# W2D1: Tutorial 1 - Demo widget thoughts from exploration

Hi!

When I play with this demo, I have some thoughts I may want to discuss with you.

I suppose that the location of a visual target is clear and close, so people have a small standard deviation (1) in the visual distribution. And participants have good auditory ability to distinguish and identify the location of the audio source, thus I use a small standard deviation(1) in auditory likelihood distribution as well.

If the two distributions are very far apart: the visual target(+4) is probably not the source of audio, and people hear that the sound comes from the opposite direction (-4). The real location of the sound source lies between these two distributions, when both of their probabilities are low.
If the two distributions are very close to each other: the visual target(+1) is nearly the right place of sending the sound, and the audio judgment(-0.5) is almost correct, too. The real location distribution is at the overlapping area of these two.

The first phenomenon is interestingâ€¦
The second phenomenon shows that the real location is probably a reinforcement result of the two observations. Visual and audio perceptions could put together to make a more accurate judgment, and people could make a better prediction with two of these inputs - but the brainâ€™s judgment â€śauditory distributionâ€ť is within a certain confidence level, anyway. This shows how the brain works. BTW, I would expect the posterior integration to be 1, which should seem to be equivalent to other curves in area.

The widget uses your code for my_gaussian; are you sure everythingâ€™s okay there?

The peak of a Gaussian with sigma=1 on a grid with dx = 0.1 should be ~0.04 units, rather than 0.4.

When I run it on the instructor version with the same parameters, I get this instead:

1 Like

check if you normalised the posterior.

2 Likes

In the normalization, I didnâ€™t use the method in the solution(which I donâ€™t remember). I did an integration of the probability data * time_step_interval (which is 0.01 something).
There might be a problem. Thanks for reminding that the widget depends on my codeâ€¦