Reduce deepcopy

.travis.yml

@@ -29,7 +29,7 @@ matrix:
          env: GROUP=MOC
        - os: osx
          env: GROUP=SV
-         
+
     allow_failures:
        - env: GROUP=Test
          os: osx

benchmarks/benchmark.jl

@@ -6,14 +6,14 @@ using Stan
 CONFIG = Dict(
   "model-list" => [
     "gdemo-geweke",
-    "MoC",
+    # "MoC",
     #"normal-loc",
     "normal-mixture",
-    "simplegauss",
+    "gdemo",
     "gauss",
     "bernoulli",
     #"negative-binomial",
-    "lda"
+    # "lda"
   ],
 
   "test-level" => 2   # 1 = model lang, 2 = whole interface

benchmarks/bernoulli-stan.run.jl

@@ -4,7 +4,7 @@ using Stan
 
 include(Pkg.dir("Turing")*"/benchmarks/benchmarkhelper.jl")
 include(Pkg.dir("Turing")*"/example-models/stan-models/bernoulli-stan.data.jl")
-include(Pkg.dir("Turing")*"/example-models/stan-models/bernoulli-stan..model.jl")
+include(Pkg.dir("Turing")*"/example-models/stan-models/bernoulli-stan.model.jl")
 
 stan_model_name = "bernoulli"
 berstan = Stanmodel(name=stan_model_name, model=berstanmodel, nchains=1);

benchmarks/bernoulli.run.jl

@@ -4,7 +4,7 @@ using Stan
 
 include(Pkg.dir("Turing")*"/benchmarks/benchmarkhelper.jl")
 include(Pkg.dir("Turing")*"/example-models/stan-models/bernoulli.data.jl")
-include(Pkg.dir("Turing")*"/example-models/stan-models/bernoulli..model.jl")
+include(Pkg.dir("Turing")*"/example-models/stan-models/bernoulli.model.jl")
 
 bench_res = tbenchmark("HMC(1000, 0.25, 5)", "bermodel", "berdata")
 logd = build_logd("Bernoulli Model", bench_res...)

benchmarks/lda.run.jl

@@ -6,12 +6,15 @@ include(Pkg.dir("Turing")*"/benchmarks/benchmarkhelper.jl")
 include(Pkg.dir("Turing")*"/example-models/stan-models/lda-stan.data.jl")
 include(Pkg.dir("Turing")*"/example-models/stan-models/lda.model.jl")
 
-bench_res = tbenchmark("HMCDA(1000, 0.65, 1.5)", "ldamodel", "data=ldastandata[1]")
-bench_res[4].names = ["phi[1]", "phi[2]"]
-logd = build_logd("LDA", bench_res...)
-
 include(Pkg.dir("Turing")*"/benchmarks/"*"lda-stan.run.jl")
-logd["stan"] = lda_stan_d
-logd["time_stan"] = lda_time
 
-print_log(logd)
+setchunksize(60)
+
+for alg in ["HMC(1000, 0.25, 6)", "HMCDA(1000, 0.65, 1.5)", "NUTS(2000, 1000, 0.65)"]
+  bench_res = tbenchmark(alg, "ldamodel", "data=ldastandata[1]")
+  bench_res[4].names = ["phi[1]", "phi[2]"]
+  logd = build_logd("LDA", bench_res...)
+  logd["stan"] = lda_stan_d
+  logd["time_stan"] = lda_time
+  print_log(logd)
+end

example-models/benchmarks/gauss.model.jl

@@ -1,6 +1,6 @@
 @model gaussmodel(x) = begin
     N = length(x)
-    lam = tzeros(Float64,N)
+    lam = tzeros(Real,N)
     mu ~ Normal(0, sqrt(1000));
     for i = 1:N
       lam[i] ~ Gamma(1, 1)

example-models/notebooks/.ipynb_checkpoints/BayesHmm-checkpoint.ipynb

@@ -0,0 +1,110 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "using Turing, Distributions\n",
+    "using PyPlot, PyCall"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "y = [ 1.0, 1.0, 1.0, 1.0, 2.0, 2.0, 2.0, 3.0, 3.0, 3.0, 2.0, 2.0, 2.0, 1.0, 1.0 ];\n",
+    "N = length(y);  K = 3;"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "@model BayesHmm(y) = begin\n",
+    "    s = tzeros(Int64, N)\n",
+    "    m = tzeros(Real, K)\n",
+    "    T = Array{Array}(K)\n",
+    "    for i = 1:K\n",
+    "        T[i] ~ Dirichlet(ones(K)/K)\n",
+    "        # m[i] ~ Normal(1, 0.1) # Defining m this way causes label-switching problem.\n",
+    "        m[i] ~ Normal(i, 0.01)\n",
+    "    end\n",
+    "    s[1] ~ Categorical(ones(Float64, K)/K)\n",
+    "    for i = 2:N\n",
+    "        s[i] ~ Categorical(vec(T[s[i-1]]))\n",
+    "        y[i] ~ Normal(m[s[i]], 0.01)\n",
+    "    end\n",
+    "    return(s, m)\n",
+    "end"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "g = Gibbs(300, HMC(1, 0.2, 5, :m, :T), PG(50, 1, :s))\n",
+    "c = sample(BayesHmm(y), g);"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "describe(c)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "m = c[:m][111];\n",
+    "s = c[:s][111];\n",
+    "PyPlot.plot(y, linestyle=\"None\", marker=\"+\", color = \"r\")\n",
+    "PyPlot.plot(m[s], linestyle=\"-\", marker=\"*\", color = \"b\")"
+   ]
+  }
+ ],
+ "metadata": {
+  "anaconda-cloud": {},
+  "kernel_info": {
+   "name": "julia-0.5"
+  },
+  "kernelspec": {
+   "display_name": "Julia 0.5.0",
+   "language": "julia",
+   "name": "julia-0.5"
+  },
+  "language_info": {
+   "file_extension": ".jl",
+   "mimetype": "application/julia",
+   "name": "julia",
+   "version": "0.5.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 1
+}

