waveorder 2.2.1__py3-none-any.whl → 3.0.0__py3-none-any.whl

waveorder/calib/Optimization.py

@@ -0,0 +1,470 @@
+import logging
+
+import numpy as np
+from scipy import optimize
+
+
+class BrentOptimizer:
+    def __init__(self, calib):
+        self.calib = calib
+
+    def _check_bounds(self, lca_bound, lcb_bound):
+        current_lca = self.calib.get_lc("LCA")
+        current_lcb = self.calib.get_lc("LCB")
+
+        # check that bounds don't exceed range of LC
+        lca_lower_bound = (
+            0.01
+            if (current_lca - lca_bound) <= 0.01
+            else current_lca - lca_bound
+        )
+        lca_upper_bound = (
+            1.6
+            if (current_lca + lca_bound) >= 1.6
+            else current_lca + lca_bound
+        )
+
+        lcb_lower_bound = (
+            0.01
+            if current_lcb - lcb_bound <= 0.01
+            else current_lcb - lcb_bound
+        )
+        lcb_upper_bound = (
+            1.6 if current_lcb + lcb_bound >= 1.6 else current_lcb + lcb_bound
+        )
+
+        return (
+            lca_lower_bound,
+            lca_upper_bound,
+            lcb_lower_bound,
+            lcb_upper_bound,
+        )
+
+    def opt_lca(
+        self,
+        cost_function,
+        lower_bound,
+        upper_bound,
+        reference,
+        cost_function_args,
+    ):
+        xopt, fval, ierr, numfunc = optimize.fminbound(
+            cost_function,
+            x1=lower_bound,
+            x2=upper_bound,
+            disp=0,
+            args=cost_function_args,
+            full_output=True,
+        )
+
+        lca = xopt
+        lcb = self.calib.get_lc(self.calib.mmc, self.calib.PROPERTIES["LCA"])
+        abs_intensity = fval + reference
+        difference = fval / reference * 100
+
+        logging.debug("\tOptimize lca ...")
+        logging.debug(f"\tlca = {lca:.5f}")
+        logging.debug(f"\tlcb = {lcb:.5f}")
+        logging.debug(f"\tIntensity = {abs_intensity}")
+        logging.debug(f"\tIntensity Difference = {difference:.4f}%")
+
+        return [lca, lcb, abs_intensity, difference]
+
+    def opt_lcb(
+        self,
+        cost_function,
+        lower_bound,
+        upper_bound,
+        reference,
+        cost_function_args,
+    ):
+        xopt, fval, ierr, numfunc = optimize.fminbound(
+            cost_function,
+            x1=lower_bound,
+            x2=upper_bound,
+            disp=0,
+            args=cost_function_args,
+            full_output=True,
+        )
+
+        lca = self.calib.get_lc(self.calib.mmc, self.calib.PROPERTIES["LCA"])
+        lcb = xopt
+        abs_intensity = fval + reference
+        difference = fval / reference * 100
+
+        logging.debug("\tOptimize lcb ...")
+        logging.debug(f"\tlca = {lca:.5f}")
+        logging.debug(f"\tlcb = {lcb:.5f}")
+        logging.debug(f"\tIntensity = {abs_intensity}")
+        logging.debug(f"\tIntensity Difference = {difference:.4f}%")
+
+        return [lca, lcb, abs_intensity, difference]
+
+    def optimize(self, state, lca_bound, lcb_bound, reference, thresh, n_iter):
+        converged = False
+        iteration = 1
+        self.calib.inten = []
+        optimal = []
+
+        while not converged:
+            logging.debug(f"iteration: {iteration}")
+
+            (
+                lca_lower_bound,
+                lca_upper_bound,
+                lcb_lower_bound,
+                lcb_upper_bound,
+            ) = self._check_bounds(lca_bound, lcb_bound)
+
+            if state == "ext":
+                results_lca = self.opt_lca(
+                    self.calib.opt_lc,
+                    lca_lower_bound,
+                    lca_upper_bound,
+                    reference,
+                    (self.calib.PROPERTIES["LCA"], reference),
+                )
+
+                self.calib.set_lc(self.calib.mmc, results_lca[0], "LCA")
+
+                optimal.append(results_lca)
+
+                results_lcb = self.opt_lcb(
+                    self.calib.opt_lc,
+                    lcb_lower_bound,
+                    lcb_upper_bound,
+                    reference,
+                    (self.calib.PROPERTIES["LCB"], reference),
+                )
+
+                self.calib.set_lc(self.calib.mmc, results_lca[1], "LCB")
+
+                optimal.append(results_lcb)
+
+                results = results_lcb
+
+            if state == "45" or state == "135":
+                results = self.opt_lcb(
+                    self.calib.opt_lc,
+                    lca_lower_bound,
+                    lca_upper_bound,
+                    reference,
+                    (self.calib.PROPERTIES["LCB"], reference),
+                )
+
+                optimal.append(results)
+
