def evolve()

in syne_tune/optimizer/schedulers/searchers/bore/de.py [0:0]

• 48 lines of code
• 24 McCabe index (conditional complexity)

    def evolve(self, j):
        best_idv = self.de_pop[self.idxbest]
        current_idv = self.de_pop[j]

        # perform mutation operation
        if self.strategy == "rand1":
            idxs = [idx for idx in range(self.popsize) if idx != j]
            r1, r2, r3 = self.de_pop[np.random.choice(idxs, 3, replace=False)]

            # Step 3.1: Perform mutation and checking
            temp = r1 + self.mut * (r2 - r3)
            vi = np.clip(temp, self.lower_bound, self.upper_bound)

        if self.strategy == "best1":
            idxs = [idx for idx in range(self.popsize) if idx != j]
            r1, r2 = self.de_pop[np.random.choice(idxs, 2, replace=False)]
            temp = best_idv + self.mut * (r1 - r2)
            vi = np.clip(temp, self.lower_bound, self.upper_bound)

        if self.strategy == "rand2":
            idxs = [idx for idx in range(self.popsize) if idx != j]
            r1, r2, r3, r4, r5 = self.de_pop[np.random.choice(idxs, 5, replace=False)]
            temp = r1 + self.mut * (r1 - r2) + self.mut * (r3 - r4)
            vi = np.clip(temp, self.lower_bound, self.upper_bound)

        if self.strategy == "best2":
            idxs = [idx for idx in range(self.popsize) if idx != j]
            r1, r2, r3, r4 = self.de_pop[np.random.choice(idxs, 4, replace=False)]
            temp = best_idv + self.mut * (r1 - r2) + self.mut * (r3 - r4)
            vi = np.clip(temp, self.lower_bound, self.upper_bound)

        if self.strategy == "currenttobest1":
            idxs = [idx for idx in range(self.popsize) if idx != j]
            r1, r2 = self.de_pop[np.random.choice(idxs, 2, replace=False)]
            temp = current_idv + self.mut * (best_idv - current_idv) + self.mut * (r1 - r2)
            vi = np.clip(temp, self.lower_bound, self.upper_bound)

        if self.strategy == "randtobest1":
            idxs = [idx for idx in range(self.popsize) if idx != j]
            r1, r2, r3 = self.de_pop[np.random.choice(idxs, 3, replace=False)]
            temp = r1 + self.mut * (best_idv - r1) + self.mut * (r2 - r3)
            vi = np.clip(temp, self.lower_bound, self.upper_bound)

        # perform crossover operation
        if self.bin == 1:
            cross_points = np.random.rand(self.dimensions) < self.crossp

            if not np.any(cross_points):
                cross_points[np.random.randint(0, self.dimensions)] = True
            ui = np.where(cross_points, vi, current_idv)

        else:
            i = 0
            ui = []
            fill_point = np.random.randint(0, self.dimensions)
            while (i < self.dimensions and
                   np.random.rand(0, 1) < self.crossp):
                ui[fill_point] = vi[fill_point]
                fill_point = (fill_point + 1) % self.dimensions
                i += 1

        return ui