From db00da0838060ef361b8ca7ecbc6b1d6bc1c26cd Mon Sep 17 00:00:00 2001
From: Mahmoud Bentriou <mahmoud.bentriou@centralesupelec.fr>
Date: Mon, 25 Jan 2021 02:13:08 +0100
Subject: [PATCH] all of the experiences for property automata with ER model
---
examples/scripts/enzym_1d.jl | 96 +++++++++++++++++++++++++++++++++++
examples/scripts/enzym_2d.jl | 98 ++++++++++++++++++++++++++++++++++++
2 files changed, 194 insertions(+)
create mode 100644 examples/scripts/enzym_1d.jl
create mode 100644 examples/scripts/enzym_2d.jl
diff --git a/examples/scripts/enzym_1d.jl b/examples/scripts/enzym_1d.jl
new file mode 100644
index 0000000..6e38558
--- /dev/null
+++ b/examples/scripts/enzym_1d.jl
@@ -0,0 +1,96 @@
+
+using Distributed
+@everywhere using MarkovProcesses
+using Dates
+using Plots
+using DelimitedFiles
+
+@assert length(ARGS) > 0 "No arguments entered. Please specify at least one argument that specifies the experiment: R1, R2, R3; e.g. julia enzym_1d.jl R3."
+@assert length(ARGS) <= 2 "Too much arguments"
+exp = ARGS[1]
+@assert exp in ["R1", "R2", "R3", "R5"] "Wrong name of experiment"
+nbr_exp = parse(Int, exp[2])
+date_run = Dates.format(now(), "Y-m-d_HH:MM:SS")
+path_results = "./results_$(exp)_$(date_run)/"
+if length(ARGS) == 2
+ path_results = ARGS[2]
+end
+if !isdir(path_results) mkdir(path_results) end
+exps_p_star_k3 = [90.0, 25.0, 10.0]
+exps_prob_p_star_k3 = [1.0, 1.0, 0.9997]
+#=
+samples_abc_post = readdlm(path_results * "mat_p_end.csv", ',')
+samples_weights = readdlm(path_results * "weights_end.csv", ',')[:,1]
+=#
+
+# Chemical reaction network model
+load_model("ER")
+observe_all!(ER)
+# Choice of the automaton
+load_automaton("automaton_F")
+load_automaton("automaton_G")
+dict_automata = Dict()
+dict_automata["R1"]Â = create_automaton_F(ER, 50.0, 75.0, 0.025, 0.05, :P)
+dict_automata["R2"] = create_automaton_F(ER, 50.0, 75.0, 0.05, 0.075, :P)
+dict_automata["R3"] = create_automaton_F(ER, 25.0, 50.0, 0.05, 0.075, :P)
+aut = dict_automata[exp]
+# Synchronized model
+sync_ER = aut * ER
+pm_sync_ER = ParametricModel(sync_ER, (:k3, Uniform(0.0, 100.0)))
+nbr_pa = 1500
+
+r = automaton_abc(pm_sync_ER; nbr_particles = nbr_pa, dir_results = path_results)
+samples_abc_post = r.mat_p_end[1,:]
+samples_weights = r.weights
+
+# Histogram
+histogram(samples_abc_post, weights = r.weights, normalize = :density)
+png(path_results * "histogram.png")
+
+## Satisfaction function
+
+# Optimal bandwidth with KernelEstimator
+using KernelEstimator
+observations = samples_abc_post
+observations_scaled = observations ./ 100
+d_exp_kernel_func = Dict("R1" => betakernel,
+ "R2" => betakernel,
+ "R3" => gaussiankernel)
+kernel_func = d_exp_kernel_func[exp]
+opt_bw = bwlscv(observations_scaled, betakernel)
+@show opt_bw
+pdf_estim_abc_post(x::Float64) = kerneldensity(observations, xeval = [x], lb = 0.0, ub = 100.0, kernel = betakernel, h = opt_bw)[1]
+
+# Optimal bandwidth with BoundedKDE
+using BoundedKDE
+d_exp_kernel = Dict("R1" => "chen99",
+ "R2" => "chen99",
+ "R3" => "gaussian")
+kernel_kde = d_exp_kernel[exp]
+bw_lbound = [0.02, 0.02, 0.05]
