Building machine learning applications with images, text and table data is not easy. It requires feature engineering, or the use of domain knowledge of data to create the functions that make AI algorithms work, plus sufficient pre-processing of data sets to ensure that no bias occurs in trained models.
That is probably why Amazon has developed AutoGluon, an open source library designed to allow developers to write AI-informed apps with just a few lines of code. It has been publicly launched about a month after the silent turnout at GitHub today.
AutoGluon wants to automate many of the decisions that developers had to make themselves in the past. Normally, tasks such as fine-tuning hyperparameters are performed manually, in which scientists have to anticipate how hyperparameters – which represent the choices made when building an AI model – will influence the training of the model. Another common human-assisted task, the search for neural architecture, involves advanced engineering, at least to the extent that developers must identify the optimal design for their respective models.
To this end, AutoGluon can produce a model with only three lines of code by automatically tuning choices within standard ranges that are known to perform well for a given task. Developers simply indicate when they want to have their trained model ready, and in response, AutoGluon uses the available calculation tools to find the strongest model within the assigned runtime.
It builds on the work that Amazon and Microsoft did three years ago – Gluon – and that was later published in Apache MXNet and Microsoft's Cognitive Toolkit. While Gluon is a machine learning interface that enables developers to build models with a collection of pre-built and optimized components, AutoGluon takes care of the development process end-to-end.
Out of the box can identify AutoGluon models for table prediction, image and text classification and object detection, and it offers an API that more experienced developers can tap to further improve the predictive performance of a model. It requires Python version 3.6 or 3.7 and currently only supports Linux, but Amazon says Mac OSX and Windows versions will be available soon.
“We have developed AutoGluon to truly democratize machine learning and to make the power of in-depth learning available to all developers,” said AWS applied scientist Jonas Mueller in a statement. “AutoGluon solves this problem because all choices are automatically aligned within standard ranges that are known to perform well for the specific task and model.”
The debut of AutoGluon follows the heels of major upgrades from Amazon Web Services “(AWS”) SageMaker, a toolkit for continuous training and deployment of machine learning models for cloud and edge environments. AWS SageMaker Studio is a training and workflow management model that collects all machine learning code, notebooks and folders in one place, while SageMaker Notebooks developers can quickly create a Jupyter notebook for machine learning projects. There is also SageMaker Autopilot, which automates the creation of models by automatically choosing algorithms and tuning those models; SageMaker Experiments, which tests and validates models; SageMaker Debugger, which improves the accuracy of models; and SageMaker Model Monitor, which detects concept drift.
Amazon previously launched AWS Deep Learning Containers, a library of Docker images pre-installed with popular deep learning frameworks, and a series of fully managed services such as Personalize, Textract, Fraud Detector and CodeGuru. With these and stand-alone tools such as AutoGluon, the tech giant of Seattle is chasing a market that, according to Statista, will have a value of $ 118.6 billion by 2025.Tags: #ArtificialIntelligence, #latestNewsAI, #researchAi, #Robotics, amazon, Artificial Intelligence