AWS DeepComposer launches new learning capsule that deep dives into training an autoregressive CNN model

Posted on: Jul 31, 2020

Today, we are excited to release a new learning capsule that deep dives into training an autoregressive convolutional neural network model (AR-CNN) in AWS DeepComposer. AWS DeepComposer gives developers a creative way to get started with machine learning (ML). With AWS DeepComposer, developers can get hands-on, literally, with a musical keyboard and the latest machine learning techniques to expand their ML skills. To learn the concepts of generative AI algorithms, developers can use easy-to-consume, bite-size learning capsules in the AWS DeepComposer console. In our previous learning capsule launched in June 2020, developers were introduced to an AR-CNN model and learned how it is able to add or remove one note at a time to generate music.

In this new learning capsule, Learning Capsule 3, developers learn how to manipulate data used to train an AR-CNN model to generate a composition. The learning capsule explains the use of data augmentation to teach the model to generate target musical composition notes and the use of a custom loss function to accurately measure the learning progress of the model. Additionally, the learning capsule helps developers learn the optimal dataset size, instrument limitations, and data organization to train a custom genre model using the provided Amazon Sagemaker notebook.  

You can get started with AWS DeepComposer and the Learning Capsules at no cost with AWS Free Tier. The AWS DeepComposer keyboard is available on and is priced at $99. An additional 3-month free trial of AWS DeepComposer services are provided with purchase of the keyboard.

Learn more about AWS DeepComposer.