YOUR TARGET / TWO CONTROLLERS
Challenge the Fish.
Place a target, make it move or hide it briefly. See how the original and saved learned controller handle the same task.
The saved controller is frozen for this session. This challenge does not update the shared learner.
Original controller
Visual input onSaved controller
Visual input onWhat will the model do?
NO RESULT YETChoose a task and press Run challenge. Results are measured from the actual controller trajectories.
After a completed trial, the server independently repeats the task with the same parameters.
Trial events
MODEL TIMEMETHOD & SCOPE
One task. Two fixed sets of parameters.
A fair comparison. Both controllers start at the same position and see the same scheduled target movement and visibility changes. Their own movement changes their retinal input. Each uses the same model equations.
A new test, not new training. The learned gains come from the separate target-learning experiment. They are not guaranteed to win on your task. No weights change during a challenge, and no helper reveals the target.
An inspectable result. The server repeats completed trials from the supplied task and parameters. Shared links preserve those inputs. This confirms reproducibility, not biological validity or the origin of user-supplied weights.