+bw_ubound = [1.2, 1.2, 1.2]
+@show kernel_kde
+estim_abc_post = UnivariateKDE(samples_abc_post; kernel = kernel_kde, weights = r.weights, lower_bound = 0.0, upper_bound = 100.0)
+#lscv_bandwidth = select_bandwidth(estim_abc_post, bw_lbound[nbr_exp], bw_ubound[nbr_exp], 10; verbose = true)
+lscv_bandwidth = minimize_lscv(estim_abc_post, bw_lbound[nbr_exp], bw_ubound[nbr_exp]; verbose = true)
+@show lscv_bandwidth, lscv_bandwidth / asymptotic_bandwidth(estim_abc_post)
+estim_abc_post = change_bandwidth(estim_abc_post, lscv_bandwidth)
+#estim_abc_post = change_bandwidth(estim_abc_post, opt_bw)
+pdf_estim_abc_post_pkg(x) = pdf(estim_abc_post, x)
+
+# Estimation of the constant with a probability estimated by Statistical Model Checking
+constant = exps_prob_p_star_k3[nbr_exp] / pdf_estim_abc_post(exps_p_star_k3[nbr_exp])
+constant_pkg = exps_prob_p_star_k3[nbr_exp] / pdf_estim_abc_post_pkg(exps_p_star_k3[nbr_exp])
+
+# Plot of satisfaction probability function
+prob_func(x) = pdf_estim_abc_post(x) * constant
+prob_func_pkg(x) = pdf_estim_abc_post_pkg(x) * constant_pkg
+xaxis = 0:0.1:100
+plot(xaxis, prob_func.(xaxis), label = "ABC spf KernelEstimator", dpi = 480)
+plot!(xaxis, prob_func_pkg.(xaxis), label = "ABC spf BoundedKDE", dpi = 480)
+y_MC = readdlm("/home/moud/plot_R1-3/estim_MC/$(exp)/satisfaction_func.csv", ',')[:,1]
+inf_x, sup_x = 0.0, 100.0
+x_MC = inf_x:((sup_x-inf_x)/(length(y_MC)-1)):sup_x
+plot!(x_MC, y_MC, label = "MC spf")
+png(path_results * "satisfaction_prob_function.png")
+
diff --git a/examples/scripts/enzym_2d.jl b/examples/scripts/enzym_2d.jl
new file mode 100644
index 0000000..dc74706
--- /dev/null
+++ b/examples/scripts/enzym_2d.jl
@@ -0,0 +1,98 @@
+
+using DelimitedFiles
+using Distributed
+@everywhere using MarkovProcesses
+using Dates
+using Plots
+
+@assert length(ARGS) > 0 "No arguments entered. Please specify at least one argument that specifies the experiment: R4, R5, R6; e.g. julia enzym_2d.jl R5."
+@assert length(ARGS) <= 2 "Too much arguments"
+exp = ARGS[1]
+@assert exp in ["R4", "R5", "R6"] "Wrong name of experiment"
+nbr_exp = parse(Int, exp[2]) - 3
+date_run = Dates.format(now(), "Y-m-d_HH:MM:SS")
+path_results = "./results_$(exp)/"
+if length(ARGS) == 2
+ path_results = ARGS[2]
+end
+if !isdir(path_results) mkdir(path_results) end
+exps_p_star_k1_k2 = [[0.01, 60.0], [0.25, 40.0], [0.50, 40.0]]
+exps_prob_p_star_k1_k2 = [0.8634, 1.0, 0.363553]
+
+samples_abc_post = readdlm(path_results * "mat_p_end.csv", ',')
+samples_weights = readdlm(path_results * "weights_end.csv", ',')[:,1]
+#=
+# Chemical reaction network model
+load_model("ER")
+observe_all!(ER)
+# Choice of the automaton
+load_automaton("automaton_F")
+load_automaton("automaton_G")
+load_automaton("automaton_G_and_F")
+dict_automata = Dict()
+dict_automata["R4"]Â = create_automaton_F(ER, 5.0, 15.0, 8.0, 10.0, :P)
+x1, x2, t1, t2 = 50.0, 100.0, 0.0, 0.8
+x3, x4, t3, t4 = 30.0, 100.0, 0.8, 0.9
+dict_automata["R5"] = create_automaton_G(ER, x1, x2, t1, t2, :E)
+dict_automata["R6"] = create_automaton_G_and_F(ER, x1, x2, t1, t2, :E, x3, x4, t3, t4, :P)
