Strange Error-Nan Iteration Results and RuntimeWarning Using t-SNE

Question

I am using the python t-SNE implementation to reduce the dimension by X, which contains 100 instances, each of which is described by 1024 parameters for cnn rendering.

X.shape = [100, 1024]

X.dtype = float32

When I run:

Y = tsne.tsne(X)

The first warning appears in tsne.py, line 23:

RuntimeWarning: division by zero, found in the log H = Math.log (sumP) + beta * Math.sum (D * P) / sumP

Then there are a few more warnings like this in the following lines:

RuntimeWarning: Invalid value found during division

And finally, I get this result after each iteration during processing:

Iterate xyz: error - nan

"", .

EDIT:

- > , . (, , , )

:

- , ? ?

Answer 1

sumP = sum(P)+np.finfo(np.double).eps
H = np.log(sumP) + beta * np.sum(D * P) / sumP;

Answer 2

, tsne tsne Python Matlab, "H = np.log(sumP) + beta * np.sum(D * P)/sumP; " - ?