Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Prediction of molecular properties using machine learning
Machine learning has allowed for fast advances in many research areas, such as natural language processing and image recognition, and is becoming increasingly important in chemical engineering. In the EPSE group, we use machine learning to predict molecular properties, such as activity coefficients in infinite solutions. For this purpose, we use deep machine learning models developed initially for natural langue processing to learn the “chemical grammar” of a molecule. Thereby, we can predict molecular properties with high accuracy from the structure of a molecule solely.
Keywords: Machine learning, property prediction, thermodynamics, chemical engineering, deep learning, Natural language processing
**Challenge**
Machine learning has allowed for fast advances in many research areas, such as natural language processing and image recognition, and is becoming increasingly important in chemical engineering. In the EPSE group, we developed the SMIELStoPropertiesTransformer(SPT) machine learning to predict molecular properties, such as activity coefficients in infinite solutions. For this purpose, we use deep machine learning models developed initially for natural langue processing to learn the “chemical grammar” of a molecule. Thereby, we can predict molecular properties with high accuracy from the structure of a molecule solely.
**Task**
In our work on machine learning for molecular properties, we can offer multiple projects for semester, bachelor, or master projects: Possible projects are:
1. refining current models by optimizing hyperparameters or model architecture,
1. transferring current models towards new molecular properties, or
1. developing an interface of machine learning with classical process simulations.
Due to the fast-moving of the field, new project ideas are constantly created. Please get in contact with us to learn more about current opportunities.
**What you need**
- Good understanding of thermodynamics
- Above-average grades
- Programming experience: Python
**What you get**
In the project, you will get to work with state-of-the-art machine learning models and have the opportunity to gain a deep understanding of these exciting modeling techniques. Insights gained during the project will transfer to deep learning challenges in other areas.
Furthermore, you will be part of a new and growing dynamic team on process design. If you are interested, please get in touch with us. (Benedikt Winter: bewinter@ethz.ch)
**Challenge**
Machine learning has allowed for fast advances in many research areas, such as natural language processing and image recognition, and is becoming increasingly important in chemical engineering. In the EPSE group, we developed the SMIELStoPropertiesTransformer(SPT) machine learning to predict molecular properties, such as activity coefficients in infinite solutions. For this purpose, we use deep machine learning models developed initially for natural langue processing to learn the “chemical grammar” of a molecule. Thereby, we can predict molecular properties with high accuracy from the structure of a molecule solely.
**Task**
In our work on machine learning for molecular properties, we can offer multiple projects for semester, bachelor, or master projects: Possible projects are:
1. refining current models by optimizing hyperparameters or model architecture, 1. transferring current models towards new molecular properties, or 1. developing an interface of machine learning with classical process simulations.
Due to the fast-moving of the field, new project ideas are constantly created. Please get in contact with us to learn more about current opportunities.
**What you need**
- Good understanding of thermodynamics - Above-average grades - Programming experience: Python
**What you get**
In the project, you will get to work with state-of-the-art machine learning models and have the opportunity to gain a deep understanding of these exciting modeling techniques. Insights gained during the project will transfer to deep learning challenges in other areas. Furthermore, you will be part of a new and growing dynamic team on process design. If you are interested, please get in touch with us. (Benedikt Winter: bewinter@ethz.ch)