+aut = dict_automata[exp]
+# Synchronized model
+sync_ER = aut * ER
+pm_sync_ER = ParametricModel(sync_ER, (:k1, Uniform(0.0, 100.0)), (:k2, Uniform(0.0, 100.0)))
+nbr_pa = 1000
+
+r = automaton_abc(pm_sync_ER; nbr_particles = nbr_pa, dir_results = path_results)
+samples_abc_post = r.mat_p_end
+samples_weights = r.weights
+
+# Histogram
+histogram2d(samples_abc_post[1,:], samples_abc_post[2,:], weights = samples_weights, normalize = :density)
+png(path_results * "histogram.png")
+=#
+## Satisfaction function
+
+# Optimal bandwidth with BoundedKDE
+using BoundedKDE
+d_exp_kernel = Dict("R4" => "chen99",
+ "R5" => "chen99",
+ "R6" => "gaussian")
+kernel_kde = d_exp_kernel[exp]
+#bw_lbound = fill.([0.01, 0.05, 0.01], 2)
+#bw_ubound = fill.([1.0, 0.5, 1.0], 2)
+bw_lbound = [fill(0.1, 2), [0.5, 0.5], fill(0.1, 2)]
+bw_ubound = [fill(1.0, 2), [1.5, 1.5], fill(1.0, 2)]
+coeff_bw_R5 = [0.025704814875365627, 0.07660069455807658] # -0.0166403765
+estim_abc_post = MultivariateKDE(samples_abc_post; kernel = kernel_kde, weights = samples_weights, lower_bound = [0.0, 0.0], upper_bound = [2.0, 100.0])
+bw_ref = asymptotic_bandwidth(estim_abc_post)
+@show kernel_kde
+@show lscv(estim_abc_post, bw_ref .* coeff_bw_R5)
+lscv_bandwidth = select_bandwidth(estim_abc_post, bw_lbound[nbr_exp], bw_ubound[nbr_exp], fill(5, 2); maxevals_int = typemax(Int))
+#=
+lscv_bandwidth = minimize_lscv(estim_abc_post, bw_lbound[nbr_exp], bw_ubound[nbr_exp];
+ coeff_init_bw = coeff_bw_R5, rt = 0.5, f_tol = 1E-5, x_tol = 1E-4,
+ verbosity = 2, maxevals_int = 200)
+=#
+#lscv_bandwidth = [0.004308901764294055, 0.07706748767996473]
+@show lscv_bandwidth, lscv_bandwidth / asymptotic_bandwidth(estim_abc_post)
+estim_abc_post = change_bandwidth(estim_abc_post, lscv_bandwidth)
+#estim_abc_post = change_bandwidth(estim_abc_post, opt_bw)
+pdf_estim_abc_post(x) = pdf(estim_abc_post, x)
+
+# Estimation of the constant with a probability estimated by Statistical Model Checking
+constant = exps_prob_p_star_k1_k2[nbr_exp] / pdf_estim_abc_post(exps_p_star_k1_k2[nbr_exp])
+
+# Plot of satisfaction probability function
+prob_func(x,y) = pdf_estim_abc_post([x,y]) * constant
+@show prob_func(exps_p_star_k1_k2[nbr_exp]...), exps_prob_p_star_k1_k2[nbr_exp]
+xaxis = 0:0.1:1.5
+yaxis = 0:1.0:100.0
+p = plot(title = "Multivariate KDE", dpi = 480, background_color_legend = :transparent)
+plot!(p, xaxis, yaxis, prob_func, st = :surface, c = :coolwarm, camera = (30, 45))
+png(path_results * "estim_abc_satisfaction_prob_function.png")
+x_MC = readdlm("/home/moud/results_last_automata/estim_satisfaction_func_MC/$(exp)/grid_X.csv", ',')
+y_MC = readdlm("/home/moud/results_last_automata/estim_satisfaction_func_MC/$(exp)/grid_Y.csv", ',')
+z_MC = readdlm("/home/moud/results_last_automata/estim_satisfaction_func_MC/$(exp)/satisfaction_func.csv", ',')
+p = plot(title = "ABC MC", dpi = 480, background_color_legend = :transparent)
+plot!(p, [x_MC...], [y_MC...], [z_MC...], st = :surface, c = :coolwarm, camera = (30, 45), label = "MC spf")
+png(path_results * "estim_MC_satisfaction_prob_function.png")
+
--
GitLab