Deep Learning: Difference between revisions
| Line 1,454: | Line 1,454: | ||
In Glow [Kingmen et al], they use invertible 1x1 convolution which is equivalent to matrix multiplication. | In Glow [Kingmen et al], they use invertible 1x1 convolution which is equivalent to matrix multiplication. | ||
Use <math>W = P L (U+diag(s))</math> and <math>| det(W)| = sum(\log |S|)</math> | Use <math>W = P L (U+diag(s))</math> and <math>| det(W)| = sum(\log |S|)</math> | ||
===Dequantization=== | |||
Fitting a ''density'' function to discrete data can have crazy peaks. | |||
For dequantization, add a uniform to the data to get a more stable density function. | |||
==Misc== | ==Misc== | ||