Brain-Inspired Computing In Qualcomm Zeroth Processors

Sunday, October 13, 2013

Qualcomm has a new processor in the works that aims to bring a wealth of new enhancements to mobile processors as we know it. The biologically inspired Zeroth processors can learn as they go and can be taught through positive reinforcement.

T aking a ground-up approach to their latest design, Qualcomm’s latest Zeroth processor has been made with speed and power efficiency in mind.

The company's research teams have been working on the new computer architecture that mimics the human brain and nervous system so devices can have embedded cognition driven by brain-inspired computing.

There are three main goals for Qualcomm Zeroth processors:

1. Biologically Inspired Learning

"We want Qualcomm Zeroth products to not only mimic human-like perception but also have the ability to learn how biological brains do," writes Samir Kumar, Qualcomm's Director of Business Development, "Instead of preprogramming behaviors and outcomes with a lot of code, we’ve developed a suite of software tools that enable devices to learn as they go and get feedback from their environment."

In the video below, a robot outfitted with a Qualcomm Zeroth processor is placed it in an environment with colored boxes. The researchers were then able to teach it to visit white boxes only. They did this through dopaminergic-based learning, a.k.a. positive reinforcement—not by programming lines of code.

2. Enable Devices To See and Perceive the World as Humans Do

Brain Architecture

Related articles
Another major pillar of Zeroth processor function is striving to replicate the efficiency with which our senses and our brain communicate information says Kumar. Neuroscientists have created mathematical models that accurately characterize biological neuron behavior when they are sending, receiving or processing information. Neurons send precisely timed electrical pulses referred to as “spikes” only when a certain voltage threshold in a biological cell’s membrane is reached.

These spiking neural networks (SNN) encode and transmit data very efficiently in both how our senses gather information from the environment and then how our brain processes and fuses all of it together.

3. Creation and definition of an Neural Processing Unit—NPU

Qualcomm Neural Processing Unit

The final goal of Qualcomm Zeroth is to create, define and standardize this new processing architecture called a Neural Processing Unit (NPU.) Qualcomm envisions NPU’s in a variety of different devices, but also able to live side-by-side in future system-on-chips. "This way you can develop programs using traditional programing languages, or tap into the NPU to train the device for human-like interaction and behavior," writes Kumar.

SOURCE  Qualcomm

By 33rd SquareSubscribe to 33rd Square