some new funcs

Author: Kramer84

notebooks/OptimizationBasedModels/Modele1.5D_Auto_Optim_CMC.ipynb

@@ -10967,7 +10967,7 @@
     "\n",
     "result_dict={}\n",
     "\n",
-    "def store_results(x, fp_gld, fp_slack, gld_params, experiment_key=None):\n",
+    "def store_results(x, fp_gld, fp_slack, gld_params, experiment_key=None, result_dict=result_dict):\n",
     "    x_key = otaf.common.bidirectional_string_to_array_conversion(x)\n",
     "    x_dict = {\"FP_GLD\": fp_gld, \"FP_SLACK\":fp_slack, \"GLD_PARAMS\": gld_params}\n",
     "    if experiment_key is None:\n",
@@ -10992,7 +10992,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "SIZE_MC_PF = 500000 #int(1e6) #1e4\n",
+    "SIZE_MC_PF = 100000 #int(1e6) #1e4\n",
     "\n",
     "sample_gld = otaf.sampling.generate_and_transform_sequence(NDim_Defects, SIZE_MC_PF, RandDeviationVect) \n",
     "scale_factor = 1.0\n",
@@ -11015,7 +11015,7 @@
     "    return np.squeeze(neural_model.evaluate_model_non_standard_space(x).detach().numpy())\n",
     "\n",
     "@otaf.optimization.scaling(scale_factor)\n",
-    "def optimization_function_mini(x, failure_slack=0.0, model=model2, experiment_key=None):\n",
+    "def optimization_function_mini(x, failure_slack=0.0, model=model2, experiment_key=None, result_dict=result_dict):\n",
     "    # Here we search the minimal probability of failure\n",
     "    slack = model(x)\n",
     "    gld_params = gld.fit_LMM(slack,  disp_fit=False, disp_optimizer=False)\n",
@@ -11028,12 +11028,12 @@
     "        fp_gld = gld.CDF_num(failure_slack, gld_params)\n",
     "        fp_out = fp_gld\n",
     "    \n",
-    "    store_results(x, fp_gld, fp_slack, gld_params, experiment_key)\n",
+    "    store_results(x, fp_gld, fp_slack, gld_params, experiment_key, result_dict)\n",
     "    return fp_out\n",
     "\n",
     "\n",
     "@otaf.optimization.scaling(scale_factor)\n",
-    "def optimization_function_maxi(x, failure_slack=0.0, model=model2, experiment_key=None):\n",
+    "def optimization_function_maxi(x, failure_slack=0.0, model=model2, experiment_key=None, result_dict=result_dict):\n",
     "    # Here we search the maximal probability of failure so negative output\n",
     "    slack = model(x)\n",
     "    gld_params = gld.fit_LMM(slack, disp_fit=False, disp_optimizer=False)\n",
@@ -11046,7 +11046,7 @@
     "        fp_gld = gld.CDF_num(failure_slack, gld_params)\n",
     "        fp_out = fp_gld\n",
     "    \n",
-    "    store_results(x, fp_gld, fp_slack, gld_params, experiment_key)\n",
+    "    store_results(x, fp_gld, fp_slack, gld_params, experiment_key, result_dict)\n",
     "\n",
     "    return fp_out*-1"
    ]
@@ -11078,7 +11078,7 @@
     "        midof_funcs.two_parallel_straight_lines,\n",
     "        mif_args = (t, X3),\n",
     "        n_dof = 2,\n",
-    "        n_sample = 100000,\n",
+    "        n_sample = 40000,\n",
     "        target = \"std\", #\"prob\",\n",
     "        target_val = sigma_e_pos*np.sqrt(1-(2/np.pi)), #To be folded normal variance!\n",
     "        isNormal = True, #The component distributions are normal\n",
@@ -11115,7 +11115,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "def assembly_constraint_no_mean(x, scale_factor=1.0, experiment_key=None):\n",
+    "def assembly_constraint_no_mean(x, scale_factor=1.0, result_dict=result_dict, experiment_key=None):\n",
     "    \"\"\" The functions takes directly the concatenated list of all normalized parameters (between 0 and 1), \n",
     "    and reconstructs the statistical constraint violation for each feature.\n",
     "    \"\"\"\n",
@@ -11167,8 +11167,66 @@
    "id": "701e5c11-de56-4437-99f5-9019839cc18a",
    "metadata": {},
    "source": [
-    "# Optimizations to find the paramter values that maximize and minimize the rejection rate.\n",
+    "# Optimizations to find the paramter values that maximize and minimize the rejection rate."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "63241638-cc40-480a-9692-8e4dc7c9859b",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def pf_min_max_optimizer(failure_slack=0.0):\n",
+    "    # Initial guess\n",
+    "    x0 = [0.5] * NDim_Defects  # Initial guess\n",
+    "    \n",
