svd() [3/3]

template<Rank rank>

std::tuple<SRDef, Tensor<rank>, Tensor<2>, Tensor<2> > syten::SVD::svd	(	Tensor< rank > const &	in,
		Index	uncombinedLeg,
		Truncation const &	trunc = Truncation(),
		DecompRescaleCGC const	rescale = DecompRescaleCGC::Yes,
		GaugeFixSVD	gf = GaugeFixSVD::n(),
		SourceLocation	location = SourceLocation::current()
	)

A nicer interface to a named left-normalising SVD.

Use as follows:

auto [err, u, s, v] = svd(in, uncombLeg, trunc);

err will contain the sum of squared singular values, u, s and v will contain the tensors returned by the SVD, where u has the same rank as in and s and v are rank-2 tensors connecting to the uncombLeg of u.

Remarks

With the syntax above, you do not have to declare err, u, s and v beforehand.

References std::make_tuple(), std::move(), and svd().

Here is the call graph for this function: