This article is part of a series in which we look at the business of artificial intelligence.With many industries interested in using machine learning algorithms to improve efficiency and lower operating costs, ML products have become one of the main battlegrounds for big tech companies.
In recent years, there has been a proliferation of services that facilitate the creation, training, fine-tuning, and deployment of machine learning models for various organisations.Not to be outdone, Amazon announced new machine learning products at this year’s AWS re:Invent conference, including a no-code ML tool, a data-labeling platform, and a service for optimising machine learning model deployment.
The advantages of the new tools are twofold. These tools will enable organisations that lack the in-house talent and resources to develop their own ML models to get started with machine learning and put their data stores to productive use.
Organizations that are already running machine learning projects will be able to increase the speed and productivity of the machine learning development cycle with the new applied ML tools.Canvas, a no-code interface for SageMaker, AWS’s machine learning development platform, was the most exciting announcement for me.
Canvas is a graphical tool that allows you to manage the entire machine learning pipeline without having to write any code. If you have a basic understanding of machine learning concepts, you can use Canvas to create, train, and test your own models.SageMaker Canvas handles many of the details behind the scenes, such as cleaning and consolidating data, testing different models and algorithms, and running single or batch predictions.
Canvas, like all machine learning applications, begins by importing your data. Canvas allows you to upload files directly from your computer, store them in the Amazon S3 cloud, or store them in data lakes and warehouses like RedShift and Snowflake.
Canvas assists you in merging multiple files into a single dataset and extracting information such as means, data types, and missing and invalid values for each column when you import multiple files. If you don’t manually adjust the blanks, Canvas will fill them in for you.When your dataset is complete, you can build a supervised ML model by selecting a target column to predict. Canvas recognises the type of machine learning task you require (binary/multiclass classification, regression). It also provides useful information such as the relevance of each column in your dataset to the target value, which is useful for dimensionality reduction and model optimization.
Canvas begins training hundreds of different ML models, testing different algorithms and hyperparameters until it finds an optimal solution to the problem after preparing the data and objective. When the ML model is complete, Canvas reports the accuracy as well as additional metrics such as precision, recall, and F1 score. You can also see the model’s confusion matrix, which reports the false positive and negative rates for each class.
You can run the model directly in Canvas to make single or batch predictions. You can also use SageMaker Studio to import the model and integrate it with other ML products. Canvas is an excellent tool for organizations that lack in-house machine learning talent and want to go from raw data to a usable model without writing a single line of code. It is also beneficial to experienced machine learning coders because it eliminates the frustration of manually coding the data analysis, model preparation, and training processes.
Finally, it is already part of a working product, so it will be available to many product teams without requiring any changes to infrastructure or workflows. Faster labelling of data for machine learning modelsData labelling is one of the most difficult aspects of supervised machine learning. Annotating training data with the ground truth requires a tremendous amount of energy and, at times, complicated workflows as well as the hiring and management of a large data-labeling workforce.
To address these issues, Amazon has launched Ground Truth Plus, a service that automates the process of creating labelled training datasets for machine learning models. Ground Truth Plus, like Canvas, is part of the SageMaker product, which means it will be available to Amazon customers who already use the ML development suite.
Ground Truth Plus manages the process for you after you provide it with your raw data and labeling requirements. Amazon handles the annotation process behind the scenes with the help of an expert workforce that specializes in the use case you specify.
Amazon augments human annotators’ efforts with machine learning models that automate the labelling process. The ML models label the datasets in advance, reducing the amount of manual labor required in the annotation process. As the labeling process progresses, the ML models improve, and manual labeling becomes more of a review and adjustment process.
While labelling is taking place, you can keep track of the progress via the Ground Truth Plus Project Portal. According to Amazon, Ground Truth Plus can cut data labelling costs by up to 40%.This tool complements Canvas because it eliminates the need for in-house talent or the management of an outsourced workforce. It will make it simpler for small businesses to begin applied machine learning projects.
lowering the cost of implementing machine learning modelsWhen deploying machine learning models, you must choose a compute instance that will deliver your model in a scalable and cost-effective manner. Managing the compute stack is a complex process that often necessitates weeks of testing and can result in unwelcome failures.
A new service announced at re:Invent, SageMaker Inference Recommender, reduces the complexities of the ML deployment process. Inference Recommender recommends which types of Amazon compute instances will best serve the needs of your application based on your ML model and requirements. You can then deploy your ML model on one of the recommended compute instances right away.
Other features of Inference Recommender include load-testing ML models in simulated environments and setting constraints such as required throughput and latency. This makes evaluating the performance of the ML model and computes instances much easier before deploying your application to production.
Inference Recommender can be a valuable tool for businesses that lack engineers with years of experience tinkering with various cloud computing infrastructures.
It can also help experienced MLOps engineers who want to accelerate the deployment of machine learning models. Amazon is one of several companies focusing on bringing machine learning to talent and resource-constrained industries. With machine learning becoming more accessible and affordable, it will be fascinating to see what new applications emerge in the coming months and years.
Amazon | Don’t forget to follow us on Twitter @njtimesofficial. To get the latest updates