example-models/notebooks/BayesHmm2.ipynb

@@ -0,0 +1,210 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "using Turing, Distributions\n",
+    "using PyPlot, PyCall"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "y = [ 1.0, 1.0, 1.0, 1.0, 2.0, 2.0, 2.0, 3.0, 3.0, 3.0, 2.0, 2.0, 2.0, 1.0, 1.0 ];\n",
+    "N = length(y);\n",
+    "K = 3;"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      " Assume - `T` is a parameter\n",
+      " Assume - `m` is a parameter\n",
+      " Assume - `s` is a parameter\n",
+      " Observe - `y` is an observation\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "BayesHmm (generic function with 4 methods)"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "@model BayesHmm(y) = begin\n",
+    "    s = tzeros(Int64, N)\n",
+    "    m = tzeros(Dual, K)\n",
+    "    T = Array{Array}(K)\n",
+    "    for i = 1:K\n",
+    "        T[i] ~ Dirichlet(ones(K)/K)\n",
+    "        # m[i] ~ Normal(1, 0.1) # Defining m this way causes label-switching problem.\n",
+    "        m[i] ~ Normal(i, 0.01)\n",
+    "    end\n",
+    "    s[1] ~ Categorical(ones(Float64, K)/K)\n",
+    "    for i = 2:N\n",
+    "        s[i] ~ Categorical(vec(T[s[i-1]]))\n",
+    "        y[i] ~ Normal(m[s[i]], 0.01)\n",
+    "    end\n",
+    "    return(s, m, T)\n",
+    "end"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Turing.PG(50,1000,Turing.resampleSystematic,0.5,Set{Any}())"
+      ]
+     },
+     "execution_count": 4,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "g1 = Gibbs(300, HMC(2, 0.2, 5, :m, :T), PG(50, 5, :s))\n",
+    "g2 = PG(50, 1000)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "T[1] = ForwardDiff.Dual{0,Float64}[Dual(0.117584),Dual(0.744224),Dual(0.138192)], T[1] ~ Distributions.Dirichlet{Float64}(alpha=[0.333333,0.333333,0.333333])\n",
+      "m[1] = Dual(0.9967711359430271), m[1] ~ Distributions.Normal{Float64}(μ=1.0, σ=0.01)\n",
+      "T[2] = ForwardDiff.Dual{0,Float64}[Dual(0.0064186),Dual(0.0065692),Dual(0.987012)], T[2] ~ Distributions.Dirichlet{Float64}(alpha=[0.333333,0.333333,0.333333])\n",
+      "m[2] = Dual(2.007863731847725), m[2] ~ Distributions.Normal{Float64}(μ=2.0, σ=0.01)\n",
+      "T[3] = ForwardDiff.Dual{0,Float64}[Dual(0.713092),Dual(0.234245),Dual(0.0526634)], T[3] ~ Distributions.Dirichlet{Float64}(alpha=[0.333333,0.333333,0.333333])\n",
+      "m[3] = Dual(2.9919946593762465), m[3] ~ Distributions.Normal{Float64}(μ=3.0, σ=0.01)\n",
+      "s[1] = Dual(3), s[1] ~ Distributions.Categorical{Float64}(K=3, p=[0.333333,0.333333,0.333333])\n"
+     ]
+    }
+   ],
+   "source": [
+    "s1 = @sample(BayesHmm(y), g1);"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "describe(s1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "m = s1[:m][111];\n",
+    "s = s1[:s][111];\n",
+    "PyPlot.plot(y, linestyle=\"None\", marker=\"+\", color = \"r\")\n",
+    "PyPlot.plot(m[s], linestyle=\"-\", marker=\"*\", color = \"b\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "s2 = @sample(BayesHmm(y), g2);"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "describe(s2)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "m = s2[:m][150];\n",
+    "s = s2[:s][150];\n",
+    "PyPlot.plot(y, linestyle=\"None\", marker=\"+\", color = \"r\")\n",
+    "PyPlot.plot(m[s], linestyle=\"-\", marker=\"*\", color = \"b\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "anaconda-cloud": {},
+  "kernelspec": {
+   "display_name": "Julia 0.5.0",
+   "language": "julia",
+   "name": "julia-0.5"
+  },
+  "language_info": {
+   "file_extension": ".jl",
+   "mimetype": "application/julia",
+   "name": "julia",
+   "version": "0.5.0"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 1
+}