itensor / website Goto Github PK

Here are some updates needed in the "Book" (http://www.itensor.org/docs.cgi?vers=cppv3&page=book):

Index Objects section (http://www.itensor.org/docs.cgi?vers=cppv3&page=book/index):

• noprime -> noPrime

ITensor Basics section (http://www.itensor.org/docs.cgi?vers=cppv3&page=book/itensor_basics):

• T.ord(), ord(T), rank(T), T.r() -> maybe don't mention these alternatives? We could just mention that "order" is called different things in different communities, but I think it is better not to mention them as supported anymore, I would rather undocument them so they are not floating around forever.
• Maybe change T.elt(...) -> elt(T,...) if we decide on that notation.

Factorizing ITensors (SVD Example) section (http://www.itensor.org/docs.cgi?vers=cppv3&page=book/itensor_factorizing):

• "Maxm","Minm" -> "MaxDim", "MinDim"
• Mention setting custom tags on new Index? I think we should encourage users to do this, but maybe it is too much too early on.

Case Study: TRG Algorithm section (http://www.itensor.org/docs.cgi?vers=cppv3&page=book/trg):

• Matt: I plan to clean up the TRG code, and then update this page accordingly. For example, I will use the Indices l,r,u,d with tags "left","right","up","down", and then distinguish indices at different levels just by the tags "orig" and "renorm", instead of e.g. "scale=4", "scale=5".
• Don't name the delta tensors for doing the trace Trx, Tr, maybe just multiply with T*delta(l,r)*delta(u,d) is clear enough.
• We should keep track of kappa, the "partition function per site" defined by Z = kappa^N. The current way of keeping track of lnZ overflows with more iterations. Also we can mention the exact free energy, etc.

IQTensor sections:

• Update to make IQTensor -> ITensor with QDense storage.

SiteSet docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/siteset) need to be updated for V3:

• Should sites.op("Sz",3) be made a free function op(sites,"Sz",3)?
• sites.N() -> length(sites) (should this be named nsites(sites)?

Also, for specific SiteSet implementation docs (SpinHalf, SpinOne, etc.):

• .real -> .elt
• Maybe sites.op(...) -> op(sites,...)?
• For each, list that the default Arg is {"ConsverveQNs=",true}

Should the exact diagonalization code formula (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/simple_ed) show the use of davidson instead of a power method for finding the ground state? It should be more efficient.

Maybe it could be mentioned as an alternative, since the power method is better for teaching purposes.

We should add the LinQu package: https://github.com/YiqingZhouKelly/LinQu.jl
to the codes page once we begin documenting ITensors.jl on the website.

This is the quantum simulation package developed by Yiqing Zhou during her summer CCQ internship.

It seems like we can merge "The ITensor Book", "Tutorials", and "Code Formulas" together (maybe all into "The ITensor Book"?).

They all have useful information, but it is too spread out right now. I often spend time looking for a tutorial/sample code I know is there but forgot which link it was put under.

As described in this support answer: http://www.itensor.org/support/866/initialize-the-random-seed-of-mps-sites

@emstoudenmire, could you have the homepage point to v3.1.1? It has an important bugfix for the removeQNs(ITensor) function.

It also has a few other improvements, like that svd now doesn't truncate by default (which it was doing accidentally before) and Cutoff=0 now truncates singular values that are exactly zero.

Thanks,
Matt

Tensor Decomposition docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/decomp) need some updates:

• randomTensor -> randomITensor everywhere in examples
• Change svd args:
  • Change "Maxm"/"Minm" (maybe to "MaxDim"/"MinDim" or "MaxRank"/"MinRank"? This isn't implemented yet so maybe don't change the docs)
  • Need to remove "LeftIndexName", "LeftIndexType", etc. to instead be "LeftTags", etc.
  • List the default tag convention, and add example of getting the Index with commonIndex or findIndex
• Update factor for tags
• Update diagHermitian:
  • "Maxm"/"Minm" -> something else (Matt: this isn't implemented yet so maybe don't change the docs).
  • Update to reflect tag args and tag convention
  • Show how more general input index structure can be used, like ia,ja,...,ib,jb,... where replaceTags(ia,tags(a),tags(b)) == ib? (Matt: I am not sure if I added this feature yet, so maybe not).
• Update denmatDecomp docs
  • "Maxm"/"Minm" -> something else (Matt: this isn't implemented yet so maybe don't change the docs).
  • Show the tags convention for the new index, and how it can be changed.