+            if state == "60":
+                results = self.opt_lca(
+                    self.calib.opt_lc_cons,
+                    lca_lower_bound,
+                    lca_upper_bound,
+                    reference,
+                    (reference, "60"),
+                )
+
+                swing = (self.calib.lca_ext - results[0]) * self.calib.ratio
+                lca = results[0]
+                lcb = self.calib.lcb_ext + swing
+
+                optimal.append([lca, lcb, results[2], results[3]])
+
+            if state == "90":
+                results = self.opt_lca(
+                    self.calib.opt_lc,
+                    lca_lower_bound,
+                    lca_upper_bound,
+                    reference,
+                    (self.calib.PROPERTIES["LCA"], reference),
+                )
+
+                optimal.append(results)
+
+            if state == "120":
+                results = self.opt_lca(
+                    self.calib.opt_lc_cons,
+                    lca_lower_bound,
+                    lca_upper_bound,
+                    reference,
+                    (reference, "120"),
+                )
+
+                swing = (self.calib.lca_ext - results[0]) * self.calib.ratio
+                lca = results[0]
+                lcb = self.calib.lcb_ext - swing
+
+                optimal.append([lca, lcb, results[2], results[3]])
+
+            # if both LCA and LCB meet threshold, stop
+            if results[3] <= thresh:
+                converged = True
+                optimal = np.asarray(optimal)
+
+                return optimal[-1, 0], optimal[-1, 1], optimal[-1, 2]
+
+            # if loop preforms more than n_iter iterations, stop
+            elif iteration >= n_iter:
+                logging.debug(
+                    f"Exceeded {n_iter} Iterations: Search discontinuing"
+                )
+
+                converged = True
+                optimal = np.asarray(optimal)
+                opt = np.where(optimal == np.min(np.abs(optimal[:, 0])))[0]
+
+                logging.debug(
+                    f"Lowest Inten: {optimal[opt, 0]}, lca = {optimal[opt, 1]}, lcb = {optimal[opt, 2]}"
+                )
+
+                return optimal[-1, 0], optimal[-1, 1], optimal[-1, 2]
+
+            iteration += 1
+
+
+class MinScalarOptimizer:
+    def __init__(self, calib):
+        self.calib = calib
+
+    def _check_bounds(self, lca_bound, lcb_bound):
+        current_lca = self.calib.get_lc("LCA")
+        current_lcb = self.calib.get_lc("LCB")
+
+        if self.calib.mode == "voltage":
+            # check that bounds don't exceed range of LC
+            lca_lower_bound = (
+                0.01
+                if (current_lca - lca_bound) <= 0.01
+                else current_lca - lca_bound
+            )
+            lca_upper_bound = (
+                2.2
+                if (current_lca + lca_bound) >= 2.2
+                else current_lca + lca_bound
+            )
+
+            lcb_lower_bound = (
+                0.01
+                if current_lcb - lcb_bound <= 0.01
+                else current_lcb - lcb_bound
+            )
+            lcb_upper_bound = (
+                2.2
+                if current_lcb + lcb_bound >= 2.2
+                else current_lcb + lcb_bound
+            )
+
+        else:
+            # check that bounds don't exceed range of LC
+            lca_lower_bound = (
+                0.01
+                if (current_lca - lca_bound) <= 0.01
+                else current_lca - lca_bound
+            )
+            lca_upper_bound = (
+                1.6
+                if (current_lca + lca_bound) >= 1.6
+                else current_lca + lca_bound
+            )
+
+            lcb_lower_bound = (
+                0.01
+                if current_lcb - lcb_bound <= 0.01
+                else current_lcb - lcb_bound
+            )
+            lcb_upper_bound = (
+                1.6
+                if current_lcb + lcb_bound >= 1.6
+                else current_lcb + lcb_bound
+            )
+
+        return (
+            lca_lower_bound,
+            lca_upper_bound,
+            lcb_lower_bound,
+            lcb_upper_bound,
+        )
+
+    def opt_lca(
+        self,
+        cost_function,
+        lower_bound,
+        upper_bound,
+        reference,
+        cost_function_args,
+    ):
+        res = optimize.minimize_scalar(
+            cost_function,
+            bounds=(lower_bound, upper_bound),
+            method="bounded",
+            args=cost_function_args,
+        )
+
+        lca = res.x
+        lcb = self.calib.get_lc("LCB")
+        abs_intensity = res.fun + reference
+        difference = res.fun / reference * 100
+
+        logging.debug("\tOptimize lca ...")
