mirror of
https://github.com/kalmarek/PropertyT.jl.git
synced 2024-11-13 22:05:27 +01:00
180 lines
4.5 KiB
Julia
180 lines
4.5 KiB
Julia
using JLD
|
||
using JuMP
|
||
import Primes: isprime
|
||
import SCS: SCSSolver
|
||
import Mosek: MosekSolver
|
||
|
||
using Mods
|
||
|
||
using Groups
|
||
using ProgressMeter
|
||
|
||
|
||
function SL_generatingset(n::Int)
|
||
|
||
indexing = [(i,j) for i in 1:n for j in 1:n if i≠j]
|
||
|
||
S = [E(i,j,N=n) for (i,j) in indexing];
|
||
S = vcat(S, [convert(Array{Int,2},x') for x in S]);
|
||
S = vcat(S, [convert(Array{Int,2},inv(x)) for x in S]);
|
||
return unique(S)
|
||
end
|
||
|
||
function E(i::Int, j::Int; val=1, N::Int=3, mod=Inf)
|
||
@assert i≠j
|
||
m = eye(Int, N)
|
||
m[i,j] = val
|
||
if mod == Inf
|
||
return m
|
||
else
|
||
return [Mod(x,mod) for x in m]
|
||
end
|
||
end
|
||
|
||
function cofactor(i,j,M)
|
||
z1 = ones(Bool,size(M,1))
|
||
z1[i] = false
|
||
|
||
z2 = ones(Bool,size(M,2))
|
||
z2[j] = false
|
||
|
||
return M[z1,z2]
|
||
end
|
||
|
||
import Base.LinAlg.det
|
||
|
||
function det(M::Array{Mod,2})
|
||
if size(M,1) ≠ size(M,2)
|
||
d = Mod(0,M[1,1].mod)
|
||
elseif size(M,1) == 2
|
||
d = M[1,1]*M[2,2] - M[1,2]*M[2,1]
|
||
else
|
||
d = zero(eltype(M))
|
||
for i in 1:size(M,1)
|
||
d += (-1)^(i+1)*M[i,1]*det(cofactor(i,1,M))
|
||
end
|
||
end
|
||
# @show (M, d)
|
||
return d
|
||
end
|
||
|
||
function adjugate(M)
|
||
K = similar(M)
|
||
for i in 1:size(M,1), j in 1:size(M,2)
|
||
K[j,i] = (-1)^(i+j)*det(cofactor(i,j,M))
|
||
end
|
||
return K
|
||
end
|
||
|
||
import Base: inv, one, zero, *
|
||
|
||
one(::Type{Mod}) = 1
|
||
zero(::Type{Mod}) = 0
|
||
zero(x::Mod) = Mod(x.mod)
|
||
|
||
function inv(M::Array{Mod,2})
|
||
d = det(M)
|
||
d ≠ 0*d || thow(ArgumentError("Matrix is not invertible!"))
|
||
return inv(det(M))*adjugate(M)
|
||
return adjugate(M)
|
||
end
|
||
|
||
function prepare_Δ_sdp_constraints(identity, S)
|
||
@show length(S)
|
||
|
||
B₁ = vcat([identity], S)
|
||
B₂ = products(B₁, B₁);
|
||
B₃ = products(B₁, B₂);
|
||
B₄ = products(B₁, B₃);
|
||
@assert B₄[1:length(B₂)] == B₂
|
||
|
||
product_matrix = create_product_matrix(B₄,length(B₂));
|
||
sdp_constraints = constraints_from_pm(product_matrix, length(B₄))
|
||
L_coeff = splaplacian_coeff(S, B₂, length(B₄));
|
||
Δ = GroupAlgebraElement(L_coeff, product_matrix)
|
||
|
||
return Δ, sdp_constraints
|
||
end
|
||
|
||
function load_Δ_sdp_constraints(name::String;cached=true)
|
||
pm_filename = "$name.product_matrix.jld"
|
||
Δ_coeff_filename = "$name.delta.coefficients.jld"
|
||
f₁ = isfile(pm_filename)
|
||
f₂ = isfile(Δ_coeff_filename)
|
||
if cached && f₁ && f₂
|
||
println("Loading precomputed pm, Δ, sdp_constraints...")
|
||
product_matrix = load(pm_filename, "pm")
|
||
L = load(Δ_coeff_filename, "Δ")[:, 1]
|
||
Δ = GroupAlgebraElement(L, Array{Int,2}(product_matrix))
|
||
sdp_constraints = constraints_from_pm(product_matrix)
|
||
else
|
||
println("Computing pm, Δ, sdp_constraints...")
|
||
ID = eye(Int, 3)
|
||
S = SL₃ℤ_generatingset()
|
||
Δ, sdp_constraints = prepare_Δ_sdp_constraints(ID, S)
|
||
|
||
save(pm_filename, "pm", Δ.product_matrix)
|
||
save(Δ_coeff_filename, "Δ", Δ.coefficients)
|
||
|
||
end
|
||
return Δ, sdp_constraints
|
||
end
|
||
|
||
|
||
function compute_κ_A(name::String, Δ, sdp_constraints;
|
||
cached = true,
|
||
tol = 1e-7,
|
||
verbose = false,
|
||
# solver = MosekSolver(INTPNT_CO_TOL_REL_GAP=tol, QUIET=!verbose))
|
||
solver = SCSSolver(eps=tol, max_iters=20000, cg_rate=3, verbose=verbose))
|
||
|
||
f₁ = isfile("$name.kappa")
|
||
f₂ = isfile("$name.SDPmatrixA")
|
||
|
||
if cached && f₁ && f₂
|
||
println("Loading precomputed κ, A...")
|
||
A = readdlm("$name.SDPmatrixA")
|
||
κ = readdlm("$name.kappa")[1]
|
||
else
|
||
println("Solving SDP problem maximizing κ...")
|
||
κ, A = solve_SDP(sdp_constraints, Δ, solver, verbose=verbose)
|
||
# writedlm("$name.kappa", kappa)
|
||
# writedlm("$name.SDPmatrixA", A)
|
||
end
|
||
return κ, A
|
||
end
|
||
|
||
function main()
|
||
const NAME = "SL3Z"
|
||
const VERBOSE = true
|
||
const TOL=1e-7
|
||
const Δ, sdp_constraints = load_Δ_sdp_constraints(NAME)
|
||
const κ, A = compute_κ_A(NAME, Δ, sdp_constraints, cached=false, verbose=VERBOSE)
|
||
|
||
if maximum(A) < 1e-2
|
||
warn("Solver might not solved the problem successfully and the positive solution is due to floating-point error, proceeding anyway...")
|
||
end
|
||
|
||
if κ > 0
|
||
@assert A == Symmetric(A)
|
||
const A_sqrt = real(sqrtm(A))
|
||
|
||
T = ℚ_distance_to_positive_cone(Δ, κ, A, tol=TOL, verbose=VERBOSE)
|
||
|
||
if T < 0
|
||
println("$NAME HAS property (T)!")
|
||
else
|
||
println("$NAME may NOT HAVE property (T)!")
|
||
end
|
||
|
||
else
|
||
println("$κ < 0: $NAME may NOT HAVE property (T)!")
|
||
end
|
||
end
|
||
|
||
@everywhere push!(LOAD_PATH, "./")
|
||
using GroupAlgebras
|
||
include("property(T).jl")
|
||
|
||
main()
|