+    "    # Perform the local optimization using COBYQA directly\n",
+    "    res_maxi = minimize(\n",
+    "        optimization_function_maxi, x0,\n",
+    "        args=(failure_slack),\n",
+    "        method=\"COBYQA\", \n",
+    "        jac=None, \n",
+    "        bounds=Bounds(0.0, 1.0, keep_feasible=True),\n",
+    "        constraints = nonLinearConstraint,\n",
+    "        options={\n",
+    "            \"f_target\": -0.1, \n",
+    "            \"maxiter\": 1000,\n",
+    "            \"maxfev\": 4000,\n",
+    "            \"feasibility_tol\": 1e-9,\n",
+    "            \"initial_tr_radius\": np.sqrt(2*10),\n",
+    "            \"final_tr_radius\": 1e-5,\n",
+    "            \"disp\": True,\n",
+    "            \"scale\": False #True\n",
+    "        }\n",
+    "    )\n",
     "\n",
+    "    # Perform the local optimization using COBYQA directly\n",
+    "    res_mini = minimize(\n",
+    "        optimization_function_mini, x0, \n",
+    "        args=(failure_slack),\n",
+    "        method=\"COBYQA\", \n",
+    "        jac=None, \n",
+    "        bounds=Bounds(0.0, 1.0, keep_feasible=True),\n",
+    "        constraints = nonLinearConstraint,\n",
+    "        options={\n",
+    "            \"f_target\": -0.01,\n",
+    "            \"maxiter\": 1000,\n",
+    "            \"maxfev\": 4000,\n",
+    "            \"feasibility_tol\": 1e-9,\n",
+    "            \"initial_tr_radius\": np.sqrt(2*10),\n",
+    "            \"final_tr_radius\": 1e-5,\n",
+    "            \"disp\": True,\n",
+    "            \"scale\": False #True\n",
+    "        }\n",
+    "    )\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5cb89faf-31a3-42f5-860c-043667015974",
+   "metadata": {},
+   "source": [
     "## First maximize Pf"
    ]
   },
@@ -11358,29 +11416,6 @@
     }
    ],
    "source": [
-    "# Initial guess\n",
-    "x0_maxi = [0.5] * NDim_Defects  # Initial guess\n",
-    "\n",
-    "# Perform the local optimization using COBYQA directly\n",
-    "res_maxi = minimize(\n",
-    "    optimization_function_maxi, \n",
-    "    x0_maxi,\n",
-    "    args=(0.05),\n",
-    "    method=\"COBYQA\", \n",
-    "    jac=None, \n",
-    "    bounds=Bounds(0.0, 1.0, keep_feasible=True),\n",
-    "    constraints = nonLinearConstraint,\n",
-    "    options={\n",
-    "        \"f_target\": -0.1, \n",
-    "        \"maxiter\": 1000,\n",
-    "        \"maxfev\": 4000,\n",
-    "        \"feasibility_tol\": 1e-9,\n",
-    "        \"initial_tr_radius\": np.sqrt(2*10),\n",
-    "        \"final_tr_radius\": 1e-5,\n",
-    "        \"disp\": True,\n",
-    "        \"scale\": False #True\n",
-    "    }\n",
-    ")\n",
     "\n",
     "print(\"Maximization Result with COBYQA:\")\n",
     "print(res_maxi)"
@@ -11619,26 +11654,7 @@
     "# Initial guess\n",
     "x0_mini = [0.5] * NDim_Defects  # Initial guess\n",
     "\n",
-    "# Perform the local optimization using COBYQA directly\n",
-    "res_mini = minimize(\n",
-    "    optimization_function_mini, \n",
-    "    x0_mini, \n",
-    "    args=(0.05),\n",
-    "    method=\"COBYQA\", \n",
-    "    jac=None, \n",
-    "    bounds=Bounds(0.0, 1.0, keep_feasible=True),\n",
-    "    constraints = nonLinearConstraint,\n",
-    "    options={\n",
-    "        \"f_target\": -0.01,\n",
-    "        \"maxiter\": 1000,\n",
-    "        \"maxfev\": 4000,\n",
-    "        \"feasibility_tol\": 1e-9,\n",
-    "        \"initial_tr_radius\": np.sqrt(2*10),\n",
-    "        \"final_tr_radius\": 1e-5,\n",
-    "        \"disp\": True,\n",
-    "        \"scale\": False #True\n",
-    "    }\n",
-    ")\n",
+    "\n",
     "\n",
     "print(\"minimization Result with COBYQA:\")\n",
     "print(res_mini)"

src/otaf/common/_common.py

@@ -15,7 +15,8 @@
     "get_tqdm_range",
     "alphabet_generator",
     "threshold_for_percentile_positive_values_below",
-    "bidirectional_string_to_array_conversion"
+    "bidirectional_string_to_array_conversion",
+    "arrays_close_enough"
 ]
 
 import re
@@ -654,3 +655,6 @@ def bidirectional_string_to_array_conversion(x):
         return str_to_arr(x)
     else:
         return arr_to_str(x)
+
+def arrays_close_enough(arr1, arr2, tolerance=1e-6):
+    return np.allclose(arr1, arr2, atol=tolerance)