+        logging.debug(f"\tlca = {lca:.5f}")
+        logging.debug(f"\tlcb = {lcb:.5f}")
+        logging.debug(f"\tIntensity = {abs_intensity}")
+        logging.debug(f"\tIntensity Difference = {difference:.4f}%")
+
+        return [lca, lcb, abs_intensity, difference]
+
+    def opt_lcb(
+        self,
+        cost_function,
+        lower_bound,
+        upper_bound,
+        reference,
+        cost_function_args,
+    ):
+        res = optimize.minimize_scalar(
+            cost_function,
+            bounds=(lower_bound, upper_bound),
+            method="bounded",
+            args=cost_function_args,
+        )
+
+        lca = self.calib.get_lc("LCA")
+        lcb = res.x
+        abs_intensity = res.fun + reference
+        difference = res.fun / reference * 100
+
+        logging.debug("\tOptimize lcb ...")
+        logging.debug(f"\tlca = {lca:.5f}")
+        logging.debug(f"\tlcb = {lcb:.5f}")
+        logging.debug(f"\tIntensity = {abs_intensity}")
+        logging.debug(f"\tIntensity Difference = {difference:.4f}%")
+
+        return [lca, lcb, abs_intensity, difference]
+
+    def optimize(
+        self, state, lca_bound, lcb_bound, reference, thresh=None, n_iter=None
+    ):
+        (
+            lca_lower_bound,
+            lca_upper_bound,
+            lcb_lower_bound,
+            lcb_upper_bound,
+        ) = self._check_bounds(lca_bound, lcb_bound)
+
+        if state == "ext":
+            optimal = []
+
+            results_lca = self.opt_lca(
+                self.calib.opt_lc,
+                lca_lower_bound,
+                lca_upper_bound,
+                reference,
+                ("LCA", reference),
+            )
+
+            self.calib.set_lc(results_lca[0], "LCA")
+
+            optimal.append(results_lca)
+
+            results_lcb = self.opt_lcb(
+                self.calib.opt_lc,
+                lcb_lower_bound,
+                lcb_upper_bound,
+                reference,
+                ("LCB", reference),
+            )
+
+            self.calib.set_lc(results_lcb[1], "LCB")
+
+            optimal.append(results_lcb)
+
+            # ============BEGIN FINE SEARCH=================
+
+            logging.debug(f"\n\tBeginning Finer Search\n")
+            lca_lower_bound = results_lcb[0] - 0.01
+            lca_upper_bound = results_lcb[0] + 0.01
+            lcb_lower_bound = results_lcb[1] - 0.01
+            lcb_upper_bound = results_lcb[1] + 0.01
+
+            results_lca = self.opt_lca(
+                self.calib.opt_lc,
+                lca_lower_bound,
+                lca_upper_bound,
+                reference,
+                ("LCA", reference),
+            )
+
+            self.calib.set_lc(results_lca[0], "LCA")
+
+            optimal.append(results_lca)
+
+            results_lcb = self.opt_lcb(
+                self.calib.opt_lc,
+                lcb_lower_bound,
+                lcb_upper_bound,
+                reference,
+                ("LCB", reference),
+            )
+
+            self.calib.set_lc(results_lcb[1], "LCB")
+
+            optimal.append(results_lcb)
+
+            # Sometimes this optimization can drift away from the minimum,
+            # this makes sure we use the lowest iteration
+            optimal = np.asarray(optimal)
+            opt = np.where(optimal == np.min(optimal[:][2]))[0]
+
+            lca = float(optimal[opt][0][0])
+            lcb = float(optimal[opt][0][1])
+            results = optimal[opt][0]
+
+        if state == "45" or state == "135":
+            results = self.opt_lcb(
+                self.calib.opt_lc,
+                lcb_lower_bound,
+                lcb_upper_bound,
+                reference,
+                ("LCB", reference),
+            )
+
+            lca = results[0]
+            lcb = results[1]
+
+        if state == "60":
+            results = self.opt_lca(
+                self.calib.opt_lc_cons,
+                lca_lower_bound,
+                lca_upper_bound,
+                reference,
+                ("LCA", reference, "60"),
+            )
+
+            swing = (self.calib.lca_ext - results[0]) * self.calib.ratio
+            lca = results[0]
+            lcb = self.calib.lcb_ext + swing
+
+        if state == "90":
+            results = self.opt_lca(
+                self.calib.opt_lc,
+                lca_lower_bound,
+                lca_upper_bound,
+                reference,
+                ("LCA", reference),
+            )
+            lca = results[0]
+            lcb = results[1]
+
+        if state == "120":
+            results = self.opt_lca(
+                self.calib.opt_lc_cons,
+                lca_lower_bound,
+                lca_upper_bound,
+                reference,
+                ("LCB", reference, "120"),
+            )
+
+            swing = (self.calib.lca_ext - results[0]) * self.calib.ratio
+            lca = results[0]
+            lcb = self.calib.lcb_ext - swing
+
+        return lca, lcb, results[2]