An Effective Backward Filter Pruning Algorithm Using K1,n Bipartite Graph-Based Clustering and the Decreasing Pruning Rate Approach

Setting a fixed pruning rate and/or specified threshold for pruning filters in convolutional layers has been widely used to reduce the number of parameters required in the convolutional neural networks (CNN) model. However, it fails to fully prune redundant filters for different layers whose redundant filters vary with layers. To overcome this disadvantage, we propose a new backward filter pruning algorithm using a sorted bipartite graph- and binary search-based (SBGBS-based) clustering and decreasing pruning rate (DPR) approach. We first represent each filter of the last layer by a bipartite graph 𝐾1 𝑛, with one root mean set and one 𝑛-weight set, where 𝑛 denotes the number of weights in the filter. Next, according to the accuracy loss tolerance, an SBGBS-based clustering method is used to partition all filters into clusters as maximal as possible. Then, for each cluster, we retain the filter corresponding to the bipartite graph with the median root mean among 𝑛 root means in the cluster, but we discard the other filters in the same cluster. Following the DPR approach, we repeat the above SBGBS-based filtering pruning approach to the backward layer until all layers are processed. Based on the CIFAR-10 and MNIST datasets, the proposed filter pruning algorithm has been deployed into VGG-16, AlexNet, LeNet, and ResNet. With similar accuracy, the thorough experimental results have demonstrated the substantial parameters and floating-point operations reduction merits of our filter pruning algorithm relative to the existing filter pruning methods.

An Effective Backward Filter Pruning Algorithm Using K1,n Bipartite Graph-Based Clustering and the Decreasing Pruning Rate Approach

Setting a fixed pruning rate and/or specified threshold for pruning filters in convolutional layers has been widely used to reduce the number of parameters required in the convolutional neural networks (CNN) model. However, it fails to fully prune redundant filters for different layers whose redundant filters vary with layers. To overcome this disadvantage, we propose a new backward filter pruning algorithm using a sorted bipartite graph- and binary search-based (SBGBS-based) clustering and decreasing pruning rate (DPR) approach. We first represent each filter of the last layer by a bipartite graph 𝐾1 𝑛, with one root mean set and one 𝑛-weight set, where 𝑛 denotes the number of weights in the filter. Next, according to the accuracy loss tolerance, an SBGBS-based clustering method is used to partition all filters into clusters as maximal as possible. Then, for each cluster, we retain the filter corresponding to the bipartite graph with the median root mean among 𝑛 root means in the cluster, but we discard the other filters in the same cluster. Following the DPR approach, we repeat the above SBGBS-based filtering pruning approach to the backward layer until all layers are processed. Based on the CIFAR-10 and MNIST datasets, the proposed filter pruning algorithm has been deployed into VGG-16, AlexNet, LeNet, and ResNet. With similar accuracy, the thorough experimental results have demonstrated the substantial parameters and floating-point operations reduction merits of our filter pruning algorithm relative to the existing filter pruning methods.

GRAPH BASED CLUSTERING WITH CONSTRAINTS AND ACTIVE LEARNING

During the past few years, semi-supervised clustering has emerged as a new interesting direction in machine learning research. In a semi-supervised clustering algorithm, the clustering results can be significantly improved by using side information, which is available or collected from users. There are two main kinds of side information that can be learned in semi-supervised clustering algorithms: the class labels - called seeds or the pairwise constraints. The first semi-supervised clustering was introduced in 2000, and since that, many algorithms have been presented in literature. However, it is not easy to use both types of side information in the same algorithm. To address the problem, this paper proposes a semi-supervised graph based clustering algorithm that tries to use seeds and constraints in the clustering process, called MCSSGC. Moreover, we introduces a simple but efficient active learning method to collect the constraints that can boost the performance of MCSSGC, named KMMFFQS. In order to verify effectiveness of the proposed algorithm, we conducted a series of experiments not only on real data sets from UCI, but also on a document data set applied in an Information Extraction of Vietnamese documents. These obtained results show that the proposed algorithm can significantly improve the clustering process compared to some recent algorithms.