Build a Deep Understanding of Machine Learning Tools Using Small Targeted Projects

Last Updated on

Once you have chosen a machine learning tool you need to improve your skill in using it.

Machine learning tools have a large number of features. Having a recipe for using a feature is not the same as deeply knowing how to use the feature, its limitations or its capabilities.

Also, once you build skills in a machine learning tool you need to be able to demonstrate that you know how to use it well. This is more than using it on a project to deliver a result. This helps, but there are ways to show deep knowledge of a tool as you are building your knowledge.

In this post you will discover how you can investigate features of machine learning tools in order to improve your skill using the tool and to demonstrate your growing skills to others. This can be most helpful for you at the start of new projects and in interviews.

Investigate Machine Learning Tools

Investigate Machine Learning Tools
Photo by paurian, some rights reserved.

Cultivate Skills In Machine Learning Tools

Once you have chosen or are using a machine learning tool, you need to grow your skills in using it. This can happy naturally as you use the tool on the project, but it will be slow, narrow and out of your control.

You need to systematically improve your skills on the features of your machine learning tool.

Show Your Growing Skills

As you are building your skills, you can talk about how well you are doing and how skillful you are becoming. This has far less impact than actually showing the results you have achieved with the tool. Again, you can point to the results you have achieved on projects, but this can be very slow on large projects and frustrating if projects are shut down.

You need to be able to demonstrate your skills using a machine learning tool independent of the projects you may be working on.

Short and Narrow Investigation of Tool Features

The strategy that you can use is to systematically investigate features of your chosen machine learning tools.

A systematic investigation is a short narrowly focused project that uses a feature of a machine learning tool to deliver a result that is useful in the context of the process of applied machine learning.

It is an investigation that you design and have complete control over.

It is also an investigation whose result you own and can share. You can turn it into a video, blog post, open source project or technical report to demonstrate your growing skill with the machine learning tool.

You can investigate features of machine learning tools by using a specific step-by-step process.

Investigate Machine Learning Tools

You can investigate machine learning tools by using a systematic process.

Quick 5-Step Process

  1. Select a Tool. Select the tool that you want to investigate. This may be a tool that you have short listed, a tool that you have described or a tool that you have proceduralized. It may be a tool that you are already using and that you want to learn more about or get better at using.
  2. Select a Feature. Select the feature of the tool that you would like to investigate. Ideally, this would be a feature that covers a sub-task in the process of applied machine learning. This would be a feature that would be useful in the process in making accurate predictions on a machine learning problem. It should have a clear input and result.
  3. Research Feature. Research how to use the feature in the documentation for the tool. This may be API documentation, tutorials, books and websites. Learn the attributes of the feature and how it can be used in one or a variety of ways.
  4. Use Feature on a Dataset. Apply the feature on a dataset. This could be your own dataset, a datasets from competitive machine learning or a standard research dataset from the UCI Machine Learning repository. Use the feature to get a result on the dataset. For example, perform an analysis if the feature is a data analysis technique or create a model and make predictions if the feature is predictive model. Demonstrate the capabilities, limitations, and results that the feature can deliver.
  5. Share Results. Share the results you achieved with friends, colleagues or publicly. You could create a video demonstrating the feature of the tool. Alternatively, you could create a step-by-step tutorial for how to get a result with the tool’s feature on a dataset. Share your code, your procedures and your results.

Share Your Results

Don’t neglect this last important step of sharing.

  • Sharing helps you be focused.
  • Sharing helps others learn.
  • Sharing makes you learn more.
  • Sharing shows your skills.
  • Sharing forces you to finish.

Tips For Great Tool Investigations

Below are tips for getting the most out of your investigation into machine learning tools.

  • Be detailed but not exhaustive. Your investigation of the feature does not need to be exhaustive, but it should be detailed. You should cover variations for the way the feature can be used and try to show its limitations and capabilities.
  • One-off. You do not need to keep the results of the investigation. It is useful to share your results, but the investigation is not a template or procedure for the best use of the feature.
  • Share using suitable media. If you are a video person, share the results using a short screen capture. If you are a writer, consider a blog post or technical report. If you like to teach, consider creating a step-by-step tutorial. If you are a programmer, consider providing code on GitHub.
  • Don’t take too long. Try to limit your investigation to perhaps 1-to-2 hours. A detailed summary for sharing may require the same amount of time again.
  • Don’t stop at one. If you want to get good at using a tool, you should take the time and investigate many of the key features the tool provides.
  • Make the time. If the tool is good enough for you to use on your project, it is important enough for you to make the time to investigate an learn to a deeper level.

Case Studies

Below are some example case studies for tool feature investigations you could consider.

  • Data Analysis Technique: Investigate how to use a tool to perform a specific type of data analysis such as data summary or a graph.
  • Data Preparation Technique: Investigate how to use a tool to perform a specific type of data preparation such as impute missing values or take a sample of a dataset.
  • Predictive Modeling Technique: Investigate how to use a specific machine learning algorithm provided by a tool to create a predictive model.
  • Model Evaluation Technique: Investigate how to use a tool to perform a specific model evaluation technique such as cross validation.
  • Result Improvement Technique: Investigate how to use a tool to perform a specific result improvement technique such as algorithm tuning or model ensembling.

You Can Investigate Machine Learning Tools

You do not need to be an expert in the tool. In fact, this tactic will help you improve your skills and eventually become an expert in the tool. Few people take the time to actually study the tools that they use. If you do, you will know more and accelerate your knowledge leaps and bounds over those that don’t.

You do not need to be an expert in machine learning. There may be features of the tool that you do not fully understand. You can use the process of investigating the feature to both improve your skills with the tool and learn more about machine learning in the process.

You do not need to be a programmer. Not all machine learning tools require programming. You can investigate those machine learning tools that provide graphical, web or command line interfaces instead.

An investigation is not a recipe. You are not creating a recipe for best using a feature. You are investigating the facets of a feature and how it may or may not be used on your project. It may even be a feature that is not directly useful on your project. The goal is learning and demonstrating skill, not creating a jump-start recipe (although these concerns are not mutually exclusive).

Summary

In this post you discovered how to investigate machine learning tools.

You learned that you can use investigations of features of machine learning tools as a way of improving your skill with the tool. You also learned that the results of your investigation can be used as an indicator to generally demonstrate your skill with the tool.

You discovered a quick 5-step process that you can use to investigate any machine learning tool.

  1. Select the machine learning tool that you want to investigate.
  2. Identify the feature of the tool that you want to investigate.
  3. Research the feature so that you know how to use it effectively to get a result.
  4. Use the feature of the tool on a dataset and get a result.
  5. Share what you learned about the feature, the tool and the result that you achieved.

Your Next Step

Is there a machine learning tool or a feature of a tool that you want to investigate?

Investigate the feature Right Now! Take action.

  1. Use the process described above to investigate the feature.
  2. Don’t take any longer than 1 hour to get a result.
  3. Share what you learned and the result in the comments below.

Do you have any questions about this process? Email me or leave a comment.

No comments yet.

Leave a Reply