Google introduces TensorFlow Lite: During its past two years as an open-source project, TensorFlow has emerged as one of the leading maching learning frameworks in the industry, drawing interest and investment from well beyond its creator and core contributor, Google. Now the TensorFlow team has released a Lite version of the framework aimed at mobile and embedded devices, in a move that will draw excitement but may also seem belated to some.
Here are some of the key details from the team's announcement:
TensorFlow has always run on many platforms, from racks of servers to tiny IoT devices, but as the adoption of machine learning models has grown exponentially over the last few years, so has the need to deploy them on mobile and embedded devices. TensorFlow Lite enables low-latency inference of on-device machine learning models. It is designed from scratch to be:
Lightweight Enables inference of on-device machine learning models with a small binary size and fast initialization/startup
Cross-platform A runtime designed to run on many different platforms, starting with Android and iOS
Fast Optimized for mobile devices, including dramatically improved model loading times, and supporting hardware acceleration
To the last point, TensorFlow Lite includes support for Android Neural Networks API, which works in conjunction with custom hardware accelerators that are found more increasingly on mobile devices these days. It also comes with support for some key machine learning models, including Inception v3 for image recognition; and Smart Reply for quick, automated responses to incoming chats.
TensorFlow Lite should be viewed as an "evolution" of the existing TensorFlow Mobile and will eventually supercede it through a much larger scope, according to the team.
POV: "In the battle of AI platforrms distribution is key," says Constellation VP and principal analyst Holger Mueller. "The more places algorithms can run, the more value there is for developers and enterprises. Google delivering on the TensorFlow Lite announcement gives developers broader reach for one of the popular neural networks and AI platforms, and it's an important move for TensorFlow.
That being said, the move is a fairly obvious one, particularly since Google is the company behind Android, notes Constellation VP and principal analyst Doug Henschen. "Since most of us are carrying supercomputers around with us in our pockets, we should be able to take advantage of machine learning and nueral nets," he says. "What's more, application developers will want these capabilities accessible to users no matter the platform. MXNet and Torch already run on Android and iOS, so it stands to reason that Google would want to extend the versatility of TensorFlow to be a first-class citizen on mobile devices."
ADT buys Datashield for hybrid cybersecurity: While ADT has long been known for its physical security systems—alarms and sensors for businesses and homes alike—in fact, the company has its own offerings for network security, and has partnered with security software providers for some time. Now ADT is taking the strategy a step further with the acquisition of Datashield, which makes a platform for managed threat detection and response. Terms of the deal were not disclosed.
POV: Datashield uses a combination of its software and assigned security experts who investigate and rectify validated threats. This is an advantage over simply sending customers a torrent of security alerts for their own teams to monitor and assess, according to the company. It competes with the likes of Dell SecureWorks.
As others have noted, Datashield presents ADT with ample cross-selling opportunities. Where the challenge lies is educating its customer base on the value proposition, as well as living up to the cost savings Datashield claims its dedicated security monitoring centers can deliver as an outsourcing play, compared to hiring more infosec staffers. ADT is rolling Datashield into a new division called ADT Cybersecurity, which could get some initial traction in a market wracked of late by some of the highest-profile and damaging data breaches ever.
Microsoft embraces MariaDB on Azure: After Oracle acquired Sun Microsystems in 2010, MySQL co-creator Monty Widenius responded to uncertainty over the database's future under Oracle's ownership by leading a fork of the codebase called MariaDB. While experiencing some fits and starts, MariaDB today has become a mature, widely used relational database. A point of validation to that end came this week, with Microsoft's announcement that it is joining the MariaDB Foundation, which oversees the platform's development.
MariaDB will also be offered as a fully managed service on Microsoft Azure, with a preview version coming in the near future. MariaDB will join MySQL and Postgres on Azure.
POV: Microsoft's move isn't as monumental as others it has made in the open-source realm, such as its initial support of MySQL and decision to join the Linux Foundation. Still, it provides more open-source credibility for Redmond while sticking a thumb in the eye of Oracle and its MySQL unit, notes Constellation VP and principal analyst Doug Henschen. Microsoft has taken pains to position Azure as a broadly open cloud platform focused on customer choice, and supporting MariaDB both as a service and on the foundation level falls squarely in line with that attitude.