Hi. Welcome to Ingram Micro.

Please choose your role, so we can direct you to what you’re looking for.

If you’d like to learn more about Ingram Micro global initiatives and operations, visit ingrammicro.com.

FAQ: Deep learning

May 11, 2017

FAQ: Deep learning

What is deep learning?

In short, deep learning is a set of algorithms that take multiple inputs and rationalize them into a high-level output. As a subset of machine learning, its algorithms are inspired by neural networks, a structure and function of the brain. Deep learning uses these networks to learn levels of representation and abstraction that make sense of data such as sound, images and text.

Deep learning frameworks—such as Caffe, Theano and Torch—have been developed to expedite the process of designing and training these neural networks.

If deep learning were to be assigned a human personality trait, it would be “follower.” In deep learning, rules are set and executed, which isn’t always the case for humans. For example, we have a choice to drive vehicles at unsafe speeds by exceeding the speed limit—autonomous, self-driving vehicles are programmed to remain within the parameters of which it was programmed, usually within the constraints of legal law.

What's an example of deep learning?

If a social media application has ever identified you in a photo, based on facial recognition alone, deep learning played a key role in this process. After you’ve been tagged just a few times, algorithms can recognize your face with remarkable precision. Facial key points are detected and the best facial recognition application can identify you with up to 97% accuracy. Deep learning and facial recognition can also be used to identify escaped convicts who appear on security cameras while on the lam.

Other practical applications of deep learning include speech recognition and translation, connected vehicles and life sciences (e.g., predicting the look of cells via 3D imaging data).

How is deep learning technology like human learning?

Think of a typical young child who enters school. In first grade, they learn to solve problems they couldn’t solve in kindergarten; in second grade they learn to solve problems they couldn’t solve in first grade. Deep learning technology is designed to learn progressively like this, only at a faster rate and with greater precision—free of human error.

Is deep learning a type of software? Do hardware and system architecture play a role?

Deep learning requires both hardware and software. Its software is nothing without power, and power comes from the hardware in which its deployed.

At its core, deep learning requires an immensely powerful, efficient and expensive computer. Take the system you may be running at home or in the office and supercharge it with an immensely powerful CPU, graphics card, motherboard, memory, etc. As for cost, world-class deep learning boxes start at six figures.

Who is a major player in deep learning?

The NVIDIA DGX SATURNV is ranked the world’s #1 most efficient supercomputer on the Green500 list, earning a rating of 9.46 gigaflops per watt. NVIDIA claims its deep learning software stack delivers AI-accelerated analytics, data processing, time to insights and visualization of large datasets. The company’s deep learning technology is being used to design more sophisticated neural networks for healthcare and medical research applications—from real-time pathology assessment to predictive analytics for clinical decision-making.