•Filter selection and model fine-tuning are integrated into a single end-to-end trainable framework.•Adaptive compression ratio and multi-layer compression.•Good generalization ability.
Channel ...pruning is an important method to speed up CNN model’s inference. Previous filter pruning algorithms regard importance evaluation and model fine-tuning as two independent steps. This paper argues that combining them into a single end-to-end trainable system will lead to better results. We propose an efficient channel selection layer, namely AutoPruner, to find less important filters automatically in a joint training manner. Our AutoPruner takes previous activation responses as an input and generates a true binary index code for pruning. Hence, all the filters corresponding to zero index values can be removed safely after training. By gradually erasing several unimportant filters, we can prevent an excessive drop in model accuracy. Compared with previous state-of-the-art pruning algorithms (including training from scratch), AutoPruner achieves significantly better performance. Furthermore, ablation experiments show that the proposed novel mini-batch pooling and binarization operations are vital for the success of model pruning.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
We propose an efficient and unified framework, namely ThiNet, to simultaneously accelerate and compress CNN models in both training and inference stages. We focus on the filter level pruning, i.e., ...the whole filter would be discarded if it is less important. Our method does not change the original network structure, thus it can be perfectly supported by any off-the-shelf deep learning libraries. We formally establish filter pruning as an optimization problem, and reveal that we need to prune filters based on statistics information computed from its next layer, not the current layer, which differentiates ThiNet from existing methods. Experimental results demonstrate the effectiveness of this strategy, which has advanced the state-of-the-art. We also show the performance of ThiNet on ILSVRC-12 benchmark. ThiNet achieves 3.31 x FLOPs reduction and 16.63× compression on VGG-16, with only 0.52% top-5 accuracy drop. Similar experiments with ResNet-50 reveal that even for a compact network, ThiNet can also reduce more than half of the parameters and FLOPs, at the cost of roughly 1% top-5 accuracy drop. Moreover, the original VGG-16 model can be further pruned into a very small model with only 5.05MB model size, preserving AlexNet level accuracy but showing much stronger generalization ability.
In this paper, two online schemes for an integrated design of fault-tolerant control (FTC) systems with application to Tennessee Eastman (TE) benchmark are proposed. Based on the data-driven design ...of the proposed fault-tolerant architecture whose core is an observer/residual generator based realization of the Youla parameterization of all stabilization controllers, FTC is achieved by an adaptive residual generator for the online identification of the fault diagnosis relevant vectors, and an iterative optimization method for system performance enhancement. The performance and effectiveness of the proposed schemes are demonstrated through the TE benchmark model.
Abstract
DoriC, a database of replication origins, was initially created to present the bacterial oriCs predicted by Ori-Finder or determined by experiments in 2007. DoriC 5.0, an updated database of ...oriC regions in both bacterial and archaeal genomes, was published in the 2013 Nucleic Acids Research database issue. Now, the latest release DoriC 10, a large-scale update of replication origins in prokaryotic genomes including chromosomes and plasmids, has been presented with a completely redesigned user interface, which is freely available at http://tubic.org/doric/ and http://tubic.tju.edu.cn/doric/. In the current release, the database of DoriC has made significant improvements compared with version 5.0 as follows: (i) inclusion of oriCs on more bacterial chromosomes increased from 1633 to 7580; (ii) inclusion of oriCs on more archaeal chromosomes increased from 86 to 226; (iii) inclusion of 1209 plasmid replication origins retrieved from NCBI annotations or predicted by in silico analysis; (iv) inclusion of more replication origin elements on bacterial chromosomes including DnaA-trio motifs. Now, DoriC becomes the most complete and scalable database of replication origins in prokaryotic genomes, and facilitates the studies in large-scale oriC data mining, strand-biased analyses and replication origin predictions.
