Stanford Artificial Intelligence Lab director Andrew Ng has released a paper aimed at making deep learning more accessible to researchers by showing how to make a neural network that costs about $20,000 using powerful but inexpensive graphics-processing units (GPUs). Deep learning relies on a combination of hardware and software to imitate the functioning of the human brain.
Last year at Google, Ng built a $1-million computerized brain that detects cat videos on Youtube. Ng's system taught itself to find the videos using a 1-billion-connection network on 1,000 computers.
However, Ng says some researchers wondered how they could make progress in deep learning without that level of funding. "I hope that the ability to scale up using much less expensive hardware opens up another avenue for everyone around the world," Ng says. "That’s the reason I’m excited--you can now build a 1-billion-connection model with $20,000 worth of hardware. It opens up the world for researchers to improve the performance of speech recognition and computer vision."
Ng's research also paves the way for more powerful GPU-based applications in the future.
From Wired News
View Full Article
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA
No entries found