Suche
Mein Konto
A large-scale model of primary visual cortex can accurately solve multiple visual processing tasks
HBP researchers have trained a large-scale model of the mouse's primary visual cortex to solve visual tasks in an extremely robust manner. The model forms the basis for a new generation of neural network models. Because of their versatility and energy-efficient processing, these models can contribute to advances in neuromorphic computing. Modeling the brain can have a massive impact on artificial intelligence (AI): Since the brain processes images much more energy-efficiently than artificial networks, scientists are taking inspiration from neuroscience to create neural networks that function much more closely to biological ones, saving energy. In this sense, brain-inspired neural...
A large-scale model of primary visual cortex can accurately solve multiple visual processing tasks
HBP researchers have trained a large-scale model of the mouse's primary visual cortex to solve visual tasks in an extremely robust manner. The model forms the basis for a new generation of neural network models. Because of their versatility and energy-efficient processing, these models can contribute to advances in neuromorphic computing.
Modeling the brain can have a massive impact on artificial intelligence (AI): Since the brain processes images much more energy-efficiently than artificial networks, scientists are taking inspiration from neuroscience to create neural networks that function much more closely to biological ones, saving energy.
In this sense, brain-inspired neural networks are likely to have an impact on future technologies by serving as blueprints for visual processing in more power-efficient neuromorphic hardware. Now a study by researchers at the Human Brain Project (HBP) at Graz University of Technology (Austria) showed how a large data-based model can reproduce a range of the brain's visual processing abilities in a versatile and accurate way. The results were published in the journal Science Advances.
Neuroscience eBook
Compilation of the top interviews, articles and news from the last year. Download a copy today
Using the PCP pilot systems at the Jülich Supercomputing Center, developed in collaboration between the HBP and the software company Nvidia, the team analyzed a biologically detailed large-scale model of the primary mouse visual cortex that can solve multiple visual processing tasks. This model provides the greatest integration of anatomical detail and neurophysiological data currently available for visual cortex area V1, which is the first cortical region to receive and process visual information.
The model is built with a different architecture than the deep neural networks used in current AI, and the researchers found that it has interesting advantages in terms of learning speed and visual processing power over models commonly used for visual processing in AI.
The model was able to solve all five visual tasks set by the team with high accuracy. These tasks involved, for example, classifying images of handwritten numbers or recognizing visual changes in a long sequence of images. Remarkably, the virtual model achieved the same high performance as the brain, even when the researchers exposed the model to noise in the images and network that it had not encountered during training.
One reason for the model's superior robustness - or its ability to cope with errors or unexpected inputs such as noise in the images - is that it reproduces several distinctive coding properties of the brain.
The authors have developed a unique tool for studying brain-style visual processing and neural coding and describe their new model as an “unprecedented window into the dynamics of this brain area.”
Source:
Reference:
Chen, G., et al. (2022) A data-driven large-scale model for the primary visual cortex enables brain-like robust and versatile visual processing. Scientific advances. doi.org/10.1126/sciadv.abq7592.
.