IBM updates PowerAI to make deep learning more accessible

The software distribution package now includes tools that make it easier and faster for application developers to deploy deep learning models.

IBM Watson can now do your taxes for you

IBM on Wednesday is announcing significant updates to PowerAI, its deep learning software distribution package, making it faster for data scientists to deploy deep learning models and easier for developers to integrate computer vision into their applications.

Analysts say that artificial intelligence has reached a tipping point where it's being integrated into just about every service, product, or integration, but there are still major challenges for the data scientists and developers interested in exploiting AI.

Some sectors like financial services have had data scientists on staff for at least five or 10 years, but they've only recently started deploying deep learning methods. PowerAI "gives them these higher level tools that much it make easier and automated," IBM VP Sumit Gupta told ZDNet. "You still have data scientists guiding the whole process, but we're removing some of the steps."


PowerAI, launched last year, is designed to help data scientists deploy open source deep learning frameworks -- including such TensorFlow, CAFFE, Torch, Theano, Chainer, and Anaconda -- that run on IBM Power System severs built for AI, along with Nvidia GPUs. The system uses Nvidia's NVLink technology for fast data transfer between the processors.

"In effect, PowerAI is the Red Hat of deep learning," Gupta said.

To make it easier to deploy these deep learning frameworks, the PowerAI updates include new tools for data preparation. IBM's new cluster virtualization software, Spectrum Conductor, integrates Apache Spark to help automate the step of inputting and tuning data sets.

"It sound like a trivial step but it actually is a huge pain point in this community," Gupta said of data preparation.

Once the data is ready, the new PowerAI also reduces training time with a distributed version of TensorFlow. Taking advantage of a virtualized cluster of GPU-accelerated servers, this distributed version can bring deep learning training time down from weeks to hours.

Additionally, a new software tool called DL Insight automates the process of adjusting deep learning parameters. There are thousands of parameters a data scientist could adjust, and "figuring out which knobs to turn takes a lot of experience," Gupta said. With DL Insight, a deep learning model will quickly deliver more accurate results without all the manual tuning.

Lastly, the updated PowerAI includes a new software tool called AI Vision, enabling developers with limited knowledge about deep learning to train and deploy deep learning models targeted at computer vision.

"You have thousands of companies around the world trying to do image analysis because cameras are becoming extremely inexpensive," Gupta said. "The challenge is most people who use deep learning today have to become fairly knowledgeable in terms of how the deep learning algorithms are working."

With AI Vision, a developer can input an image data set, and the software then helps choose the best model, what framework to use, how to tune it, and it delivers an essentially a trained model. It presents all of this to an app developer via simple frontend interface.

One application, Gupta said, could be electric company equipment inspections. Rather than manually inspecting infrastructure, utilities can use drones to collect images and then use computer vision to spot problems.

"Instead of every electric company in the world hiring a bunch of data scientists, they need the tools that make it easier [to conduct] cognitive image analysis," he said.

Related stories:


You have been successfully signed up. To sign up for more newsletters or to manage your account, visit the Newsletter Subscription Center.
See All
See All