Deep Learning
Internal training, University of Medicine and Pharmacy at Ho Chi Minh city, Department of Organic Chemistry, 2023
This is a set of four (4+) tutorials on basic information of deep learning in drug discovery incoporating tensorflow and pytorch framework, using the Google Colab free cloud-computing environment in Spring 2023.
Introduction
These tutorials were created between Jan-April 2023 as part of the MedAI Training Session 2 for full execution over Google Colab and remote accesibility via web browsers.
Each tutorial includes a brief introduction of the activities to be performed, installation instructions of the open-source software to be used in each session and several programming, visualization and data analysis activities to be achieved during the tutorial.
After this training session, you will know:
- Build a feed forward neural network using tensorflow and pytorch framework
- Validate the model performance using validation data set
- Visualize loss plot
- Optimize model by changing nodes, layers, dropout, regularization,…
- Early stopping technique
- Save and load model
- Basic understanding of transfer learning
Description of the Tutorials
The following is a brief description of each tutorial, along with the open-source software used for each task:
| Tutorial | Description | Software |
|---|---|---|
Lab.01  | Overview of neural networks and tensorflow framework for regression and classification tasks | RDKit, tensorflow |
| Lab.02 | Overview of neural networks and Pytorch framework for regression and classification tasks | RDKit, pytorch, catboost |
| Lab.03 | Building neural network from scratch | RDKit, tensorflow, scikeras |
| Lab.04 | ANN visualization and optimization | RDKit, tensorflow, scikeras |