unsum: Generate unsummed adjacency matrix
In dougmet/unsumnet: Solve Row and Column Sum Network Constraint Problems

Description Usage Arguments Value Author(s)

Given the row and column sums for a positive weighted adjacency matrix, this function will generate candidate matrices that are sampled uniformly from the space of possible matrices. This is done using simulated annealing.

unsum(constraints, nEdges = NULL, verbose = FALSE, maxEdges = FALSE,
  noReturn = FALSE, mctSchedule = 100, hotTime = 2000, beta0 = 0.01,
  betaMax = 1e+05, mu0 = 0.1, coolingRate = 1.02, maxTime = 1e+06,
  minError = 0.001)

`constraints`	A matrix or data frame containing the row and column sums. NA or negative values are interpretted as unconstrained.
`nEdges`	Integer The desired number of edges (non-zero elements) the adjacency matrix should have.
`verbose`	Logical if set to TRUE then maximum output will be written to screen. Default is FALSE.
`maxEdges`	Logical if set to TRUE then all possible edges will be used. The number of possible edges is given by `calc_max_edges`. Default is FALSE.
`noReturn`	If set to TRUE then only one directed edge can go between two nodes. This would be appropriate if only modelling netted positions for example. Default is FALSE.
`mctSchedule`	Sets the number of MC loops between measurements. Also effects the quench speed as beta is only changed every mctSchedule loops.
`hotTime`	How many MC loops to run at the high temperature before starting the quench. The bigger the number the more time the network has to truly randomise before quenching.
`beta0`	Starting inverse temperature. The smaller the number the hotter, or more random, the starting configuration is.
`betaMax`	The maximum beta value.
`mu0`	The field that couples to the number of edges. The bigger this number the more strict the edge restriction is during the quench.
`coolingRate`	A number greater than, but close to, 1 that multiplies by `beta` every `mctSchedule` loops.
`maxTime`	The maximum number of loops before giving up on that quench and starting again.
`minError`	When the mean error drops below this threshold it is decided that it is close enough to a solution. The remaining error is resolved using the row/col iterator. A bigger number will give better performance but risks skewing the distribution of networks.

an unsumnet object, which is a list containing:

AW: The weighted adjacency matrix that satisfies the row/col sum constraints
A: The (unweighted) adjacency matrix containing network structure.
W: The final weights matrix (for on and off edges) before iterating to the final solution
Results: Counts for the number of outcomes from dense_hybrid.
targetEdges Number of edges requested
nEdges Number of edges obtained