-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathptable_heatmap_mace.py
269 lines (226 loc) · 11.2 KB
/
ptable_heatmap_mace.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
import pandas as pd
from collections import defaultdict
from tqdm import tqdm
from pymatgen.core import Structure
from pymatgen.io.ase import AseAtomsAdaptor
import matplotlib.pyplot as plt
from pymatviz import ptable_heatmap
import numpy as np
class QMOFAnalyzer:
def __init__(self, qmof_struct_completed, dict_, e0s, models=None):
self.qmof_struct_completed = qmof_struct_completed
self.dict_ = dict_
self.e0s = e0s
self.models = models
self.element_mae_data = defaultdict(lambda: defaultdict(list))
self.element_frequency = defaultdict(int)
self.element_mae = {}
self.mae_data = {}
self.mae_dfs = {}
self.dict_atomic = defaultdict(dict)
def calculate_errors(self):
for qmof_st in tqdm(self.qmof_struct_completed, desc="Processing"):
qmof_id_st = qmof_st["qmof_id"]
try:
extract_st = qmof_st["structure"]
struct = Structure.from_dict(extract_st)
atoms = AseAtomsAdaptor.get_atoms(struct)
element_counts = struct.composition.get_el_amt_dict()
total_atoms = sum(element_counts.values())
e0_isolated = 0.0
for element, count in element_counts.items():
e0_isolated += self.e0s[element][0] * count
self.element_frequency[element] += count
energies_per_atom = {}
self.dict_atomic[qmof_id_st] = {}
for model in self.models:
energy_key = f"{model}_atomic"
energy = self.dict_[qmof_id_st].get(model)
if energy is not None:
if model != "pbe_energy" and model != "macemof":
energy += self.dict_[qmof_id_st].get("pbe_energy_vdw", 0.0)
energy -= e0_isolated
energy_per_atom = energy / total_atoms
self.dict_atomic[qmof_id_st][energy_key]= energy_per_atom
energies_per_atom[model] = energy_per_atom
pbe_energy_per_atom = energies_per_atom.get("pbe_energy")
if pbe_energy_per_atom is None:
continue
for model, energy_per_atom in energies_per_atom.items():
if model != "pbe_energy":
total_error = abs(energy_per_atom - pbe_energy_per_atom)
for element, count in element_counts.items():
fraction = count / total_atoms
self.element_mae_data[element][model].append(total_error * fraction)
except Exception:
continue
def compute_mae(self):
self.element_mae = {
element: {
model: (sum(errors[model]) / self.element_frequency[element] * 1000)
if self.element_frequency[element] else None
for model in self.models if model != "pbe_energy"
}
for element, errors in self.element_mae_data.items()
}
def format_dict_atomic(self):
formatted_data = []
for qmof_id, values in self.dict_atomic.items():
pbe_energy_per_atom = self.dict_atomic[qmof_id].get("pbe_energy_atomic")
for energy_key, energy_value in values.items():
if energy_key == "pbe_energy_atomic" or energy_value is None:
continue
else:
formatted_data.append({
"qmof_id": qmof_id,
"model": energy_key,
"mae_energy_per_atom": abs(energy_value - pbe_energy_per_atom),
"model_energy_per_atom": energy_value
})
df = pd.DataFrame(formatted_data)
return df
def display_dict_atomic(self):
df = self.format_dict_atomic()
return df
def filter_outliers(self, qmof_props):
df = self.display_dict_atomic()
merged_df = df.set_index('qmof_id').join(qmof_props, on="qmof_id")
merged_df['pbe_energy_atomic'] = [
self.dict_atomic[qid]['pbe_energy_atomic'] for qid in merged_df.index
]
outliers_list = []
inliers_list = []
dict_std_error = {}
x = np.array([data.get("pbe_energy_atomic") for data in self.dict_atomic.values()])
for model in self.models:
if model != "pbe_energy":
y = np.array([data.get(f"{model}_atomic") for data in self.dict_atomic.values()])
valid_indices = [i for i in range(len(x)) if x[i] is not None and y[i] is not None]
xp = x[valid_indices]
yp = y[valid_indices]
dict_std_error[model] = np.std(xp - yp)
upper_limit_ori = xp + 3 * dict_std_error[model]
lower_limit_ori = xp - 3 * dict_std_error[model]
cond = (yp > upper_limit_ori) | (yp < lower_limit_ori)
model_pd = merged_df[merged_df["model"] == f"{model}_atomic"]
outliers = model_pd[cond]
inliers = model_pd[~cond]
print(f"Outliers for model {model}: {outliers.shape}")
print(f"Inliers for model {model}: {inliers.shape}")
outliers_list.append(outliers)
