Update to SpecialFunctions 0.8

Project.toml

@@ -1,7 +1,7 @@
 name = "DPMMSubClusters"
 uuid = "2841fd70-8698-11e9-176d-6dfa142d2ee7"
 authors = ["Or Dinari <[email protected]>"]
-version = "0.1.6"
+version = "0.1.7"
 
 [deps]
 CatViews = "81a5f4ea-a946-549a-aa7e-2a7f63a27d31"
@@ -20,10 +20,10 @@ StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 CatViews = "1"
 Clustering = "0.13.3"
 DistributedArrays = "0, 1"
-Distributions = "0.21.3"
+Distributions = "0, 1"
 JLD2 = "0, 1"
 NPZ = "0, 1"
-SpecialFunctions = "0, 1"
+SpecialFunctions = "0.8,0.9"
 StatsBase = "0,1"
 julia = "1"
 

src/DPMMSubClusters.jl

@@ -33,10 +33,6 @@ include("global_params.jl")
 include("dp-parallel-sampling.jl")
 include("data_generators.jl")
 
-if length(procs()) == 1
-    new_procs = addprocs(1)
-end
-
 
 export generate_gaussian_data, generate_mnmm_data, dp_parallel_sampling, dp_parallel, run_model_from_checkpoint, save_model, calculate_posterior, fit, get_labels_histogram
 

src/dp-parallel-sampling.jl

@@ -444,15 +444,15 @@ end
 
 
 function calculate_posterior(model::dp_parallel_sampling)
-    log_posterior = lgamma(model.model_hyperparams.α) - lgamma(size(model.group.points,2)+model.model_hyperparams.α)
+    log_posterior = logabsgamma(model.model_hyperparams.α)[1] - logabsgamma(size(model.group.points,2)+model.model_hyperparams.α)[1]
     for cluster in model.group.local_clusters
         if cluster.cluster_params.cluster_params.suff_statistics.N == 0
             continue
         end
         log_posterior += log_marginal_likelihood(cluster.cluster_params.cluster_params.hyperparams,
             cluster.cluster_params.cluster_params.posterior_hyperparams,
             cluster.cluster_params.cluster_params.suff_statistics)
-        log_posterior += log(model.model_hyperparams.α) + lgamma(cluster.cluster_params.cluster_params.suff_statistics.N)
+        log_posterior += log(model.model_hyperparams.α) + logabsgamma(cluster.cluster_params.cluster_params.suff_statistics.N)[1]
     end
     return log_posterior
 end

src/global_params.jl

@@ -7,8 +7,8 @@ data_prefix = "data_prefix"  #If the data file name is bob.npy, this should be '
 iterations = 100
 hard_clustering = false  #Soft or hard assignments
 initial_clusters = 1
-argmax_sample_stop = 0 #Change to hard assignment from soft at iterations - argmax_sample_stop
-split_stop  = 0 #Stop split/merge moves at  iterations - split_stop
+argmax_sample_stop = 5 #Change to hard assignment from soft at iterations - argmax_sample_stop
+split_stop  = 5#Stop split/merge moves at  iterations - split_stop
 
 random_seed = nothing #When nothing, a random seed will be used.
 

src/local_clusters_actions.jl

@@ -359,9 +359,9 @@ function should_split_local!(should_split::AbstractArray{Float32,1},
     log_likihood = log_marginal_likelihood(cp.hyperparams, post, cp.suff_statistics)
 
     log_HR = log(α) +
-        lgamma(cpl.suff_statistics.N) + log_likihood_l +
-        lgamma(cpr.suff_statistics.N) + log_likihood_r -
-        (lgamma(cp.suff_statistics.N) + log_likihood)
+        logabsgamma(cpl.suff_statistics.N)[1] + log_likihood_l +
+        logabsgamma(cpr.suff_statistics.N)[1] + log_likihood_r -
+        (logabsgamma(cp.suff_statistics.N)[1] + log_likihood)
     if log_HR > log(rand())
         should_split .= 1
     end

src/priors/multinomial_prior.jl

@@ -34,7 +34,7 @@ end
 function log_marginal_likelihood(hyper::multinomial_hyper, posterior_hyper::multinomial_hyper, suff_stats::multinomial_sufficient_statistics)
     D = length(suff_stats.points_sum)
     logpi = log(pi)
-    val = lgamma(sum(hyper.α)) -lgamma(sum(posterior_hyper.α)) + sum(lgamma.(posterior_hyper.α) - lgamma.(hyper.α))
+    val = logabsgamma(sum(hyper.α))[1] -logabsgamma(sum(posterior_hyper.α))[1] + sum((x-> logabsgamma(x)[1]).(posterior_hyper.α) - (x-> logabsgamma(x)[1]).(hyper.α))
     return val
 end
 

src/shared_actions.jl

@@ -25,9 +25,9 @@ function should_merge!(should_merge::AbstractArray{Float32,1}, cpl::cluster_para
     log_likihood_l = log_marginal_likelihood(cpl.hyperparams, cpl.posterior_hyperparams, cpl.suff_statistics)
     log_likihood_r = log_marginal_likelihood(cpr.hyperparams, cpr.posterior_hyperparams, cpr.suff_statistics)
     log_likihood = log_marginal_likelihood(cp.hyperparams, cp.posterior_hyperparams, cp.suff_statistics)
-    log_HR = (-log(α) + lgamma(α) -2*lgamma(0.5*α) + lgamma(cp.suff_statistics.N) -lgamma(cp.suff_statistics.N + α) +
-        lgamma(cpl.suff_statistics.N + 0.5*α)-lgamma(cpl.suff_statistics.N)  - lgamma(cpr.suff_statistics.N) +
-        lgamma(cpr.suff_statistics.N + 0.5*α)+ log_likihood- log_likihood_l- log_likihood_r)
+    log_HR = (-log(α) + logabsgamma(α)[1] -2*logabsgamma(0.5*α)[1] + logabsgamma(cp.suff_statistics.N)[1] -logabsgamma(cp.suff_statistics.N + α)[1] +
+        logabsgamma(cpl.suff_statistics.N + 0.5*α)[1]-logabsgamma(cpl.suff_statistics.N)[1]  - logabsgamma(cpr.suff_statistics.N)[1] +
+        logabsgamma(cpr.suff_statistics.N + 0.5*α)[1]+ log_likihood- log_likihood_l- log_likihood_r)
     # log_HR = -(log(α) +
     #     lgamma(cpl.suff_statistics.N) + log_likihood_l +
     #     lgamma(cpr.suff_statistics.N) + log_likihood_r -

src/utils.jl

@@ -66,7 +66,7 @@ end
 function log_multivariate_gamma(x::Number, D::Number)
     res::Float32 = D*(D-1)/4*log(pi)
     for d = 1:D
-        res += lgamma(x+(1-d)/2)
+        res += logabsgamma(x+(1-d)/2)[1]
     end
     return res
 end