Turi, formerly Dato, announced the launch of GraphLab Create 2.0, Turi Distributed and Turi Predictive Services, machine learning products that allow data scientists and software developers to add machine learning features to applications faster.
GraphLab Create, Turi Distributed and Turi Predictive Services help make machine learning accessible to far more people and allow them to add machine learning features to their applications faster,” said Carlos Guestrin, Dato CEO and Amazon Professor of Machine Learning at the University of Washington. “Today we’re enhancing our most popular GraphLab Create tools, like the Recommender Toolkit, as well as addressing new market opportunities like Lead Scoring.”
GraphLab Create now includes a Lead Scoring Toolkit to allow sales, marketing and other teams to predict which prospects are likely to take an action like subscribing to a service. And the popular Recommender Toolkit has been expanded to cover content-based recommendations (like similar blog posts) when user interaction data like ratings are unavailable.
Trustworthy Machine Learning
Not everyone is comfortable putting a predictive model into production, not even data scientists,” said Guestrin. “Today we’re addressing the need for trustworthy machine learning with new ‘views’ that let our customers visualize the performance of their models before they’re deployed.”
Turi Toolkit Views make it easier for users to understand and validate their models. Users can visually explore, explain and evaluate models with Toolkit Views without writing more code. GraphLab Create introduced Toolkit Views for Recommender and Churn Predictor Toolkits.
GraphLab Create also includes an object detection toolkit for deep learning applications. A previous release integrated MXNet to provide users with easy access to an efficient, flexible open source deep learning framework. Recent benchmarks show MXNet performing as well or better than other frameworks like TensorFlow, Torch or Caffe.
Turi Distributed allows customers to train Boosted Trees and Random Forest models in a distributed fashion using a cluster of machines. This drastically speeds up the creation of tree models from billions of rows, scaling linearly with the number of nodes. Distributed tree model training is robust by periodically saving checkpoints to enable resuming training after interruptions.
Turi Distributed is integrated with Cloudera Manager allowing simplified installation and lifecycle management through Cloudera’s enterprise ecosystem. Users can easily install, configure, and upgrade Turi Distributed. In addition, managed execution environments on nodes allow for faster job launches.
Turi Predicitve Services
Customers can deploy R models as a service with Turi Predictive Services. R developers can leverage an R client library without leaving their development environment.
Turi Predictive Services allows customers to access their deployments through a web-based dashboard, which provides the operational status of the deployment, including metrics, logs, and endpoint metadata. The dashboard makes the hand-off from data science to application development faster and smoother. The dashboard includes model quality metrics that describe the performance of recommender models in real-time, based on live data.
Sign up for the free insideBIGDATA newsletter.