HIGHLIGHTS
- who: Axel Laborieux from the CNRS have published the paper: Synaptic metaplasticity in binarized neural networks, in the Journal: NATURE COMMUNICATIONS NATURE COMMUNICATIONS
- what: Insight, the authors develop a learning strategy using binarized neural_networks to alleviate catastrophic forgetting with strong biological-type constraints: previously presented data can not be stored, nor generated, and the loss function is not task-dependent with weight penalties. Through the example of the progressive learning of datasets, the authors show that the metaplastic binarized neural_network, by contrast, can continue to learn a task when new data becomes available, without seeing . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.