Amazon Web Services (AWS) recently announced the launch of a new service offering called AutoGluon which enables the developers to get started with developing AI models with as little as 3 lines of code by automating more complex processes which are involved in creating an AI or Machine Learning model.
AutoGluon is an opensource toolkit for developers who are not yet expert in Machine learning but want to get started with it by developing applications which deals with machine learning using Tabular data sets, Texts or even Images.
It is built on the works done by Amazon and Microsoft, 3 years ago on Gluon Project which was also published in Apache MXNET and in Cognitive toolkit of Microsoft. Unlike Gluon, which was a machine learning interface to build AI/ML models, AutoGluon has the capability to handle the complete end to end development process.
You can take a look at below sample code from AutoGluon's website.
In general, if you wanted to design an ML model, you would need to have an understanding of feature engineering and in addition to that, domain knowledge is expected to be able to deal with data and create features to the make algorithm work properly. It is also required to pre-process (clean/trim/add extra information) to the data sets to avoid biases appearing in the model.
But, With AutoGluon, developers can handle all those requirements:
To get started with learning AutoGluon, you can take a look at few tutorials at their official website.
What is your review of this service? Do let me know your thoughts by commenting on the post below.
 |
| AWS AutoGluon |
AutoGluon is an opensource toolkit for developers who are not yet expert in Machine learning but want to get started with it by developing applications which deals with machine learning using Tabular data sets, Texts or even Images.
It is built on the works done by Amazon and Microsoft, 3 years ago on Gluon Project which was also published in Apache MXNET and in Cognitive toolkit of Microsoft. Unlike Gluon, which was a machine learning interface to build AI/ML models, AutoGluon has the capability to handle the complete end to end development process.
You can take a look at below sample code from AutoGluon's website.
>>> from autogluon import TabularPrediction as task >>> predictor = task.fit(train_data=task.Dataset(file_path=TRAIN_DATA.csv), label_column=COLUMN_NAME) >>> predictions = predictor.predict(task.Dataset(file_path=TEST_DATA.csv))
In general, if you wanted to design an ML model, you would need to have an understanding of feature engineering and in addition to that, domain knowledge is expected to be able to deal with data and create features to the make algorithm work properly. It is also required to pre-process (clean/trim/add extra information) to the data sets to avoid biases appearing in the model.
But, With AutoGluon, developers can handle all those requirements:
- Complex Hyper parameter tuning
- Data Pre-processing
- Neural Architecture search
With this technology offering, Developers needn't spend time manually experimenting with multiple choices on combinations while designing and developing the deep learning model. Instead they can specify the timeline within which they need their model to be ready and as a solution, AutoGluon will lookout for the available AWS Compute resources to find the best model within the allotted runtime.
AutoGluon can identify models for predicting data in Tabular Data form, Text Classification, Image Classification, Object Identification, etc. In addition to that, it also offers an API for developers to work on advanced aspects of the ML application for performance improvement and customization.
To get started with learning AutoGluon, you can take a look at few tutorials at their official website.
What is your review of this service? Do let me know your thoughts by commenting on the post below.