PDMP Types
(/!\WIP/!\)
PDMP.FactorGraph
— Type.FactorGraph
Encapsulation of a factor graph. It is made out of a structure (FactorGraphStruct) and an array of factors corresponding to the given structure. See also: Factor
, FactorGraphStruct
.
Int
forInt64
– inPDMP.jl
Float
forFloat64
– inPDMP.jl
AllowedVarType
forUnion{Float, Vector{Float}}
(variable types in the local BPS) – inlocal/event.jl
AllowedTimeType
forUnion{Vector{Float}, LinSpace{Float}}
(format of collections of times that can be passed to extract samples from events) – inpath.jl
Abstract types
MvGaussian
for multivariate gaussians – inmodels/mvgaussian.jl
Domain
for domain description (in global BPS) – ingeometry.jl
IPPSamplingMethod
for sampling methods of an IPP – inippsampler.jl
Thinning <: IPPSamplingMethod
for thinning methods of an IPP – inippsampler.jl
Specific types (global)
geometry
Unconstrained <: Domain
(immutable) defines a domain without boundary (signature only)Polygonal <: Domain
(immutable) defines a domain with affine boundaries determined bynormals
andintercepts
ippsampler
LinearBound <: Thinning
(immutable) thinning method using a linear bound following a uniform bound on the eigenvalues of the HessianNextEvent
object returned when sampling from the IPP (contains a bouncing time, whether to bounce or not and a flipindex (ZZ case))
path
Path
container for a path of the global sampling (stores list of corners and times)Segment
(immutable) container to store the two corners of a linear segment, makes it easier to sample from aPath
simulate
Simulation
container for the parameters of a global simulation
Model types (global)
LogReg
(immutable) model for a logistic regression
Specific types (local)
event
Event{T<:AllowedVarType}
(immutable) a triple(x,v,t)
corresponding to an event for one of the node.EventList{T<:AllowedVarType}
list of events (stored as lists ofxs
,vs
,ts
in order to be traversed more effectively than aVector{Event}
).AllEventList
container for theEventList
of all the nodes
factorgraph
Factor
(immutable) attaches thenextevent
function (sampling from IPP), the gradient of the corresponding log-likelihood (gll
) and an indexFactorGraphStruct
(immutable) contains the pattern of connections between nodes viaflist
andvlist
(storing what variable is attached to what factor and vice-versa)nfactors
andnvars
keep track of the number of factors and the number of nodes (variables)FactorGraph
(immutable) container for all theFactor
and theFactorGraphStruct
simulate
LocalSimulation
(immutable) container for the parameters of a local simulation (the factor graph, initial positions, etc)