diff --git a/journal_basin_linspace_clones.py b/journal_basin_linspace_clones.py index 053b997..d97b92f 100644 --- a/journal_basin_linspace_clones.py +++ b/journal_basin_linspace_clones.py @@ -6,6 +6,7 @@ import pickle import pandas as pd import numpy as np import torch +from sklearn import preprocessing from functionalities_test import is_identity_function, test_status from journal_basins import SpawnExperiment, mean_invariate_manhattan_distance @@ -21,8 +22,8 @@ class SpawnLinspaceExperiment(SpawnExperiment): number_clones = number_clones or self.nr_clones df = pd.DataFrame( - columns=['parent', 'MAE_pre', 'MAE_post', 'MSE_pre', 'MSE_post', 'MIM_pre', 'MIM_post', 'noise', - 'status_post']) + columns=['clone', 'parent', 'parent2', 'MAE_pre', 'MAE_post', 'MSE_pre', 'MSE_post', 'MIM_pre', 'MIM_post', 'noise', + 'status_pst']) # For every initial net {i} after populating (that is fixpoint after first epoch); # parent = self.parents[0] @@ -31,7 +32,7 @@ class SpawnLinspaceExperiment(SpawnExperiment): # parent_clone.apply_weights(torch.as_tensor(parent.create_target_weights(parent.input_weight_matrix()))) # parent_clone = parent_clone.apply_noise(self.noise) # self.parents.append(parent_clone) - pairwise_net_list = itertools.combinations(self.parents, 2) + pairwise_net_list = list(itertools.combinations(self.parents, 2)) for net1, net2 in pairwise_net_list: # We set parent start_time to just before this epoch ended, so plotting is zoomed in. Comment out to # to see full trajectory (but the clones will be very hard to see). @@ -50,12 +51,13 @@ class SpawnLinspaceExperiment(SpawnExperiment): # To plot clones starting after first epoch (z=ST_steps), set that as start_time! # To make sure PCA will plot the same trajectory up until this point, we clone the # parent-net's weight history as well. - # in_between_weights = np.linspace(net1_target_data, net2_target_data, number_clones, endpoint=False) - in_between_weights = np.logspace(net1_target_data, net2_target_data, number_clones, endpoint=False) + + in_between_weights = np.linspace(net1_target_data, net2_target_data, number_clones, endpoint=False) + # in_between_weights = np.logspace(net1_target_data, net2_target_data, number_clones, endpoint=False) for j, in_between_weight in enumerate(in_between_weights): clone = Net(net1.input_size, net1.hidden_size, net1.out_size, - name=f"{net1.name}_clone_{str(j)}", start_time=self.ST_steps + 100) + name=f"{net1.name}_{net2.name}_clone_{str(j)}", start_time=self.ST_steps + 100) clone.apply_weights(torch.as_tensor(in_between_weight)) clone.s_train_weights_history = copy.deepcopy(net1.s_train_weights_history) @@ -67,10 +69,16 @@ class SpawnLinspaceExperiment(SpawnExperiment): MSE_pre = MSE(net1_target_data, clone_pre_weights) MIM_pre = mean_invariate_manhattan_distance(net1_target_data, clone_pre_weights) - # Then finish training each clone {j} (for remaining epoch-1 * ST_steps) .. - for _ in range(self.epochs - 1): - for _ in range(self.ST_steps): - clone.self_train(1, self.log_step_size, self.net_learning_rate) + try: + # Then finish training each clone {j} (for remaining epoch-1 * ST_steps) .. + for _ in range(self.epochs - 1): + for _ in range(self.ST_steps): + clone.self_train(1, self.log_step_size, self.net_learning_rate) + if any([torch.isnan(x).any() for x in clone.parameters()]): + raise