Machine Learning Model for Molecular Energy Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for estimating the total electronic energy of molecules have low accuracy and low transferability, particularly when dealing with significant static correlation energy, due to limited training data and impractical computation of static correlation energy for larger molecules.
Innovation Solution
A computing system generates a training data set with larger quantities of static correlation energy data by computing training molecular structures, Hamiltonians, and energy terms using Hartree-Fock, coupled cluster, and complete active space estimations, and trains an electron energy estimation machine learning model to accurately predict total electronic energies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are trained with limited training data, then training time and computational resources are reduced, but model accuracy and transferability deteriorate
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing static correlation energy terms for a diverse set of training molecular structures before model training. This pre-computed data is then used during the training phase, allowing the model to learn from accurate reference data without requiring real-time computation of static correlation energy, thus resolving the contradiction between training efficiency and model accuracy
Solution Approach 2:
The patent creates a copy of the static correlation energy data by generating a training dataset that includes pre-computed static correlation energy terms for various molecular structures. This copied reference data serves as ground truth for training the machine learning model, enabling the model to achieve high accuracy without directly computing static correlation energy during inference, thereby improving both accuracy and efficiency
2Measurement precision
If static correlation energy is computed for larger molecules, then energy estimation accuracy is improved, but computational complexity and time increase exponentially
Solution Approach 1:
The patent segments the total energy computation into distinct components: kinetic energy, nuclear potential energy, electron repulsion energy, exchange energy, and static correlation energy. By separating static correlation energy as an independent term that can be pre-computed and stored, the method avoids the need for exponential-time computation during molecular energy estimation, thus resolving the contradiction between accuracy and computational complexity
Solution Approach 2:
The patent performs preliminary computation of static correlation energy terms for training molecules and stores these values for later use. This pre-computation approach allows accurate energy estimation without requiring real-time calculation of static correlation energy for larger molecules, effectively reducing computational complexity while maintaining accuracy
3Productivity
If existing machine learning models are used for energy estimation, then computational speed is improved, but accuracy deteriorates due to low transferability
Solution Approach 1:
The patent changes the training parameters by incorporating static correlation energy terms as explicit features in the training dataset. This parameter change enables the model to learn the relationship between molecular structure and static correlation energy, improving transferability to different molecular systems while maintaining computational speed during inference
Data Source
AI summary
A computing system including one or more processing devices configured to generate a training data set. Generating the training data set may include generating training molecular structures, respective training Hamiltonians, and training energy terms. Computing the training energy terms may include, for each of the training Hamiltonians, computing a kinetic energy term, a nuclear potential energy term, an electron repulsion energy term, and an exchange energy term using Hartree-Fock (HF) estimation. Computing the training energy terms may further include, for a first subset of the training Hamiltonians, computing dynamical correlation energy terms using coupled cluster estimation. Computing the training energy terms may further include, for a second subset of the first subset, generating truncated Hamiltonians and computing static correlation energy terms using complete active space (CAS) estimation. The one or more processing devices may train an electron energy estimation machine learning model using the training data set.