Start Machine Learning with Amazon SageMaker

Wondering how to work with Machine Learning on the Amazon Web Services (AWS)-cloud environment? Did you come across Amazon SageMaker yet?

With SageMaker, you make Machine Learning models that you take into production quite easily. You can apply these Machine Learning models to data sets. There is no specific configuration needed to train the model in the cloud. 

What does SageMaker do?

SageMaker makes basic Machine Learning models that you can apply to data sets. Afterwards, you can take these models into production quite easily. The models might be made in SageMaker, but you can use them on every platform. This way, you won't have to fine tune the models per platform. 

SageMaker also contains SageMaker Ground Truth. With this application, you can make the models more specific.

What can you use SageMaker for?

The writer of this article in Techzine uses a Machine Learning algoritm that has to recognize cats as an example. To let an algoritm recognize cats, the model has to analyze lots of images of animals. Once that analysis is finished, there is still a possibility photos of camels have ended up in your batch of cat pictures.

Looking for a way to prevent this? That's what you use Ground Truth for. If the model is about 70 or 80 percent sure there is a cat on the photograph, the photograph can pass onto the final model. If it's not sure, it puts the image in 'manual check-up', where you decide if the image can pass or not. You decide how much control Ground Truth has - with the help of Ground Truth you can train the Machine Learning models yourself.

How does SageMaker work?

SageMaker works with workflows. First you prepare a model in Jupyter notebook. Once you're done with this, you configure it into training mode. Once that's done, start training with the SageMaker software development kit (SDK). SageMaker uploads all the scrips and data that is needed for the model to train.

Done training? You can apply the model to production and make predictions via HTTP API. Taking a model into production can be tricky, but according to Hackernoon SageMaker is a safe bet. With python SDK or a web interface, you can define a HTTP endpoint for your model and then it's done! 

Want to know more?

Want to know what you could do with SageMaker? With training, one uses all the technology has to offer. Global Knowledge offers the Practical Data Science with Amazon SageMaker training course. In only a day you figure out SageMaker and the Machine Learning process in SageMaker. Will we see you in the classroom soon?

^{_{We've used information from the following sources in this article: Amazon Web Services, Techzine, Hackernoon}}