ValueError + except ValueError: + print("Ran into nan in 'in beetween weights' array.") + continue # Post Training distances for comparison clone_post_weights = clone.create_target_weights(clone.input_weight_matrix()) @@ -81,15 +89,22 @@ class SpawnLinspaceExperiment(SpawnExperiment): # .. log to data-frame and add to nets for 3d plotting if they are fixpoints themselves. test_status(clone) if is_identity_function(clone): - print(f"Clone {j} (of net_{net1.name}) is fixpoint." + print(f"Clone {j} (between {net1.name} and {net2.name}) is fixpoint." f"\nMSE({net1.name},{j}): {MSE_post}" f"\nMAE({net1.name},{j}): {MAE_post}" f"\nMIM({net1.name},{j}): {MIM_post}\n") self.nets.append(clone) - df.loc[clone.name] = [net1.name, MAE_pre, MAE_post, MSE_pre, MSE_post, MIM_pre, MIM_post, + df.loc[len(df)] = [j, net1.name, net2.name, MAE_pre, MAE_post, MSE_pre, MSE_post, MIM_pre, MIM_post, self.noise, clone.is_fixpoint] + for net1, net2 in pairwise_net_list: + value = 'MAE' + c_selector = [f'{value}_pre', f'{value}_post'] + values = df.loc[(df['parent'] == net1.name) & (df['parent2'] == net2.name)][c_selector] + this_min, this_max = values.values.min(), values.values.max() + df.loc[(df['parent'] == net1.name) & + (df['parent2'] == net2.name), c_selector] = (values - this_min) / (this_max - this_min) for parent in self.parents: for _ in range(self.epochs - 1): for _ in range(self.ST_steps): @@ -110,8 +125,8 @@ if __name__ == '__main__': ST_log_step_size = 10 # Define number of networks & their architecture - nr_clones = 100 - ST_population_size = 2 + nr_clones = 25 + ST_population_size = 10 ST_net_hidden_size = 2 ST_net_learning_rate = 0.04 ST_name_hash = random.getrandbits(32) @@ -147,26 +162,26 @@ if __name__ == '__main__': # Pointplot with pre and after parent Distances import seaborn as sns - from matplotlib import pyplot as plt + from matplotlib import pyplot as plt, ticker # ptplt = sns.pointplot(data=exp.df, x='MAE_pre', y='MAE_post', join=False) - ptplt = sns.pointplot(data=exp.df, x='MIM_pre', y='MIM_post', join=False) - ptplt.set(xscale='log', yscale='log') + ptplt = sns.scatterplot(x=exp.df['MAE_pre'], y=exp.df['MAE_post']) + # ptplt.set(xscale='log', yscale='log') x0, x1 = ptplt.axes.get_xlim() y0, y1 = ptplt.axes.get_ylim() lims = [max(x0, y0), min(x1, y1)] # This is the x=y line using transforms ptplt.plot(lims, lims, 'w', linestyle='dashdot', transform=ptplt.axes.transData) ptplt.plot([0, 1], [0, 1], ':k', transform=ptplt.axes.transAxes) - ptplt.set(xlabel='Invariant Manhattan Distance befor Training', - ylabel='Invariant Manhattan Distance after Training') - plt.xticks(rotation=45) - for ind, label in enumerate(ptplt.get_xticklabels()): - if ind % 10 == 0: # every 10th label is kept - label.set_visible(True) - label.set_text(round(float(label.get_text()), 3)) - else: - label.set_visible(False) + ptplt.set(xlabel='Mean Absolute Distance before Self-Training', + ylabel='Mean Absolute Distance after Self-Training') + # ptplt.axes.xaxis.set_major_formatter(ticker.FuncFormatter(lambda x, pos: round(float(x), 2))) + # ptplt.xticks(rotation=45) + #for ind, label in enumerate(ptplt.get_xticklabels()): + # if ind % 10 == 0: # every 10th label is kept + # label.set_visible(True) + # else: + # label.set_visible(False) filepath = exp.directory / 'mim_dist_plot.png' plt.tight_layout()