Auto ML explained in 500 words! (Driverless AI example)
Auto ML or Automated Machine Learning is one of the world’s leading developments in AI and ML and has enabled people with almost no coding skills to perform predictive modelling and other ML algorithms on data within minutes. A well known example for Auto ML would be H20.ai’s Driverless AI which has a very interesting methodology for auto ML using Supervised Machine Learning.
Note: This article explains Auto ML using the open-source tool ‘DriverlessAI’ by H20.ai.
Auto ML enables users to custom build a Machine learning model from scratch using their own preferences.
The process is simple. A very vague process would by this way, you load the data, you visualize it and conduct your EDA to see if your data is clean and then move on select your model’s specifications and what do you want to focus on. You select the target column.
You can also change the model’s parameters and have it focus more or less on one of the following:
- Accuracy
- Time
- Interpretability
The scorer can be set to optimize the model for a greater LogLoss score, Accuracy, F1, F2 score etc. for the optimal use case as per your needs. You can also set the systems to be GPU enabled for greater performance.
Once your parameters are set, you can choose the type of result you need and then conduct the experiment for a classification/regression result etc. as required.
To change all the hyperparameters and have complete control on the ML model, advanced users can also use “Expert Settings” in Driverless AI.
When the experiment is initiated, the model tells us what it is currently doing. It takes minutes to complete the experiment. A document which has the entire project report as needed by Data Scientists is provided by Driverless with the entire summary of the experiment.
The summary has extensive reports which are a collection of Dashboards, Scoring pipelines, MLI docs, Surrogate models, Disparate impact analysis, Partial Dependency plots etc. to give the user a holistic view of the entire model.
To productionalize the final model, a python scoring pipeline or a mojo scoring pipeline is provided for the production ready artifact. This is also possible over a REST server or the Amazon Lambda for one click deployments of the model.
Auto ML tools also enable users to compare different models based on different settings and hyperparameters to choose the model best suited to their use case.
Some other amazing open source tools for Auto ML include AutoGluon by Amazon, TransmogrifAI by Apache Spark and Auto-sklearn etc. Do check out this official YouTube video by H20.ai which shows exactly how DriverlessAI works to be amazed by how fast this field is growing! What took hours of time to write efficient and effective codes to perform ML on datasets is now done in minutes.
For more such articles, stay tuned with us as we chart out paths on understanding data and coding and demystify other concepts related to Data Science and Coding. Please leave a review down in the comments. It was a long article, thank you very much for reading it all the way here! Great going!
