Visualizer feedback

[bruno-f-cruz]

People want something closer to this:

The bias plot uses this function:
https://github.com/AllenNeuralDynamics/dynamic-foraging-task/blob/main/src/foraging_gui/bias_indicator.py
https://github.com/AllenNeuralDynamics/aind-dynamic-foraging-models/blob/main/src/aind_dynamic_foraging_models/logistic_regression/model.py

[bruno-f-cruz]

• Add rolling p_non_aborted
• Add manual water modifier to rewarded ticks
• Add bias plot

[bruno-f-cruz]

Current planned bonsai ui.

[bruno-f-cruz]

Improved in #34 Looks like:

[alexpiet]

personal opinion - double axis plots are a bad idea. Why not have two rows?

[bruno-f-cruz]

Sure, can you check with everyone else if that is what they prefer so we dont need to go back and forth?

[ellahiltonvano]

@bruno-f-cruz could you attach the legend for the graph above? Something that might be helpful to have a little clearer is p_L and p_R

[bruno-f-cruz]

I turned off the legend by default to save space (but can re-enable by default if needed). Otherwise, people can right-click:

Unfortunately, the library i am using doesnt seem to use the symbols in the legend, so some of them will be repeated.

Using the legend, you can also enable/disable series (in case you want to make something clearer online)

Something that might be helpful to have a little clearer is p_L and p_R

By clearer, you mean less transparent? I increased it to 0.5 in all the plots above. Is this what you had in mind?

[ellahiltonvano]

@bruno-f-cruz I think what is unclear to me is what p_R and p_L might look like when the probabilities are not alternating so consistently, as shown above, For example, when they are decoupled, it seems like this way of visualizing might be harder for users to understand than the current plot. Is is possible for you to show what p_R and p_L would look like in a session where the probabilities fluctuate as shown here?

[bruno-f-cruz]

Ignore the issue where no color exists. But something like this:

[ellahiltonvano]

@bruno-f-cruz thanks! This looks good to me, I will circulate to other RAs as well.

[bruno-f-cruz]

@ellahiltonvano, adding the video I sent you for feedback to this thread for future reference

output.2.mp4

[ellahiltonvano]

Feedback from Margaret:

• Thinks its helpful to have p_R and p_L on the same axis, the direct comparison makes it easier to see what the optimal strategy is. How it is plotted above makes this difficult to visualize
• Shading makes it look messy
• Need to test under red light to make sure we can see the colors that are plotted
• Wants difference between unrewarded choices and rewarded choices to be more apparent
• Wants to be able to see the licks over the entire session, to get a sense of early licking, double dipping, etc.
• Ignored trials are hard to see with all the other lines

[bruno-f-cruz]

Wants to be able to see the licks over the entire session, to get a sense of early licking, double dipping, etc.

I can plot this, but it will look super busy (like the example from the current GUI) since you would be cramming data that needs more resolution than pixels you have on the screen. The current plot I showed in the video can go back N seconds, where N can be any number you want. So in theory what you are asking is already implemented, but I don't think you will be able to see it that clearly :P

Otherwise, colors I can change, the only thing that we need to decide is whether to plot lines in the same scale or not. I think that adding more lines to the plot will make it super busy, so if we go in that direction, we may want to do what Alex suggested and split the plot into 2.

@ellahiltonvano Are there any more people that would like to give feedback on this?