Conclusion and next steps

Overview

Duration: 10 min

Questions

• Where can you learn more about machine learning?

Objectives

• Assess how well you understand a machine learning workflow

• Provide feedback on the workshop

• Discuss additional machine learning resources

Model Selection

Choosing the proper machine learning model for a given task requires knowledge of both machine learning models and the domain of the task. Finding the best model for a new task in machine learning is often a research question in itself. Finding a model that performs reasonably well, however, can often be accomplished by carefully considering the task domain and a little trial and error with the validation set.

Some of the questions to consider when choosing a model are:

• How much data is there to train with?
• Does the data contain roughly the same number of instances from each class?
• How many features does the dataset have? Are all of the features relevant, or might some of them not be related to the data’s class?
• What types are the features (numeric, categorical, image, text)?
• What might the decision boundary look like? Is the data likely linearly separable?
• How noisy is the data?

Additional resources

The References page links to additional resources on machine learning concepts and introductory tools. It includes a Jupyter notebook that shows Python code to execute the type of machine learning workflow you ran with the ml4bio software. The Glossary contains definitions of the machine learning terms used in this workshop. You can also use the additional real and simulated datasets that you downloaded to continue exploring how compatible different types of classifiers are with different data patterns.

Key Points

• You are now prepared to consider how machine learning may benefit your research.

• There are many excellent introductory and intermediate resources to help you continue to learn about machine learning.