Latest in APDRC
March 8, 2019
Deep learning is a subset of machine learning, both of which are forms of artificial intelligence. Deep learning algorithms “learn” via a multi-layered approach (hence the term “deep”) that is defined by an architecture of algorithms. Layers of pattern recognition/learning are built upon one another to advance the process.
“There are a lot of buzzwords being used in this space to simplify communication,” says David Patschke, director, AI/ML Strategy, Precision workstations at Dell. “Unfortunately, it’s doing the exact opposite. It's creating more confusion. A better way to evaluate all these disciplines is to examine what they have in common. When we do this, we see that producing a model is central to each of them. This model, then, drives the outcome in the form of a prediction or recommendation, for example.”
The type of data the model is trained on is one differentiator.
“Deep learning differs from traditional machine learning techniques in that they can automatically learn representations from data such as images, video or text, without introducing hand-coded rules or human domain knowledge,” according to NVIDIA.
Deep learning and AI have come to the fore in all kinds of systems and engineering applications given the huge amounts of data now being collected from the web and IoT, and the increasing availability and deployment of graphics processing units (GPUs), according to Jim Tung, MathWorks Fellow.
“More and more systems rely on object recognition in imaging and vision, whether you’re talking about computer vision for security purposes, robotics through things like cameras, or autonomous vehicles, which rely on perception through cameras and Lidar,” Tung told DE. “It’s all about perception and identifying objects in patterns of data, and deep learning has proven to be an accurate and fast way to meet the needs of these applications.”
That speed is key, because deep learning training involves huge data sets. NVIDIA Volta and Turing GPUs, such as the NVIDIA Quadro RTX family, use Tensor Cores to speed up training and deep learning performance.
“We are currently just scratching the surface of what deep learning can do,” says Naji El Masri, CEO of Noesis Solutions in the March 2019 issue of Digital Engineering. “Right now, we are using it to provide accurate models that enable engineering teams to better understand the behavior of a system. This approach is very effective, especially when handling massive amounts of data that would be unmanageable with any conventional approach. Deep neural networks are capable of much more than this, however. Their capability to reproduce the behavior of complex, non-linear systems with almost arbitrary accuracy … enables a large number of applications.”
Have a Plan for AI
Determining which new applications to focus on involves some old-fashioned business sense. The “garbage-in-garbage-out” maxim still applies.
“Where things are currently, being able to explicitly define the use case and why you want that model is important,” Patschke says. “When designing a product, ask what you are solving for and is it relevant to moving the needle to making business more efficient, customers happy or a better product.”
According to Gartner estimates released last year, the global enterprise value derived from AI was expected to total $1.2 trillion in 2018, a 70% increase from 2017. “AI technologies can only deliver value if they are part of the organization’s strategy and used in the right way,” said Alexander Linden, research vice president at Gartner.