Search

Visual Attention Networks (VAN) with Large Kernel Attention (LKA) modules have been shown to provide remarkable performance, that surpasses Vision Transformers (ViTs), on a range of vision-based tasks. However, the depth-wise convolutional layer in these LKA modules incurs a quadratic increase in the computational and memory footprints with increasing convolutional kernel size. To mitigate these problems and to enable the use of extremely large convolutional kernels in the attention modules of VAN, we propose a family of Large Separable Kernel Attention modules, termed LSKA. LSKA decomposes the 2D convolutional kernel of the depth-wise convolutional layer into cascaded horizontal and vertical 1-D kernels. In contrast to the standard LKA design, the proposed decomposition enables the direct use of the depth-wise convolutional layer with large kernels in the attention module, without requiring any extra blocks. We demonstrate that the proposed LSKA module in VAN can achieve comparable performance with the standard LKA module and incur lower computational complexity and memory footprints. We also find that the proposed LSKA design biases the VAN more toward the shape of the object than the texture with increasing kernel size. Additionally, we benchmark the robustness of the LKA and LSKA in VAN, ViTs, and the recent ConvNeXt on the five corrupted versions of the ImageNet dataset that are largely unexplored in the previous works. Our extensive experimental results show that the proposed LSKA module in VAN provides a significant reduction in computational complexity and memory footprints with increasing kernel size while outperforming ViTs, ConvNeXt, and providing similar performance compared to the LKA module in VAN on object recognition, object detection, semantic segmentation, and robustness tests. [ABSTRACT FROM AUTHOR]

For an efficient detection of intrusion in the network, the system requires sensitive information. This information contains various features depending upon the network. The selection of useful features and relevant information from the dataset has always been difficult for the user. These features bound the performance of the system and also restrict them in making accurate decisions. This irrelevant feature plays a major role in making confusion between the normal data pattern against the attack data pattern. This article proposes a novel Maximum correlation-based mutual information technique for efficient feature selection (MCMIFS) in a data network. The proposed method is utilized with Kernel Extreme Learning Machine (KELM) based multiclass classifier for effective intrusion detection. The performance of the proposed scheme is evaluated using standard intrusion detection datasets with feature ranging from 42 to 155. The hybrid of MCMIFS with KELM reveals the effectiveness of the proposed scheme by improving the accuracy of detection, i.e., 99.97%, decreasing the FPR, i.e., 0.19, by decreasing the computational complexity. The proposed Intrusion Detection Scheme (IDS) results are analyzed and compared with the other existing techniques. • An optimized MCMIFS based feature selection scheme is proposed. • The proposed MCMIFS is utilized with Kernel ELM for multiclass intrusion detection. • The proposed IDS is tested using standard intrusion detection datasets. • Standard performance metrics i.e. Accuracy, Detection Rate and FPR are utilized. • Proposed technique outperforms the existing ones for all used cases. [ABSTRACT FROM AUTHOR]