We should add the "C++v2 | C++v3" links at the top of the "Codes" page (http://www.itensor.org/codes.html) and only list ones that work for each version.

Update Index docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/index):

• Add i=2 as a type of constructor for IndexVals
• Show free function versions of .write, .read, .dag, .dir (and make those versions if they don't exist)

The SiteSet documentation page (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/siteset) says you can do SiteSet(N) to get a generic SiteSet with N sites, but I don't think it works. Maybe it should default to SiteSet(2,N) SiteSet(N,2)?

The script should write html files to the same directory as the md files. Each time it's run, the script checks whether the md file is newer than the html file - if so, it replaces the html file.

Here are updates needed to the tutorials (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials) for V3:

ITensor Quickstart (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/quickstart)

• Fix Index constructors
• i(1) -> i=1

Understanding Tensor Diagrams (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/diagrams)

• Fix Index constructors
• IQMPS/IQTensor -> MPS/ITensor with QNs/QDense storage

Priming Indices in ITensor (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/primes)

• Remove IndexType, explain using tags

Choosing DMRG Parameters (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/dmrg_params)

• maxm -> maxDim, etc.
• niter -> maxKrylovSize?

Matrix Product States (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/MPS)

• psi.A(3) -> psi(3), psi.Aref(3) -> psi.ref(3), etc.
• In prime function, use "Site" instead of Site

Matrix Product Operator (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/MPO)

• MPO -> toMPO?
• Site -> "Site"

Introduction to AutoMPO (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/AutoMPO)

• MPO -> toMPO?

Calculating Two-Site Correlation Functions (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/correlations)

• psi.A(2) -> psi(2), etc.
• Link -> "Link", etc.
• C.real() -> C.elt(). Maybe to get a scalar it should be elt(C)?

The Args System (http://www.itensor.org/docs.cgi?vers=cppv3&page=tutorials/args)

• "Maxm" -> "MaxDim"

The link to Tutorials is missing on the C++v3 page. It appears only on C++v2.

Sweeps docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/sweeps):

• maxm, minm options -> maxdim, mindim
• niter -> something else (Matt: we should add restarts, and then have two arguments, like MaxKrylovSize and MaxIter, but maybe that should wait until post V3?)

DMRG docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/dmrg):

• IQMPS/IQMPO -> MPS/MPO
• "WriteM" -> "WriteDim"

IndexVal docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/indexval) need to be updated to show free functions. Also, I need to check that the free functions are all valid.

Some updates are needed for the AutoMPO docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/autompo) for V3:

• IQMPO/MPO -> toMPO/removeQNs
• "MaxM" -> "MaxDim" in toMPO args
• toExpH<ITensor> -> toExpH

Here are some things we should make sure to emphasize in the transition guide/changelog:

• Changed the prime level convention for the output of nmultMPO and applyMPO.
• Removed exactApplyMPO and fitApplyMPO interface in favor of applyMPO.
• Instead of mapprime, use replaceTags (maybe we will add mapPrime back to ease the transition).
• ITensor functions assume MPS and MPO have certain index topology conventions (neighboring tensors have one shared link).

How come the iDMRG tutorial (http://itensor.org/docs.cgi?page=tutorials/iDMRG) is not referenced on the actual website? Is it obsolete or incomplete or something? Currently, iDMRG has no documentation at all on the website.

We want to direct people with questions about the library to the appropriate resources. These could include a mailing list system and discussion forums on the website.

The ITensor docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/itensor) need some more updates for V3:

• rank -> order.
• Should .elt and .eltC be free functions, i.e. elt(T,i=1,j=2)? (Yes)
• T.inds() -> inds(T).
• Add the tags(T, "... -> ...") function, once it is created. Decided not to implement this.
• Instead of referencing IndexSet docs for *Tags functions, just put the description directly on the ITensor page (these functions will become much simpler once prime levels can be accessed through tags). The interface for tagging functions is the same, so the docs just point to IndexSet.
• Put .conj, .dag, etc next to conj, dag, etc.
• Put the section "Analyzing ITensor Indices" closer to the top, since these are used a lot.

Add TimeEvoMPS.jl to the codes list. Even though it hasn't been updated in a while, people might still find it to be useful.

