Opened 3 years ago

Last modified 3 years ago

#28176 closed defect

Fix few bugs in ClusterAlgebra discovered while working on #26771 — at Initial Version

Reported by: Salvatore Stella Owned by:
Priority: major Milestone: sage-8.9
Component: algebra Keywords: ClusterAlgebras
Cc: Travis Scrimshaw, Gregg Musiker, Christian Stump, gh-EBanaian, Dylan Rupel, gh-kelleye Merged in:
Authors: Salvatore Stella Reviewers:
Report Upstream: N/A Work issues:
Branch: Commit:
Dependencies: Stopgaps:

Status badges

Description

As pointed out by the people at #26771, ClusterAlgebra has few bugs.

  • Default arguments do not play nicely with UniqueRepresentation

  sage: A1 = ClusterAlgebra(['A',2])
  sage: A2 = ClusterAlgebra(['A',2], cluster_variable_prefix='x')
  sage: A1 is A2
  False
  • mutate_initial keeps the same variable names; In theory this is not a problem but may create confusion. For example suppose we are looking at the Grassmannian of planes in a 4 dimensional vector space. Then the associate cluster algebra is
  sage: A = ClusterAlgebra( matrix(5,[0,1,-1,1,-1]), cluster_variable_names=['p13'], coefficient_names=['p12','p23','p34','p41'], scalars=QQbar); A
  A Cluster Algebra with cluster variable p13 and coefficients p12, p23, p34, p41 over Algebraic Field

Swapping out p13 from the initial seed should give p24 or, at the very least *not* p13. Getting p24 requires some understanding of what the ring in question really is that goes beyond the scope of this class. Currently we get

  sage: A.mutate_initial(0)
  A Cluster Algebra with cluster variable p13 and coefficients p12, p23, p34, p41 over Algebraic Field
  • mutate_initial does not compute all the required F-polynomials and it does not enforce the fact that they are not rational expressions
  sage: A1 = ClusterAlgebra(['A',[2,1],1])
  sage: A2 = A1.mutate_initial([0,1,0])
  sage: len(A2.g_vectors_so_far()) == len(A2.F_polynomials_so_far())
  False
  sage: all(parent(f) == A2._U for f in A2.F_polynomials_so_far())
  False
  • finally mutate_initial does not enforce the fact that initial g-vectors belong to the initial cluster but rather prepend the performed mutation sequence to all the paths already known.
  sage: A2.find_g_vector((0,0,1))
  [0, 1, 0]

This, while technically not false, may generate confusion.

This ticket deals with all these issues.

Change History (0)

Note: See TracTickets for help on using tickets.