AutoLearning provides plug-ins of computer vision model training and general-purpose model training. You can use these plug-ins to label data on the AutoLearning platform, automatically train models, tune hyperparameters, and evaluate models. To obtain a deeply optimized model, you need only to prepare a small amount of labeled data and specify the training duration. These plug-ins are highly compatible with Elastic Algorithm Service (EAS) of Machine Learning Platform for AI (PAI). You can use these plug-ins to deploy models as RESTful services with ease.

Computer vision model training

The computer vision model training plug-in allows you to label training data, train commonly used computer vision models, and deploy models. This plug-in deeply optimizes the models that are used on mobile devices. You can test the performance of a model on your mobile device by scanning the QR code of the model. You can also deploy a model on your mobile device as needed. You can use the computer vision model training plug-in in the following scenarios:

General-purpose model training

The general-purpose model training plug-in allows you to automatically train models, tune hyperparameters, and evaluate models in the following scenarios. In addition, this plug-in is highly compatible with EAS of PAI. You can use this plug-in to deploy a model as a RESTful service with ease.
  • Matching recall
    Matching recall lies in the recall and sort processes. In the recall process, items to be recommended are selected from a large number of recommendation candidates. Then, these items are sorted in the sort process. You can perform matching recall by using the matching algorithms of Machine Learning Studio to establish a complete recall process with ease. For more information, see Matching recall examples. To perform a matching recall, perform the following steps:
    • Configure a matching strategy: In this step, configure a matching strategy in a Tablestore instance. Collaborative filtering recall, semantic recall, and custom recall strategies are supported.
    • Configure a data filtering strategy: In this step, specify the users and items to be filtered out from the matching results. For example, if you want to filter out the 001 product from the matching list, set 001 as the item to be filtered out in the Tablestore instance. Then, AutoLearning automatically filters out the item.
    • Deploy and test the model: In this step, the matching recall model is tested. If the recommendation result meets your requirements, you can deploy the model as an online service in EAS.