Figure 9: Relative improvement of 4S over PCA

from pathlib import Path
import pandas as pd

import pickle
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

from fours.utils.data_handling import read_fours_root_dir

Load the contrast curves

root_dir = Path(read_fours_root_dir())
result_file = root_dir / Path("70_results/x1_fake_planet_experiments/contrast_curves/")

Data in the FOURS_ROOT_DIR found. Location: /fast/mbonse/s4

s4_keys = [
    "s4_median_lambda_000100",
    "s4_median_lambda_001000",
    "s4_median_lambda_010000",
    "s4_median_lambda_100000",
    "s4_mean_lambda_000100",
    "s4_mean_lambda_001000",
    "s4_mean_lambda_010000",
    "s4_mean_lambda_100000",
]

pca_results = dict()
s4_results = dict()

for tmp_file in result_file.iterdir():
    if not tmp_file.name.endswith(".pkl"):
        continue

    tmp_datset_name = tmp_file.name[:-5]

    with open(tmp_file, 'rb') as f:
        tmp_data = pickle.load(f)

    # merge all data
    tmp_result_table = pd.concat(tmp_data.values(), keys=tmp_data.keys(), axis=1)
    pca_columns = tmp_result_table.columns.get_level_values(0).difference(s4_keys)

    pca_results[tmp_datset_name] = tmp_result_table[pca_columns]
    s4_results[tmp_datset_name] = tmp_result_table[s4_keys]

Get the best of PCA and 4S

def get_best_off(dataset_name):
    separations_FWHM = pca_results[dataset_name].index.values
    overall_best_pca = pca_results[dataset_name].max(axis=1).values
    overall_best_s4 = s4_results[dataset_name].max(axis=1).values

    return overall_best_s4, overall_best_pca, separations_FWHM

results = []
separations = []
merged = []

for tmp_dataset_name in pca_results.keys():
    overall_best_s4, overall_best_pca, separations_FWHM = get_best_off(tmp_dataset_name)

    improvement = overall_best_s4 - overall_best_pca
    tmp_dataframe = pd.DataFrame(improvement, separations_FWHM)
    tmp_dataframe.columns = [tmp_dataset_name]
    merged.append(tmp_dataframe)
merged_improvement = pd.concat(merged, axis=1)
merged_improvement.index = np.round(merged_improvement.index, 2)
merged_improvement.index = merged_improvement.index.astype(str)

merged_improvement

	HD7570_331_1101_C-0092_C	HD115892_143_1101_C-0092_E	HD177724_231_091_C-0654_A	HD169022_140_1101_C-0092_E	HD40136_333_1101_C-0092_C	HD2262_305_199_C-0065_C	HD209952_196_089_C-0149_A	HD22049_303_199_C-0065_C	HD11171_332_1101_C-0092_C	HD38678_331_084_C-0396_A	HD22049_351_096_C-0679_A
1.5	1.336391	0.796561	NaN	NaN	1.417691	0.783961	NaN	1.320061	NaN	NaN	1.309371
2.0	1.141301	0.251640	NaN	0.39776	0.390930	0.208360	NaN	1.549092	0.616931	1.175501	0.778641
2.5	-0.084010	1.199871	NaN	0.24099	0.648501	0.692501	NaN	1.347711	0.830621	0.554311	0.542111
3.0	-0.329760	0.423820	0.38716	-0.09616	0.792411	0.438690	NaN	0.661731	0.519381	0.847631	0.260430
3.5	0.250800	0.360410	0.18092	0.33008	0.243280	0.663721	0.502531	0.172580	-0.294600	-0.171900	-0.302620
4.0	0.538801	-0.026310	0.09600	-0.09250	-0.087920	0.529091	0.397430	0.280940	0.258770	0.779031	0.194870
4.5	0.215240	-0.113590	-0.15100	0.06281	0.143750	0.039140	0.185020	-0.005130	0.265410	0.155840	0.209380
5.0	-0.229370	0.118730	0.24845	-0.19008	0.150750	0.304280	0.190390	0.316750	0.222890	-0.102150	0.222710
6.0	0.269460	-0.144220	-0.22980	-0.09785	0.308480	0.038840	-0.178240	0.076010	0.558931	0.157800	0.056870
7.0	0.096760	0.213870	0.43434	NaN	-0.296800	-0.119620	-0.071400	0.194570	0.121640	0.149900	0.227750
8.0	-0.120980	-0.252180	0.19158	0.39983	0.473850	0.186830	0.150120	0.093510	0.265350	0.151020	0.267660
9.0	0.495560	0.216250	0.08607	-0.10961	-0.047910	0.173420	-0.109640	0.155600	0.303980	0.218330	0.358320
10.0	0.196500	0.246540	0.08652	0.43829	0.212000	0.046510	-0.030880	0.321700	0.120750	0.026490	0.501501
11.0	-0.044900	0.098970	0.08107	0.20183	0.032350	-0.145520	-0.032670	0.341920	0.067110	0.072340	0.216870
12.0	-0.038250	0.102620	0.24737	0.16371	0.140500	0.055800	0.136560	0.205950	0.105600	0.100570	0.013880

agpm_datasets = [
    "HD7570_331_1101_C-0092_C",
    "HD115892_143_1101_C-0092_E",
    "HD169022_140_1101_C-0092_E",
    "HD40136_333_1101_C-0092_C",
    "HD2262_305_199_C-0065_C",
    "HD22049_303_199_C-0065_C",
    "HD22049_351_096_C-0679_A",
    "HD11171_332_1101_C-0092_C"
]

dither_datasets = [
    "HD177724_231_091_C-0654_A",
    "HD209952_196_089_C-0149_A",
    "HD38678_331_084_C-0396_A"]

Plots the results

median_results = merged_improvement.T.median(axis=0).values

fig, boxes = plt.subplots(1, 1, figsize=(8, 5.5))

sns.boxplot(merged_improvement.T,
            color="gray",
            width=0.6,
            ax=boxes,
            medianprops={"color": "r", "linewidth": 2, "alpha":0})

boxes.scatter(
    y=median_results,
    x=merged_improvement.T.columns,
    color="darkblue",
    marker="_",
    s=300,
    lw=2,
    label="median",
    zorder=10)

boxes.legend(
    fontsize=12,
    loc='lower left')

plt.hlines(0, xmin= -1, xmax=15, color="black")
plt.hlines(0.5, xmin= -1, xmax=15, ls="dotted", alpha=0.2, color="black", zorder=0)
plt.hlines(1.0, xmin= -1, xmax=15, ls="dotted", alpha=0.2, color="black", zorder=0)
plt.hlines(1.5, xmin= -1, xmax=15, ls="dotted", alpha=0.2, color="black", zorder=0)

boxes.tick_params(
    axis='both', which='major', labelsize=12)

boxes.set_xlim(-1, 15)
boxes.set_ylim(-0.5, 1.8)

boxes.set_xlabel("Separation [$\lambda /D$]",
           size=16, labelpad=10)
boxes.set_ylabel("Gained Contrast [mag]",
           size=16, labelpad=10)
boxes.set_title("4S vs. optimal PCA", fontsize=16, fontweight="bold", y=1.02)

fig.patch.set_facecolor('white')
plt.savefig("./final_plots/05_relative_improvement.pdf", bbox_inches='tight')