inliers_list.append(inliers)
merged_df_outliers = pd.concat(outliers_list, axis=0)
merged_df_inliers = pd.concat(inliers_list, axis=0)
return merged_df_outliers, merged_df_inliers
def scatter_plot(self, map_title_name=None):
from scipy.stats import gaussian_kde
x = np.array([data.get("pbe_energy_atomic", None) for data in self.dict_atomic.values()])
for model in self.models:
if model != "pbe_energy":
y = np.array([data.get(f"{model}_atomic") for data in self.dict_atomic.values()])
valid_indices = [i for i in range(len(x)) if x[i] is not None and y[i] is not None]
xp = np.array([x[i] for i in valid_indices])
y = np.array([y[i] for i in valid_indices])
if len(xp) != len(y):
continue
error = xp - y
mean_error = np.mean(error)
std_error = np.std(error)
print(mean_error, std_error)
upper_limit_ori = xp + 3 * std_error
lower_limit_ori = xp - 3 * std_error
x_min, x_max = np.min(xp), np.max(xp)
x_range = np.linspace(x_min - 0.3*abs(x_max - x_min),
x_max + 0.3*abs(x_max - x_min), 100)
upper_limit = x_range + 3*std_error
lower_limit = x_range - 3*std_error
outside_count = np.sum((y > upper_limit_ori) | (y < lower_limit_ori))
total_count = len(y)
print(f"Model: {model}")
print(f"Total Points: {total_count}, Points Outside ±3 Std Dev: {outside_count} ({outside_count / total_count:.2%})")
mae = np.mean(np.abs(xp - y))
rmse = np.sqrt(np.mean((xp - y) ** 2))
r2_score = np.corrcoef(xp, y)[0, 1] ** 2
xy = np.vstack([xp, y])
kde = gaussian_kde(xy)(xy)
scatter = plt.scatter(xp, y, c=kde, cmap='viridis', s=10)
plt.colorbar(scatter, label='# MOFs')
plt.plot(x_range, x_range, 'r--')
plt.plot(x_range, upper_limit, color='cyan', linestyle='--')
plt.plot(x_range, lower_limit, color='cyan', linestyle='--')
plt.fill_between(
x_range,
lower_limit,
upper_limit,
color='lightblue',
alpha=0.3
)
plt.xlim(-7, -3)
plt.ylim(-7, -3)
plt.xlabel('PBE Energy [eV/atom]')
if map_title_name is not None:
plt.ylabel(f'{map_title_name[model]} Energy [eV/atom]')
else:
plt.ylabel(f'{model} Energy [eV/atom]')
plt.title(f'QMOF ({int(len(y))} MOF structures)')
label_text = f'MAE = {mae:.3f} eV\nRMSE = {rmse:.3f} eV\n$r^2$ = {r2_score:.3f}\nOut. ±3 Std = {outside_count} ({outside_count / total_count:.2%})'
plt.annotate(label_text, xy=(0.05, 0.95), xycoords='axes fraction',
ha='left', va='top', fontsize=10, bbox=dict(facecolor='white', alpha=0.6))
plt.savefig(f"{model}_scatter_test.png")
plt.close()
def generate_heatmap_data(self, dict_diff_models=None):
self.mae_data = {}
for model in self.models:
if model != "pbe_energy":
self.mae_data[model] = {
element: round(maes[model], 2)
for element, maes in self.element_mae.items() if maes[model] is not None
}
if dict_diff_models:
for diff_model, models_pair in dict_diff_models.items():
model1, model2 = models_pair
self.mae_data[diff_model] = {
element: round(
self.element_mae[element][model1] - self.element_mae[element][model2], 2
)
for element in self.element_mae
if self.element_mae[element].get(model1) is not None and self.element_mae[element].get(model2) is not None
}
self.mae_dfs = {
model: pd.DataFrame(data.items(), columns=["element", model]).set_index("element")
for model, data in self.mae_data.items()
}
def plot_heatmap(self, dict_diff_models=None, map_title_name=None):
from matplotlib.colors import Normalize
for model in self.mae_dfs:
if dict_diff_models and model in dict_diff_models:
cbar_title_ = r"$\Delta E_{MAE}$ $[meV/atom]$"
max_ = round(self.mae_dfs[model].values.max(), 2)
min_ = round(self.mae_dfs[model].values.min(), 2)
cbar_range_ = (-2, 2)
colormap = 'coolwarm'
else:
cbar_title_ = "Energy MAE [meV/atom]"
max_ = round(self.mae_dfs[model].values.max(), 2)
cbar_range_ = (0, 4.5)
colormap = 'viridis'
norm = Normalize(vmin=cbar_range_[0], vmax=cbar_range_[1])
fig = ptable_heatmap(
self.mae_dfs[model],
log=False,
anno_kwargs={"fontsize": 8},
cbar_title=cbar_title_,
# cbar_kwargs=cbar_kwargs_,
cbar_range=cbar_range_,
colormap=colormap,
return_type="figure",
cbar_kwargs={"norm": norm}
)
title = map_title_name.get(model, model)
fig.suptitle(
f"Element-wise MAE {title} for QMOF Structures",
fontsize=16,
fontweight="bold"
)
fig.savefig(f"{model}_mae_test.png")
plt.show()