missing algebra gradients

[CarloLucibello]

• diag, diagm
• triu, tril
• kron
• dot
• inv
• trace
• det (see Add determinant function on variable; Add backward on svd pytorch/pytorch#3816)
• logdet, logabsdet
• chol (see Cholesky backpropagation tensorflow/tensorflow#1465 and the paper mentioned over there)
• svd (see Add determinant function on variable; Add backward on svd pytorch/pytorch#3816 and https://j-towns.github.io/papers/svd-derivative.pdf )
• qr (see Feature Request: Gradient for QR op tensorflow/tensorflow#6504 (comment))
• lq
• eig
• eigs
• \, /

useful reading https://arxiv.org/pdf/1710.08717.pdf

[denizyuret]

All done except last 3?

[CarloLucibello]

All done except last 3?

right

[denizyuret]

Some of these broke because of 0.7. I put @test_broken in test/linearalgebra.jl