Code Formulas (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas) updates for V3:

NOTE:

• Extract the Storage of a Dense ITensor
• Evaluate a Function on Each Element of a Tensor

are linked twice, update for V2 docs as well.

• Evaluate a Function on Each Element of a Tensor (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/visit):

• randomize(T) -> T.randomize()

• Extract the Storage of a Dense ITensor (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas):

• Update Index constructors

• randomize(T) -> T.randomize()

• Create a set of Site indices to use as a lattice (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/index_sites):

• nameint("site_",7) -> format("site_%d",7) (nameint was removed since there is no name anymore)

• Change Index constructors

• Change IndexVal constructors from e.g. i(1) -> i=1

• Make a single-site operator (no quantum numbers) (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/itensor_single_site_op):

• Change Index constructors

• s(1) -> s=1

• Exactly Compute the Ground State of a Small Hamiltonian (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/simple_ed)

• Code is not found, make sure it is updated for V3

• Replace an ITensor Index with Another Index (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/itensor_replace_index):

• It would be good to show x = sim(b), I think we should advertise it's use more as a way to make temporary indices as an alternative to tags and prime levels. Also, we should add a constructor x = sim(b,"...") as a way to assign new tags and prime levels to the similar index.

• Is the description of the delta correct here? I thought it didn't simply replace the Index of the ITensor, but actually did a sparse diagonal multiplication. Does it act differently for delta with more than two indices?

• Perform a basic DMRG calculation (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/basic_dmrg):

• MPO -> toMPO

• maxm -> maxDim

• Add {"ConserveQNs=",false} arg

• Make a Ladder Hamiltonian for DMRG (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/ladder):

• MPO -> toMPO

• Add {"ConserveQNs=",false} arg

• maxm -> maxDim

• Make a 2D Hamiltonian for DMRG (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas):

• IQMPO -> toMPO

• IQMPS -> MPS

• maxm -> maxDim

• Compute excited states using DMRG (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/excited_dmrg)

• Add {"ConserveQNs=",false} arg

• maxm -> maxDim

• MPO -> toMPO

Measure local properties of an MPS wavefunction (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/measure_mps):

• Change psi.A(j) -> psi(j)
• Site -> "Site"
• T.real() -> elt(T)
• sites.op(...) -> op(sites,...)

Measure two-point correlator from an MPS wavefunction (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/correlator_mps):

• IQMPS -> MPS
• sites.op("Op1",i) - > op(sites,"Op1",i), etc.
• Link -> "Link", etc.
• psi.A(i) -> psi(i), etc.
• commonIndex(psi.A(i),psi.A(i+1),Link) -> linkInd(psi,i)
• psi.A(i)*op_i*dag(prime(psi.A(i),Site,ir)) -> prime(psi(i),linkInd(psi,i-1))*op_i*dag(prime(psi(i)))
• C *= psi.A(k); -> auto sk = siteInd(psi,k); C *= prime(psi(k),sk);
• C *= dag(prime(psi.A(k),Link)); -> C *= dag(prime(psi(k)));
• C *= psi.A(j); -> auto lj = linkInd(psi,j); C *= prime(psi(j),lj);
• C *= dag(prime(psi.A(j),jl,Site)); -> C *= dag(prime(psi(j)));
• C.real() -> elt(C)

Compute entanglement entropy (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/entanglement_mps):

• psi.A(b) -> psi(b)
• commonIndex(psi.A(i),psi.A(i+1),Link) -> linkInd(psi,i)

Measure spinless fermion two-point correlator (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/spinless_correlator_mps):

• Changes similar to "Measure two-point correlator from an MPS wavefunction" page

Compute a two-site reduced density matrix from an MPS (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/mps_two_rdm):

• Changes similar to "Measure two-point correlator from an MPS wavefunction" page

Operator matrix elements involving two MPS (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/two_mps):

• Changes similar to "Measure two-point correlator from an MPS wavefunction" page

Time-evolving an MPS with Trotter Gates (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/tevol_trotter):

• BondGate<ITensor> -> BondGate
• sites.op(...) -> op(sites,...)?

