Neurala’s new neural network reduces AI training times from hours to seconds
Artificial intelligence startup Neurala Inc. is claiming a major breakthrough with its deep learning platform, saying it has reduced the time it takes to train a deep neural network from 15 hours to just 20 seconds.
Neurala claims to differentiate itself from other AI software makers because its platform learns by mimicking the human brain, integrating sight, sound and other senses into one system in a rough emulation of how the mind works. The startup added that its deep learning software further excels because it doesn’t rely on servers running in the cloud to train its systems, therefore making it suitable for a new generation of smart cars, children’s toys and industrial machines that perform computations at the edge of the network.
Now, the company said, its Lifelong Deep Neural Network technology is better than ever, slashing learning times to mere seconds and also providing the ability for neural networks to acquire new knowledge without forgetting the previous information it learned – something it claims wasn’t possible before.
“It takes a very long time to train a traditional DNN on a dataset, and once that happens, it must be completely retrained if even a single piece of new information is added,” said Anatoli Gorchetchnikov, Neurala’s chief technology officer.
Traditional deep neural networks have always been “fixed,” Neurala said, which means that once the training was complete, there can be no additional training to improve the model without relearning the knowledge it already possessed.
“[Traditional] deep neural networks still rely on a learning algorithm developed in the ’80s called backpropagation,” Neurala Chief Executive Officer Massimiliano Versace told SiliconANGLE. “This algorithm works but requires many iterations over the dataset, making it really slow, and all the data for future expansion of system knowledge must be kept.”
However, Neurala’s Lifelong DNN works differently, learning about objects it sees incrementally in a way that mimics the cortical and subcortical circuits of the human brain, thereby allowing it to add new information on the fly. The company said its latest breakthrough was made possible thanks to ImageNet, a massive database of images organized by keywords, and other more specific datasets.
The company also said there’s no negative impact on the inference times, memory or precision of its deep neural networks.
