• A bio-inspired approach to mitigate catastrophic forgetting in binarized neural networks

    Neural networks applied in synthetic intelligence (high) are topic to catastrophic forgetting. If they’re taught to acknowledge numbers (MNIST) after which garments (FMNIST), these networks lose the power to acknowledge numbers (backside, left). Because of the metaplastic strategy proposed by the researchers, these networks can successively be taught the 2 duties (backside, proper). Credit score: Laborieux et al. Deep neural networks have achieved extremely promising outcomes on a number of duties, together with picture and textual content classification. Nonetheless, many of those computational strategies are inclined to what’s referred to as catastrophic forgetting, which basically implies that when they’re educated on a brand new job, they have an inclination to…