A Hyperdimensional Computing System That Performs All Core Computations In-Memoryhttps://techxplore.com/news/2020-06-hyperdimensional-core-in-memory.htmlHyperdimensional computing (HDC) is an emerging computing approach inspired by patterns of neural activity in the human brain. This unique type of computing can allow artificial intelligence systems to retain memories and process new information based on data or scenarios it previously encountered.
... To model neural activity patterns, HDC systems use rich algebra that defines a set of rules to build, bind and bundle different hypervectors. Hypervectors are holographic 10,000-dimensional (pseudo) random vectors with independent and identically distributed components. By using these hypervectors, HDC enables the creation of powerful computing systems that can be used to complete sophisticated cognitive tasks, such as object detection, language recognition, voice and video classification, time series analysis, text categorization and analytical reasoning.
... "In our HDC architecture, the encoding of information and memory storage are separate processes by construction," Sebastian and Rahimi said. "This key disentanglement is recently appreciated in modern deep neural networks to rescue them from catastrophic forgetting and to enable few-shot learning as well as retaining for a lifetime. Our architecture and representational system will play a central role for the next generation of AI to deliver systems that can learn fast, retain information throughout their lifetime and do this efficiently even with the right materials and substrates."
Geethan Karunaratne et al.
In-memory hyperdimensional computing, Nature Electronics (2020)
http://dx.doi.org/10.1038/s41928-020-0410-3-------------------------------
Using Astrocytes to Change the Behavior of Robots Controlled By Neuromorphic Chipshttps://techxplore.com/news/2020-07-astrocytes-behavior-robots-neuromorphic-chips.html... An astrocyte is a different type of brain cell that has recently been found to do a lot more than merely fill up spaces between neurons, as researchers believed for over a century. Studies are finding that these cells also play key roles in brain functions, including learning and central pattern generation (CPG), which is the basis for critical rhythmic behaviors such as breathing and walking.
Although astrocytes are now known to underlie numerous brain functions, most existing computer systems inspired by the human brain only target the structure and function of neurons. Aware of this gap in existing literature, researchers at Rutgers University are developing brain-inspired algorithms that also account for and replicate the functions of astrocytes. In a paper pre-published on arXiv and set to be presented at the ICONS 2020 Conference in July, they introduce a neuromorphic central pattern generator (CPG) modulated by artificial astrocytes that successfully entrained several rhythmic walking behaviors in their in-house robots.
"Everything that artificial neural nets do, and they do a lot these days, is based on the neurocomputing dogma that 'brain equals neurons,'" Konstantinos Michmizos, an assistant professor of computer science at Rutgers University and the lead researcher in this study, told TechXplore. "Astrocytes are two to 10 times more plentiful than neurons. The impact of understanding or mimicking what more of half the brain is doing is enormous."
Michmizos and his team introduced a new approach to neuromorphic research aimed at understanding and mimicking the human brain in its entirety by replicating the seamless ways in which neurons and astrocytes work together to produce specific behaviors. Notably, they are the first to look at artificial intelligence (AI) development from a perspective that does not consider neurons as the only processing unit in the brain, but instead introducing astrocytes in neural networks as a second processing unit.
In their recent study, they used Intel's Loihi neuromorphic chips.
In the system devised by the researchers, robotic functions emerge naturally from the plastic interaction between artificial neurons and astrocytes. Therefore, their CPG's structure and functioning differs greatly from mainstream learning algorithms, which focus only on neurons and do not take full advantage of current knowledge about how the brain works.
"Astrocytes sense the world and change the neuronal activity and the emerging robotic behavior," Michmizos said. "By allowing astrocytes to change how neurons talk to each other, the network changes how it controls the legged robot without changing its topography. This plastic cellular function that changes how neurons transmit nerve impulses in time is a fundamentally different approach from the mainstream learning algorithms that can only change the network's structure."
Polykretis et al.,
An astrocyte-modulated neuromorphic central pattern generator for hexapod robot locomotion on Intel's Loihi.https://arxiv.org/abs/2006.04765 -----------------------------------
Automating the Search for Entirely New 'Curiosity' Algorithmshttps://techxplore.com/news/2020-04-automating-curiosity-algorithms.htmlMIT researchers used machine learning to find entirely new algorithms for encoding exploration. Their machine-designed algorithms outperformed human-designed algorithms on the wide range of simulated tasks and environments
... Engineers have discovered many ways of encoding curious exploration into machine learning algorithms. A research team at MIT wondered if a computer could do better, based on a long history of enlisting computers in the search for new algorithms.
"Algorithms designed by humans are very general," says study co-author Ferran Alet, a graduate student in MIT's Department of Electrical Engineering and Computer Science and Computer Science and Artificial Intelligence Laboratory (CSAIL). "We were inspired to use AI to find algorithms with curiosity strategies that can adapt to a range of environments."
The researchers created a "meta-learning" algorithm that generated 52,000 exploration algorithms. They found that the top two were entirely new—seemingly too obvious or counterintuitive for a human to have proposed. Both algorithms generated exploration behavior that substantially improved learning in a range of simulated tasks, from navigating a two-dimensional grid based on images to making a robotic ant walk. Because the meta-learning process generates high-level computer code as output, both algorithms can be dissected to peer inside their decision-making processes.
... Four machines searched over 10 hours to find the best algorithms. More than 99 percent were junk, but about a hundred were sensible, high-performing algorithms. Remarkably, the top 16 were both novel and useful, performing as well as, or better than, human-designed algorithms at a range of other virtual tasks, from landing a moon rover to raising a robotic arm and moving an ant-like robot in a physical simulation.
All 16 algorithms shared two basic exploration functions.
In the first, the agent is rewarded for visiting new places where it has a greater chance of making a new kind of move. In the second, the agent is also rewarded for visiting new places, but in a more nuanced way: One neural network learns to predict the future state while a second recalls the past, and then tries to predict the present by predicting the past from the future. If this prediction is erroneous it rewards itself, as it is a sign that it discovered something it didn't know before. The second algorithm was so counterintuitive it took the researchers time to figure out.
More researchers are turning to machine learning to design better machine learning algorithms, a field known as AutoML. At Google, Le and his colleagues recently unveiled a new algorithm-discovery tool called Auto-ML Zero. (Its name is a play on Google's AutoML software for customizing deep net architectures for a given application, and Google DeepMind's Alpha Zero, the program that can learn to play different board games by playing millions of games against itself.)
https://arxiv.org/abs/2003.05325https://lis.csail.mit.edu/wp-content/uploads/effective_interpretable_algorithms_for_curiosity-compressed.pdf