diff --git a/Chaos/CombinedStrangeAttractors.py b/Chaos/CombinedStrangeAttractors.py index db7eae36..5b2bd357 100644 --- a/Chaos/CombinedStrangeAttractors.py +++ b/Chaos/CombinedStrangeAttractors.py @@ -39,7 +39,7 @@ def converges_with_complexity(self): total_unique_ratio = total_unique_points / total_points return False, total_unique_points, total_points, total_unique_ratio - def get_convergence_points(self, x_min=-1, x_max=1, y_min=-1, y_max=1, resolution=100, max_iterations=10000, max_abs=100): + def get_convergence_points(self, x_min=-1, x_max=1, y_min=-1, y_max=1, resolution=100, max_iterations=10000, max_abs=1e6): print("getting convergence array") xs = np.linspace(x_min, x_max, resolution) ys = np.linspace(y_min, y_max, resolution) @@ -56,7 +56,7 @@ def get_convergence_points(self, x_min=-1, x_max=1, y_min=-1, y_max=1, resolutio print("done getting convergence array") return convergence - def plot_convergence_points(self, x_min=-1, x_max=1, y_min=-1, y_max=1, resolution=100, max_iterations=10000, max_abs=100): + def plot_convergence_points(self, x_min=-1, x_max=1, y_min=-1, y_max=1, resolution=100, max_iterations=10000, max_abs=1e6): convergence_array = self.get_convergence_points(x_min, x_max, y_min, y_max, resolution, max_iterations, max_abs) plt.imshow(convergence_array) plt.colorbar() @@ -77,7 +77,7 @@ def find_convergent_initial_condition(self, min_index, max_index): x0 = r * np.cos(theta) y0 = r * np.sin(theta) # x0, y0 = np.random.uniform(-1, 1, (2,)) - trajectory = get_late_trajectory(self, x0, y0, min_index=min_index, max_index=max_index, max_abs=100) + trajectory = get_late_trajectory(self, x0, y0, min_index=min_index, max_index=max_index, max_abs=1e6) if trajectory_diverges(trajectory): continue else: @@ -199,7 +199,7 @@ def __repr__(self): return f"" -def get_late_trajectory(mapping, x0, y0, min_index, max_index, max_abs=100): +def get_late_trajectory(mapping, x0, y0, min_index, max_index, max_abs=1e6): assert max_index > min_index res = [] x,y = x0,y0 @@ -408,8 +408,8 @@ class MappingRecord: # mapping.plot_attractor() # plt.show() - run_mixing_experiment() - # run_mixing_experiment(["AL", "AS"]) + # run_mixing_experiment() + run_mixing_experiment(["X", "Y"]) # while True: # run_perturbation_experiment(perturbation_nelda=random.choice([1, 1.5, 2, 2.5, 3]))