast-Generate a random network with no prompt
- -- Generate a random network with controls
- -- Add a new Custom/Input Neuron
- -- Export the brain to a .json file
- -- Clear all connections
- -- Quick-generate a network from template
If designer mode is launched from within the simulation (default behaviour):
- The running brain will be automatically imported into the designer.
- The purple 'Switch to Game' button passes the brain from the designer to the running brain tool
Some buttons will not be available if the designer is launched standalone (main.py -designer)
-
to add a neuron -
Mouse wheel on the canvas to zoom in/out, hold right mouse button to drag
-
Click on a neuron and hold to drag a connection line - release on another neuron
-
Select a link and use the mouse wheel to increase/decrease the connection weight (range: -1.00 to +1.00)
-
negative weights are inhibitory (red), positive weights are excitory (green).
-
Press DEL to delete a connection link
-
SPACE to change the direction of a connecting line (indicated with arrows)
-
Stronger weights are indicated with thicker lines.
-
Ctrl-click-and drag to reposition any neuron.
-
Use the Layers tab to create more complicated brains with additional layers
-
The Sensors tab shows available INPUT neurons
-
The Output tab lists Output bindings for bridging neuron activations to game behaviours:
Choose a source neuron, a trigger, and behaviour for that trigger
EXAMPLES:
IF .. can_see_food .. THEN .. change colour
IF .. happiness = <30 .. THEN .. pick up a rock
IF .. anxiety = >60 .. THEN .. seek plant
These can be combined to create reactive behaviours (simulated biological drives/urges)
brains/change_colour_when_see_food.json is an example squid that turns green when he can see food
The 'generate sparse network' button creates biologically-inspired neural networks using the core 8 neurons. Each generated brain is unique and randomised.
🦑 Raise digital squids whose brains grow & rewire themselves through Hebbian learning and Neurogenesis