AutoLearning is an automatic machine learning platform that supports online data labeling, automatic model training, hyperparameter tuning, and model evaluation. To get an optimized model, you only need to prepare a small amount of labeled data and set the training time. Models trained by AutoLearning can be deployed in Elastic Algorithm Service (EAS) as RESTful services.

General purpose model training

You can train general purpose models in AutoLearning for matching recall and image classification.

Recall and sorting are core elements of matching recall. Recall allows you to generate a list of recommended products from a large number of existing products. Sorting allows you to sort the products on the recommendation list. You can combine matching recall with the matching algorithm of Machine Learning Studio to establish a complete recall process. Matching recall includes the following modules:
  • Matching strategy configuration: Configure matching strategies in a Tablestore (OTS) instance. Collaborative filtering recall, semantic recall, and custom recall strategies are supported.
  • Data filter strategy configuration: Specify the User and Item to be filtered out from the recall results. For example, if you want to filter out the 001 product from the recall list, set 001 in the OTS instance, and the system then automatically filters out the product.
  • Model deployment and testing: Test recall models. If the recommendation list meets your requirements, you can deploy the model as an online service in EAS.

Image classification allows you to label and classify images such as photos of people, animals, and plants. Image classification includes modules of data labeling, model training and evaluation, and model testing and deployment.