via Baylor College of Medicine
UMass Amherst, Baylor researchers collaborate on imitating brain’s ‘replay’ ability
Artificial intelligence (AI) experts at the University of Massachusetts Amherst and the Baylor College of Medicine report that they have successfully addressed what they call a “major, long-standing obstacle to increasing AI capabilities” by drawing inspiration from a human brain memory mechanism known as “replay.”
First author and postdoctoral researcher Gido van de Ven and principal investigator Andreas Tolias at Baylor, with Hava Siegelmann at UMass Amherst, write in Nature Communications that they have developed a new method to protect – “surprisingly efficiently” – deep neural networks from “catastrophic forgetting” – upon learning new lessons, the networks forget what they had learned before.
Siegelmann and colleagues point out that deep neural networks are the main drivers behind recent AI advances, but progress is held back by this forgetting.
They write, “One solution would be to store previously encountered examples and revisit them when learning something new. Although such ‘replay’ or ‘rehearsal’ solves catastrophic forgetting,” they add, “constantly retraining on all previously learned tasks is highly inefficient and the amount of data that would have to be stored becomes unmanageable quickly.”
Unlike AI neural networks, humans are able to continuously accumulate information throughout their life, building on earlier lessons. An important mechanism in the brain believed to protect memories against forgetting is the replay of neuronal activity patterns representing those memories, they explain.
Siegelmann says the team’s major insight is in “recognizing that replay in the brain does not store data.” Rather, “the brain generates representations of memories at a high, more abstract level with no need to generate detailed memories.” Inspired by this, she and colleagues created an artificial brain-like replay, in which no data is stored. Instead, like the brain, the network generates high-level representations of what it has seen before.
The “abstract generative brain replay” proved extremely efficient, and the team showed that replaying just a few generated representations is sufficient to remember older memories while learning new ones. Generative replay not only prevents catastrophic forgetting and provides a new, more streamlined path for system learning, it allows the system to generalize learning from one situation to another, they state.
For example, “if our network with generative replay first learns to separate cats from dogs, and then to separate bears from foxes, it will also tell cats from foxes without specifically being trained to do so. And notably, the more the system learns, the better it becomes at learning new tasks,” says van de Ven.
He and colleagues write, “We propose a new, brain-inspired variant of replay in which internal or hidden representations are replayed that are generated by the network’s own, context-modulated feedback connections. Our method achieves state-of-the-art performance on challenging continual learning benchmarks without storing data, and it provides a novel model for abstract level replay in the brain.”
Van de Ven says, “Our method makes several interesting predictions about the way replay might contribute to memory consolidation in the brain. We are already running an experiment to test some of these predictions.”
The Latest Updates from Bing News & Google News
Go deeper with Bing News on:
- MQ-9 set to explore artificial intelligence in UAVson November 25, 2020 at 6:19 pm
General Atomics Aeronautical Systems has landed a $93.3 million contract related to artificial intelligence in unmanned aircraft.
- Pentagon's AI Ethical Priorities to Be Embedded Into its Coming Data and Development Platformon November 25, 2020 at 12:37 pm
At the heart of potential success in this next and future phases is its in-the-making data and AI development platform: the joint common foundation. On top of providing the department’s less ...
- For artificial intelligence, most of the work comes after launchon November 25, 2020 at 12:04 pm
Though building artificial intelligence applications is difficult, operationalizing a tool and monitoring behavior is where the real work begins.
- C3.ai Begins U.S. IPO Rollouton November 25, 2020 at 9:06 am
C3.ai has filed to raise $100 million in an IPO, although the final figure may differ. The firm provides artificial intelligence/machine learning software and services to enterprises worldwide.
- 5G And AI: Complementary Technologies Now And Into The Futureon November 25, 2020 at 8:25 am
Analyst Anshel Sag discusses the combined potential of AI and 5G to enable next-generation wireless use cases.
Go deeper with Google Headlines on:
Go deeper with Bing News on:
Deep neural networks
- AI Neural Networks Are Now Smart Enough To Know When They Shouldn’t Be Trustedon November 25, 2020 at 7:39 am
The idea of all-knowing, uncontrollable artificial intelligence (AI) is scary, but if we’re lucky, it looks like AI might just be able to save us from itself. A team of computer scientists from ...
- AI apocalypse avoided? Neural networks now smart enough to know when they shouldn’t be trustedon November 25, 2020 at 5:45 am
In a development that could scupper the plot of numerous sci-fi movies about an artificial intelligence apocalypse, scientists have created a neural network that is smart enough to know when it ...
- To Err Is AI, Too: Researchers Have Built Neural Networks That Tell When They Make A Mistakeon November 25, 2020 at 1:32 am
MIT researchers build AI that warns errors, in their latest work titled, “Deep Evidential Regression”, which was selected for NeurIPS 2020 ...
- Binarized neural networks show promise for fast, accurate machine learningon November 24, 2020 at 5:14 am
As anyone with a green thumb knows, pruning can promote thriving vegetation. A snip here, a snip there, and growth can be controlled and directed for a more vigorous plant.
- How Archaeologists Are Using Deep Learning to Dig Deeperon November 23, 2020 at 11:36 pm
Dr. Caspari hopes that their creation will give archaeologists a way to find new tombs and to identify important sites so that they can be protected from looters. Other convolutional neural networks ...