Applied machine learning in chemistry - block course
winter semester 2024/2025
My postdoc, Dr. Iris Guo, and I will offer an online course in February/March 2025, applied machine learning in chemistry.
Classes:
- online
- from 02.17 to 03.14 (on 02.17-19, 02.24-25, 03.03-05 and 03.14), 14h-18h30 (CET time).
I will have five spots for students outside TU Berlin. Students from underrepresented groups will have priority in the selection.
Content
The course will cover topics in supervised machine learning, such as regression and classification.
The course will address theoretical and practical aspects of the following topics:
- calculation of descriptors for small molecules using RDkit
- data types; mean, standard deviation and related estimates; data distribution
- training and test sets; strategies for cross-validation
- supervised methods for classification: k-nearest neighbors, random forests, neural networks
- metrics to evaluate classification performance
- supervised methods for regression: linear regression, random forests, support vector machines, neural networks
- metrics to evaluate regression performance
The data sets to be used in the course will focus on drug design.
Evaluation
The course has no grades. Students are approved or not approved.
To be approved you should:
- Deliver and get sufficient grade in 13/14 jupyter notebooks (grade >= 50)
- Be approved in the project (grade >= 50)
Project
Project: choose a data set from MoleculeNet; perform data analysis, build models using at least two algorithms that you learned in the course, and provide chemical interpretation of your results.
The project should be presented in a seminar (online, March 14, 14h-18h30 (CET time)).
If you are interested in the course, please fill the form below to show your interest (the form is now closed).
You will be contacted until February 3.
If you are a student from TU Berlin, please register on ISIS:
https://isis.tu-berlin.de/course/view.php?id=41759
Classes:
- online
- from 02.17 to 03.14 (on 02.17-19, 02.24-25, 03.03-05 and 03.14), 14h-18h30 (CET time).
I will have five spots for students outside TU Berlin. Students from underrepresented groups will have priority in the selection.
Content
The course will cover topics in supervised machine learning, such as regression and classification.
The course will address theoretical and practical aspects of the following topics:
- calculation of descriptors for small molecules using RDkit
- data types; mean, standard deviation and related estimates; data distribution
- training and test sets; strategies for cross-validation
- supervised methods for classification: k-nearest neighbors, random forests, neural networks
- metrics to evaluate classification performance
- supervised methods for regression: linear regression, random forests, support vector machines, neural networks
- metrics to evaluate regression performance
The data sets to be used in the course will focus on drug design.
Evaluation
The course has no grades. Students are approved or not approved.
To be approved you should:
- Deliver and get sufficient grade in 13/14 jupyter notebooks (grade >= 50)
- Be approved in the project (grade >= 50)
Project
Project: choose a data set from MoleculeNet; perform data analysis, build models using at least two algorithms that you learned in the course, and provide chemical interpretation of your results.
The project should be presented in a seminar (online, March 14, 14h-18h30 (CET time)).
If you are interested in the course, please fill the form below to show your interest (the form is now closed).
You will be contacted until February 3.
If you are a student from TU Berlin, please register on ISIS:
https://isis.tu-berlin.de/course/view.php?id=41759