Energy-Motivated Equivariant Pretraining for 3D Molecular Graphs
نویسندگان
چکیده
Pretraining molecular representation models without labels is fundamental to various applications. Conventional methods mainly process 2D graphs and focus solely on tasks, making their pretrained incapable of characterizing 3D geometry thus defective for downstream tasks. In this work, we tackle pretraining in a complete novel sense. particular, first propose adopt an equivariant energy-based model as the backbone pretraining, which enjoys merits fulfilling symmetry space. Then develop node-level loss force prediction, where further exploit Riemann-Gaussian distribution ensure be E(3)-invariant, enabling more robustness. Moreover, graph-level noise scale prediction task also leveraged promote eventual performance. We evaluate our from large-scale dataset GEOM-QM9 two challenging benchmarks: MD17 QM9. Experimental results demonstrate efficacy method against current state-of-the-art approaches, verify validity design each proposed component. Code available at https://github.com/jiaor17/3D-EMGP.
منابع مشابه
Statistically Motivated 3D Faces Reconstruction
Facial surgeons require a vast amount of knowledge about the human face, in order to decide how to reconstruct an injured or traumatized part. Does this knowledge consist only of explicit notions about face anatomy, or is there also an implicit knowledge of the normal appearance of a face? And if the answer to the latter question is affirmative, is there a way for a computer to automatically le...
متن کاملEnergy of Graphs, Matroids and Fibonacci Numbers
The energy E(G) of a graph G is the sum of the absolute values of the eigenvalues of G. In this article we consider the problem whether generalized Fibonacci constants $varphi_n$ $(ngeq 2)$ can be the energy of graphs. We show that $varphi_n$ cannot be the energy of graphs. Also we prove that all natural powers of $varphi_{2n}$ cannot be the energy of a matroid.
متن کاملOn net-Laplacian Energy of Signed Graphs
A signed graph is a graph where the edges are assigned either positive ornegative signs. Net degree of a signed graph is the dierence between the number ofpositive and negative edges incident with a vertex. It is said to be net-regular if all itsvertices have the same net-degree. Laplacian energy of a signed graph is defined asε(L(Σ)) =|γ_1-(2m)/n|+...+|γ_n-(2m)/n| where γ_1,...,γ_n are the ei...
متن کاملAverage Degree-Eccentricity Energy of Graphs
The concept of average degree-eccentricity matrix ADE(G) of a connected graph $G$ is introduced. Some coefficients of the characteristic polynomial of ADE(G) are obtained, as well as a bound for the eigenvalues of ADE(G). We also introduce the average degree-eccentricity graph energy and establish bounds for it.
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Proceedings of the ... AAAI Conference on Artificial Intelligence
سال: 2023
ISSN: ['2159-5399', '2374-3468']
DOI: https://doi.org/10.1609/aaai.v37i7.25978