BOWSR

Version: 2022.9.20 Added: 2022-11-17 Published: 2021-04-20 168k parameters Missing preds: 4,484 (1.745%) pip install maml
Repo Paper Docs DOI Files PyPI PR Checkpoint
Predictions
Geometry OptimizationDiscovery

ML vs DFT Formation Energies

ML vs DFT Convex Hull Distance

Convex hull distance prediction errors projected onto elements

1 H 0.69
2 He 0.00
3 Li 0.18
4 Be 0.27
5 B 0.37
6 C 0.27
7 N 0.43
8 O 0.67
9 F 0.67
10 Ne 0.00
11 Na 0.26
12 Mg 0.18
13 Al 0.29
14 Si 0.33
15 P 0.29
16 S 0.63
17 Cl 0.52
18 Ar 0.00
19 K 0.35
20 Ca 0.22
21 Sc 0.24
22 Ti 0.27
23 V 0.29
24 Cr 0.29
25 Mn 0.30
26 Fe 0.33
27 Co 0.25
28 Ni 0.22
29 Cu 0.25
30 Zn 0.21
31 Ga 0.23
32 Ge 0.25
33 As 0.27
34 Se 0.78
35 Br 0.98
36 Kr 0.00
37 Rb 0.33
38 Sr 0.23
39 Y 0.22
40 Zr 0.28
41 Nb 0.25
42 Mo 0.24
43 Tc 0.17
44 Ru 0.31
45 Rh 0.26
46 Pd 0.27
47 Ag 0.24
48 Cd 0.18
49 In 0.24
50 Sn 0.24
51 Sb 0.22
52 Te 0.56
53 I 0.64
54 Xe 0.80
55 Cs 0.34
56 Ba 0.24
57 La 0.20
58 Ce 0.20
59 Pr 0.19
60 Nd 0.19
61 Pm 0.13
62 Sm 0.20
63 Eu 0.20
64 Gd 0.21
65 Tb 0.21
66 Dy 0.21
67 Ho 0.21
68 Er 0.22
69 Tm 0.23
70 Yb 0.24
71 Lu 0.22
72 Hf 0.33
73 Ta 0.36
74 W 0.21
75 Re 0.25
76 Os 0.30
77 Ir 0.27
78 Pt 0.27
79 Au 0.27
80 Hg 0.17
81 Tl 0.22
82 Pb 0.24
83 Bi 0.20
84 Po 0.00
85 At 0.00
86 Rn 0.00
87 Fr 0.00
88 Ra 0.00
89 Ac 0.13
90 Th 0.25
91 Pa 0.23
92 U 0.27
93 Np 0.22
94 Pu 0.31
95 Am 0.00
96 Cm 0.00
97 Bk 0.00
98 Cf 0.00
99 Es 0.00
100 Fm 0.00
101 Md 0.00
102 No 0.00
103 Lr 0.00
104 Rf 0.00
105 Db 0.00
106 Sg 0.00
107 Bh 0.00
108 Hs 0.00
109 Mt 0.00
110 Ds 0.00
111 Rg 0.00
112 Cn 0.00
113 Nh 0.00
114 Fl 0.00
115 Mc 0.00
116 Lv 0.00
117 Ts 0.00
118 Og 0.00
57-71 La-Lu Lanthanides
89-103 Ac-Lr Actinides

Model Authors

  1. Yunxing Zuo  UC San Diego  
  2. Chi Chen  UC San Diego  
  3. Shyue Ping Ong  UC San Diego  

Model Info

  • Model Version 2022.9.20
  • Model Type BO-GNN
  • Targets E
  • Openness OSOD
  • Train Task RS2RE
  • Test Task IS2RE-SR
  • Trained for Benchmark No

Training Set

Graphs of MP 2019: 133k structures

Description

BOWSR is a Bayesian optimizer with symmetry constraints using a graph deep learning energy model to perform "DFT-free" relaxations of crystal structures.

Long

The authors show that this iterative approach improves the accuracy of ML-predicted formation energies over single-shot predictions.

Training

Uses same version of MEGNet as standalone MEGNet.

Hyperparameters

  • Optimizer Params: {"alpha":0.000676,"n_init":100,"n_iter":100}

Dependencies