Search

The increasing interest in filter pruning of convolutional neural networks stems from its inherent ability to effectively compress and accelerate these networks. Currently, filter pruning is mainly divided into two schools: norm-based and relation-based. These methods aim to selectively remove the least important filters according to predefined rules. However, the limitations of these methods lie in the inadequate consideration of filter diversity and the impact of batch normalization (BN) layers on the input of the next layer, which may lead to performance degradation. To address the above limitations of norm-based and similarity-based methods, this study conducts empirical analyses to reveal their drawbacks and subsequently introduces a groundbreaking complex hybrid weighted pruning method. By evaluating the correlations and norms between individual filters, as well as the parameters of the BN layer, our method effectively identifies and prunes the most redundant filters in a robust manner, thereby avoiding significant decreases in network performance. We conducted comprehensive and direct pruning experiments on different depths of ResNet using publicly available image classification datasets, ImageNet and CIFAR-10. The results demonstrate the significant efficacy of our approach. In particular, when applied to the ResNet-50 on the ImageNet dataset, achieves a significant reduction of 53.5% in floating-point operations, with a performance loss of only 0.6%. [ABSTRACT FROM AUTHOR]

Drug addiction is considered an aberrant form of learning, and drug-associated memories evoked by the presence of associated stimuli (drug context or drug-related cues) contribute to recurrent craving and reinstatement. Epigenetic changes mediated by DNA methyltransferase (DNMT) have been implicated in the reconsolidation of fear memory. Here, we investigated the role of DNMT activity in the reconsolidation of cocaine-associated memories. Rats were trained over 10 days to intravenously self-administer cocaine by nosepokes. Each injection was paired with a light/tone conditioned stimulus (CS). After acquisition of stable self-administration behaviour, rats underwent nosepoke extinction (10 d) followed by cue-induced reactivation and subsequent cue-induced and cocaine-priming + cue-induced reinstatement tests or subsequently tested to assess the strength of the cocaine-associated cue as a conditioned reinforcer to drive cocaine seeking behaviour. Bilateral intra-basolateral amygdala (BLA) infusion of the DNMT inhibitor5-azacytidine (5-AZA, 1 μg per side) immediately following reactivation decreased subsequent reinstatement induced by cues or cocaine priming as well as cue-maintained cocaine-seeking behaviour. In contrast, delayed intra-BLA infusion of 5-AZA 6 h after reactivation or 5-AZA infusion without reactivation had no effect on subsequent cue-induced reinstatement. These findings indicate that memory reconsolidation for a cocaine-paired stimulus depends critically on DNMT activity in the BLA. [ABSTRACT FROM AUTHOR]