Visualization of volumes of statistical distributions

function learned by neural network

Explanation of what this is

Note, in order to run this, you need a copy of MFC71.dll and MRC71R.dll.

You can download these files here (for mfc71.dll) and here (for msvcr71.dll). You may also need msvcp71.dll I think you need to put them in the same directory as the exe, but if you are sufficiently familiar with windows to do the "registering dlls" dance, then that might work too.

The application comes in two parts. Before you download the application, download the convertor. It only works if you put your source data in the same format as your test data in "C:\\ricki_data.txt". I apologize if you do not have a C:\ on your computer. Afterwards it will create a file called "C:\ricki_data.dat" Once that file is created, you can run the application , which you can download here.



If you are interested in this visualization, you might also want to see Training of an obstacle avoiding robot in simulation

Methodology

Real quick here, and I'll put more later. I treat the percentile distributions as a piecewise constant probability distributions (i.e. 10% of the distribution is uniformly distributed between the first entry in the list and the second).

For each voxel in space, I transform the voxel to an RGBA color value by taking the integral of the distribution mentioned above with the transfer function for Red, Green, Blue or Alpha. These functions are the splines shown on the right of the screenshot. The red spline is for the red fucntion, the blue for the blue, green for green, and the white spline is for the alpha (or opacity) function.

IMPORTANT NOTEThe scale on these functions is as follows. The horizontal scale is between zero and one-hundred celcius. The vertical scale is such that a constant function at max value integrates to one.

Sketching an impulse function at point p will give the propability that a given voxel has temperature p. Doing this for the red function will make the redness of the resulting color convey this. Doing this for the alpha function will make the opacity of the resulting voxel convey this. Voxels are drawn in back-to-front order such that an opaque voxel in front hides the voxels behind it.


mds at es oh ee dot you see es see dot eee dee you