Skip to main content

probability or statistics - Efficient maximisation of log-likelihood


I would like to efficiently find the maximum (σ,λ) for the log-likelihood of the derived distribution below. I only need σ and λ to one decimal place - so not very precise. I have tried lowering both the AccuracyGoal and PrecisionGoal to as low as 1, or 2, but this doesn't appear to affect the rate at which the solution is obtained.


The below code creates the function, the some test data, then attempts its maximisation:


aDist[σ_, λ_] := TruncatedDistribution[{0, ∞}, 

  MixtureDistribution[{1, 1}, {NormalDistribution[0, σ], ExponentialDistribution[λ]}]];

data = If[# > 0, #, 0] & /@ RandomVariate[aDist[4, 1/3], {20}];

NMaximize[{LogLikelihood[aDist[σ, λ], data], 10 > σ > 0, 1 > λ > 0}, {σ, λ}]

Whilst for 20 data points this method returns a solution in a short time, for my actual dataset I need to find a solution for a dataset of size 1000+. As it currently stands, this would be untenable for the above method.


I have tried some of the different methods available to NMaximize, including "RandomSearch", and "DifferentialEvolution". However, the method I choose does not seem to make the maximisation run faster.


I have also tried FindMaximum where I start the solver close to the actual parameter values, however this calculation appears to hang forever. I have also tried differentiating, then using NSolve, and FindRoot, but again I am not having any success. Finally I tried a variant on the EM algorithm here, where I alternate between maximising over σ then λ, but again this doesn't help.


Does anyone have any ideas here? I know I could use MCMC if I turn the problem into a Bayesian one, but for reasons I won't mention here I don't want to do this.



As an aside, I can see why it is difficult to find a reasonable maximum to this function, since there is a high degree of correlation between the parameters of the distribution. However, I can't help but think there's a solution of which I'm neglecting to think.




Comments

Post a Comment

Popular posts from this blog

functions - Get leading series expansion term?

Given a function f[x] , I would like to have a function leadingSeries that returns just the leading term in the series around x=0 . For example: leadingSeries[(1/x + 2)/(4 + 1/x^2 + x)] x and leadingSeries[(1/x + 2 + (1 - 1/x^3)/4)/(4 + x)] -(1/(16 x^3)) Is there such a function in Mathematica? Or maybe one can implement it efficiently? EDIT I finally went with the following implementation, based on Carl Woll 's answer: lds[ex_,x_]:=( (ex/.x->(x+O[x]^2))/.SeriesData[U_,Z_,L_List,Mi_,Ma_,De_]:>SeriesData[U,Z,{L[[1]]},Mi,Mi+1,De]//Quiet//Normal) The advantage is, that this one also properly works with functions whose leading term is a constant: lds[Exp[x],x] 1 Answer Update 1 Updated to eliminate SeriesData and to not return additional terms Perhaps you could use: leadingSeries[expr_, x_] := Normal[expr /. x->(x+O[x]^2) /. a_List :> Take[a, 1]] Then for your examples: leadingSeries[(1/x + 2)/(4 + 1/x^2 + x), x] leadingSeries[Exp[x], x] leadingSeries[(1/x + 2 + (1 - 1/x...
Post a Comment
Read more

mathematical optimization - Minimizing using indices, error: Part::pkspec1: The expression cannot be used as a part specification

I want to use Minimize where the variables to minimize are indices pointing into an array. Here a MWE that hopefully shows what my problem is. vars = u@# & /@ Range[3]; cons = Flatten@ { Table[(u[j] != #) & /@ vars[[j + 1 ;; -1]], {j, 1, 3 - 1}], 1 vec1 = {1, 2, 3}; vec2 = {1, 2, 3}; Minimize[{Total@((vec1[[#]] - vec2[[u[#]]])^2 & /@ Range[1, 3]), cons}, vars, Integers] The error I get: Part::pkspec1: The expression u[1] cannot be used as a part specification. >> Answer Ok, it seems that one can get around Mathematica trying to evaluate vec2[[u[1]]] too early by using the function Indexed[vec2,u[1]] . The working MWE would then look like the following: vars = u@# & /@ Range[3]; cons = Flatten@{ Table[(u[j] != #) & /@ vars[[j + 1 ;; -1]], {j, 1, 3 - 1}], 1 vec1 = {1, 2, 3}; vec2 = {1, 2, 3}; NMinimize[ {Total@((vec1[[#]] - Indexed[vec2, u[#]])^2 & /@ R...
Post a Comment
Read more

How to remap graph properties?

Graph objects support both custom properties, which do not have special meanings, and standard properties, which may be used by some functions. When importing from formats such as GraphML, we usually get a result with custom properties. What is the simplest way to remap one property to another, e.g. to remap a custom property to a standard one so it can be used with various functions? Example: Let's get Zachary's karate club network with edge weights and vertex names from here: http://nexus.igraph.org/api/dataset_info?id=1&format=html g = Import[ "http://nexus.igraph.org/api/dataset?id=1&format=GraphML", {"ZIP", "karate.GraphML"}] I can remap "name" to VertexLabels and "weights" to EdgeWeight like this: sp[prop_][g_] := SetProperty[g, prop] g2 = g // sp[EdgeWeight -> (PropertyValue[{g, #}, "weight"] & /@ EdgeList[g])] // sp[VertexLabels -> (# -> PropertyValue[{g, #}, "name"]...
Post a Comment
Read more