CGCNN

Version: v0.1.0 Added: 2022-12-28 Published: 2017-10-27 128k parameters Ensemble of 10 models Missing preds: 2 (0.001%)
Repo Paper DOI Files PR Checkpoint
Predictions
Discovery

ML vs DFT Formation Energies

ML vs DFT Convex Hull Distance

Convex hull distance prediction errors projected onto elements

1 H 0.52
2 He 0.00
3 Li 0.22
4 Be 0.28
5 B 0.42
6 C 0.39
7 N 0.54
8 O 0.80
9 F 0.73
10 Ne 0.00
11 Na 0.23
12 Mg 0.23
13 Al 0.35
14 Si 0.47
15 P 0.39
16 S 0.61
17 Cl 0.56
18 Ar 0.00
19 K 0.21
20 Ca 0.23
21 Sc 0.27
22 Ti 0.50
23 V 0.69
24 Cr 0.34
25 Mn 0.31
26 Fe 0.45
27 Co 0.43
28 Ni 0.39
29 Cu 0.26
30 Zn 0.26
31 Ga 0.29
32 Ge 0.34
33 As 0.35
34 Se 0.48
35 Br 0.50
36 Kr 0.00
37 Rb 0.21
38 Sr 0.21
39 Y 0.22
40 Zr 0.32
41 Nb 0.48
42 Mo 0.40
43 Tc 0.19
44 Ru 0.29
45 Rh 0.31
46 Pd 0.28
47 Ag 0.23
48 Cd 0.21
49 In 0.26
50 Sn 0.28
51 Sb 0.28
52 Te 0.38
53 I 0.43
54 Xe 0.40
55 Cs 0.21
56 Ba 0.22
57 La 0.24
58 Ce 0.24
59 Pr 0.20
60 Nd 0.20
61 Pm 0.17
62 Sm 0.21
63 Eu 0.16
64 Gd 0.23
65 Tb 0.21
66 Dy 0.20
67 Ho 0.21
68 Er 0.20
69 Tm 0.21
70 Yb 0.20
71 Lu 0.20
72 Hf 0.31
73 Ta 0.61
74 W 0.25
75 Re 0.24
76 Os 0.25
77 Ir 0.31
78 Pt 0.31
79 Au 0.25
80 Hg 0.18
81 Tl 0.21
82 Pb 0.24
83 Bi 0.25
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 0.46
91 Pa 0.23
92 U 0.29
93 Np 0.56
94 Pu 0.34
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. Tian Xie  Massachusetts Institute of Technology  
  2. Jeffrey C. Grossman  Massachusetts Institute of Technology  

Trained By

  1. Janosh Riebesell  University of Cambridge, Lawrence Berkeley National Laboratory  

Model Info

  • Model Version v0.1.0
  • Model Type GNN
  • Targets E
  • Openness OSOD
  • Train Task RS2RE
  • Test Task IS2E
  • Trained for Benchmark Yes

Training Set

MP v2022.10.28: 155k structures

Description

Published in 2018, CGCNN was the first crystal graph convolutional neural network to directly learn 8 different DFT-computed material properties from a graph representing the atoms and bonds in a crystal. Illustration of the crystal graph convolutional neural networks Aviary CGCNN model is based on the original implementation in https://github.com/txie-93/cgcnn.

Long

CGCNN was among the first to show that just like in other areas of ML, given large enough training sets, neural networks can learn embeddings that reliably outperform all human-engineered structure features directly from the data.

Hyperparameters

  • graph_construction_radius: 5
  • max_neighbors: null

Dependencies