Time-evolving an MPS with an MPO (matrix product operator) (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/tevol_mps_mpo):

• psi.A(3) -> psi(3)
• noprime -> noPrime
• toExpH<ITensor> -> toExpH
• exactApplyMPO(...) -> applyMPO(...,{"Method=","DensityMatrix"}), fitApplyMPO(...) -> ..., ect.
• normalize(psi) -> psi.normalize()

Applying a two-site 'gate' to an MPS (http://www.itensor.org/docs.cgi?vers=cppv3&page=formulas/gate):

• psi.A(3) -> psi(3)
• noprime -> noPrime
• psi.setA(3,U) -> psi.set(3,U)

It would be good to add a page somewhere with a list of performance tips for ITensor/tensor network contraction methods.

One example is a performance issue I was just running into, where I was contracting two MPS psi and phi like this:

auto N = 10;
auto sites = SiteSet(N,2);
auto psi = MPS(sites);
auto phi = MPS(sites);
auto O = psi(1)*phi(1);
for( auto n : range1(N) )
   O *= psi(n) * phi(n);

and getting really bad performance (for medium to large bond dimensions). The definition of *= in ITensor leads to a different order of operations than I was expecting (performing psi(n)*phi(n) first, then contracting with O). So a better way is to write this as:

for( auto n : range1(N) )
  {
  O *= psi(n);
  O *= phi(n);
  }

I was used to the definition of x *= y from Julia, where it is always lowered to x = x * y, so it took longer than it should have for me to catch this. Anyway, just adding a note about this for future reference.

Right now, the ITensor prime and tag functions section points to the IndexSet documentation page. It would be better to have those functions directly in the ITensor page to avoid confusion and make finding them easier when searching the website.

Right now, http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/index only mentions the i(1) type of constructor.

Update MPS docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/mps) for V3:

• psi.A(3) -> psi(3) (Matt: this still needs to be implemented)
• psi.setA(3,A) -> psi.set(3,A) (Matt: this still needs to be implemented)
• psi.Aref(3) -> psi.ref(3) (Matt: this still needs to be implemented)
• Remove .Anc from docs
• psi.N() -> length(psi)
• .rightLim(), .leftLim(), .isOrtho(), .orthoCenter(), .sites() -> free functions (Matt: this still needs to be implements)
• Update "MPS Prime Level Methods"
• averageM, maxM -> averageLinkDim, maxLinkDim
• normalize(psi) -> .normalize(psi)
• psi.norm() -> norm(psi)

MPO docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/mpo):

• H.A(j) -> H(j)

Some of the IndexSet docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/indexset) need to be updated more for V3:

• .r and rank -> order
• Update findIndex functions to only accept a TagSet for matching once prime levels can be accessed through the tagset.
• maxM and minM -> maxDim and minDim
• Move "Other IndexSet Functions" to the top since I think those are ones people would use most. The other functions are mostly internal ones that get used when they are called on an ITensor.
• Make the "Accessor Methods" free functions?
• Rename extent -> dim?

In "Single Element ITensor" docs, use i=2 instead of i(2): http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/single_itensor

Some updates to combiner docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/combiner) for V3:

• rank -> order
• C.m() -> dim(C)
• Remove IndexName arg option.
• Add example code of getting combined index with auto c = findIndex(C,"CMB")

Check to make sure diagTensor docs (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/diag_itensor) are complete. Can the constructors take other types of inputs, like IndexSet, std::vector<Index>, etc. and not just Index,Index,...?

This is something user benoit mentioned already on the website under "AutoMPO Syntax: Concatenate local terms" but miles commented that it would be best to file a feature request. I'm wondering if it would be possible to have a function that would allow you to construct an operator involving operators from an arbitrary number of sites, without explicitly needing to write
ampo += "Sz", j, "Sz", j+1, "S+", j+2, ...
I have a system where the number of operators increases with the system size, so for example for a system of size 7 I will have an operator involving all 7 sites (they will have small coefficients out front compared to more local operators so that DMRG should hopefully still be useful). See benoit's post for an example of the pseudocode that would be useful to have. Thanks!

I have no access to the documentation through the website.

Docs for nmultMPO, applyMPO, etc. (http://www.itensor.org/docs.cgi?vers=cppv3&page=classes/mps_mpo_algs) need some updating:

• Need to change args "Maxm", "Minm" -> "MaxDim", "MinDim" (Matt: this needs to be implemented in the code as well)
• Tags args to set tags of output MPS in applyMPO? (Matt: I don't think this is implemented yet so maybe not).