Deep convolutional neural network models pre-trained for the ImageNet classification task have been successfully adopted to tasks in other domains, such as texture description and object proposal ...generation, but these tasks require annotations for images in the new domain. In this paper, we focus on a novel and challenging task in the pure unsupervised setting: fine-grained image retrieval. Even with image labels, fine-grained images are difficult to classify, letting alone the unsupervised retrieval task. We propose the selective convolutional descriptor aggregation (SCDA) method. The SCDA first localizes the main object in fine-grained images, a step that discards the noisy background and keeps useful deep descriptors. The selected descriptors are then aggregated and the dimensionality is reduced into a short feature vector using the best practices we found. The SCDA is unsupervised, using no image label or bounding box annotation. Experiments on six fine-grained data sets confirm the effectiveness of the SCDA for fine-grained image retrieval. Besides, visualization of the SCDA features shows that they correspond to visual attributes (even subtle ones), which might explain SCDA's high-mean average precision in fine-grained retrieval. Moreover, on general image retrieval data sets, the SCDA achieves comparable retrieval results with the state-of-the-art general image retrieval approaches.
Permeable reactive barrier (PRB) remediation technology has been widely used in the remediation of groundwater contamination. In numerical simulations, neglecting the non-uniform distribution of ...heavy metal contamination along the depth may lead to deviations between simulation results and reality. The distribution of heavy metals in the soil layer around a non-ferrous mining area in Hezhou, Guangxi, southern China was investigated, and it was found that the standard Gaussian function could well describe the non-uniform distribution of heavy metals in the soil layer. A two-dimensional analytical model solved by the finite element method was used to simulate the migration process of heavy metal contamination in the aquifer and PRB. The results show that the uniform distribution of contaminants along the depth ignores the dilution effect, which may underestimate the service life of the PRB and lead to an overly conservative design of the PRB. The breakthrough time of the PRB decreases with the increase of the maximum initial concentration (C
) and the high concentration range (σ), and increases almost linearly with the barrier thickness (L
). An optimal design method for PRB location and thickness is proposed, which can provide a reference for the engineering application of PRB.
Convergence analysis of accelerated first-order methods for convex optimization problems are developed from the point of view of ordinary differential equation solvers. A new dynamical system, called ...Nesterov accelerated gradient (NAG) flow, is derived from the connection between acceleration mechanism and
A
-stability of ODE solvers, and the exponential decay of a tailored Lyapunov function along with the solution trajectory is proved. Numerical discretizations of NAG flow are then considered and convergence rates are established via a discrete Lyapunov function. The proposed differential equation solver approach can not only cover existing accelerated methods, such as FISTA, Güler’s proximal algorithm and Nesterov’s accelerated gradient method, but also produce new algorithms for composite convex optimization that possess accelerated convergence rates. Both the convex and the strongly convex cases are handled in a unified way in our approach.
Full text
Available for:
DOBA, EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, IZUM, KILJ, KISLJ, MFDPS, NLZOH, NUK, ODKLJ, OILJ, PILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UILJ, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
The stock market crash in 2015 had a significant impact on China’s finances, some researchers focus on the causes of the stock market crash, however, its impact on stocks does not seem to have a ...uniformity of trend, perhaps there are different inspirations for investors. Therefore, this paper achieves the goal of exploring the impact of the 2015 stock market crash on Chinese stocks by collecting a total of four sets of stock data on daily and weekly closing prices of Moutai and China Construction and performing ARIMA model analysis on the data to predict the logarithmic returns of the four sets of data. It was found that Moutai seemed to be less affected by the crash than China Construction. Meanwhile, after an in-depth analysis of the coping strategies and business environments of these two stocks, the study highlights that in addition to the financial position and risk management capabilities, a company’s brand and product quality, as well as market resilience, are key factors in coping with market risk. This provides investors with a more comprehensive perspective when making investment decisions.
This study proposes a simple but strong baseline for deep person re-identification (ReID). Deep person ReID has achieved great progress and high performance in recent years. However, many ...state-of-the-art methods design complex network structures and concatenate multi-branch features. In the literature, some effective training tricks briefly appear in several papers or source codes. The present study collects and evaluates these effective training tricks in person ReID. By combining these tricks, the model achieves 94.5% rank-1 and 85.9% mean average precision on Market1501 with only using the global features of ResNet50. The performance surpasses all existing global- and part-based baselines in person ReID. We propose a novel neck structure named as batch normalization neck (BNNeck). BNNeck adds a batch normalization layer after global pooling layer to separate metric and classification losses into two different feature spaces because we observe they are inconsistent in one embedding space. Extended experiments show that BNNeck can boost the baseline, and our baseline can improve the performance of existing state-of-the-art methods. Our codes and models are available at: https://github.com/michuanhaohao/reid-strong-baseline