Very large-scale Deep Neural Networks (DNNs) have achieved remarkable successes in a large variety of computer vision tasks. However, the high computation intensity of DNNs makes it challenging to deploy these models on resource-limited systems. Some studies used low-rank approaches that approximate the filters by low-rank basis to accelerate the testing. Those works directly decomposed the pre-trained DNNs by Low-Rank Approximations (LRA). How to train DNNs toward lower-rank space for more efficient DNNs, however, remains as an open area. To solve the issue, in this work, we propose Force Regularization, which uses attractive forces to enforce filters so as to coordinate more weight information into lower-rank space 1 . We mathematically and empirically verify that after applying our technique, standard LRA methods can reconstruct filters using much lower basis and thus result in faster DNNs. The effectiveness of our approach is comprehensively evaluated in ResNets, AlexNet, and GoogLeNet. In AlexNet, for example, Force Regularization gains 2× speedup on modern GPU without accuracy loss and 4.05× speedup on CPU by paying small accuracy degradation. Moreover, Force Regularization better initializes the low-rank DNNs such that the fine-tuning can converge faster toward higher accuracy. The obtained lower-rank DNNs can be further sparsified, proving that Force Regularization can be integrated with state-of-the-art sparsity-based acceleration methods.
Extensive efforts have been devoted to developing desulfurization catalysts to effectively remove sulfur from fuel. Active phase metals including cobalt, nickel, molybdenum, and tungsten have been extensively used in industry for hydrotreating/hydrodesulfurization catalysts for over 50 years. However, while it is desirable to use inexpensive materials to do the same job, it is a grand challenge. Herein, we report a Fe-based sulfide catalyst that is tuned by zinc with high activity for HDS, which shows an industrial application potential to replace industrial Mo-based catalysts. With an optimal configuration that has a Fe:Zn ratio close to 1:1, the reaction rate constants of the dibenzothiophene (DBT) and 4,6-dimethydibenzothiophene (4,6-DMDBT) HDS are increased by 9.2 and 17.4 times, respectively, in comparison with the sums of those on the monoiron and zinc sulfides. HDS activity for the sterically hindered 4,6-DMDBT on the FeZn sulfide catalyst is even close to that of Co-MoS2. The experimental results indicate that the addition of Zn greatly modifies the electronic properties of iron sulfide by transferring electrons from Zn to Fe, which tunes the d band center to modulate the adsorption behavior of DBT and 4,6-DMDBT. In combination with theoretical calculations, our experiments show that the addition of Zn dramatically tunes the formation of sulfur vacancies. We propose that the formation of sulfur vacancies is the critical factor for designing highly efficient Fe-based sulfide catalysts. This study provides the design principle of low-cost desulfurization catalysts for industrial refinery applications.
BackgroundAcetic acid is routinely generated during lignocelluloses degradation, syngas fermentation, dark hydrogen fermentation and other anaerobic bioprocesses. Acetate stream is commonly regarded as a by-product and detrimental to microbial cell growth. Conversion of acetate into lipids by oleaginous yeasts may be a good choice to turn the by-product into treasure.ResultsTen well-known oleaginous yeasts were evaluated for lipid production on acetate under flask culture conditions. It was found that all of those yeasts could use acetate for microbial lipid production. In particular, Cryptococcus curvatus accumulated lipids up to 73.4 % of its dry cell mass weight. When the culture was held in a 3-L stirred-tank bioreactor, cell mass, lipid content, lipid yield and acetate consumption rate were 8.1 g/L, 49.9 %, 0.15 g/g and 0.64 g/L/h, respectively. The fatty acid compositional profiles of the acetate-derived lipids were similar to those of vegetable oil, suggesting their potential for biodiesel production. Continuous cultivation of C. curvatus was conducted under nitrogen-rich condition at a dilution rate of 0.04 h−1, the maximal lipid content and lipid yield were 56.7 % and 0.18 g/g, respectively. The specific lipid formation rate, lipid content and lipid yield were all higher under nitrogen-rich conditions than those obtained under nitrogen-limited conditions at the same dilution rates. Effective lipid production by C. curvatus was observed on corn stover hydrolysates containing 15.9 g/L acetate.ConclusionsAcetate is an effective carbon source for microbial lipid production by oleaginous yeasts. Continuous cultivation of C. curvatus on acetate was promising for lipid production under both nitrogen-rich and nitrogen-limited conditions. These results provide valuable information for developing and designing more efficient acetate-into-lipids bioprocess.
Micro-expression recognition is still in the preliminary stage, owing much to the numerous difficulties faced in the development of datasets. Since micro-expression is an important affective clue for clinical diagnosis and deceit analysis, much effort has gone into the creation of these datasets for research purposes. There are currently two publicly available spontaneous micro-expression datasets—SMIC and CASME II, both with baseline results released using the widely used dynamic texture descriptor LBP-TOP for feature extraction. Although LBP-TOP is popular and widely used, it is still not compact enough. In this paper, we draw further inspiration from the concept of LBP-TOP that considers three orthogonal planes by proposing two efficient approaches for feature extraction. The compact robust form described by the proposed LBP-Six Intersection Points (SIP) and a super-compact LBP-Three Mean Orthogonal Planes (MOP) not only preserves the essential patterns, but also reduces the redundancy that affects the discriminality of the encoded features. Through a comprehensive set of experiments, we demonstrate the strengths of our approaches in terms of recognition accuracy and efficiency.
Many studies have demonstrated that upregulation of long non-coding RNA (lncRNA) antisense non-coding RNA in the INK4 locus (ANRIL) plays an oncogenic role in various tumors, including nasopharyngeal carcinoma (NPC). The aim of this study is to explore the effect of ANRIL in NPC progression and cisplatin (DDP)-induced cytotoxicity. The results showed that ANRIL was highly expressed and let-7a was downregulated in NPC tissues and cells. Luciferase assay revealed that ANRIL could negatively regulate miR-let-7a expression. ANRIL knockdown inhibited NPC cell proliferation and induced apoptosis, while anti-let-7a reversed these effects. Combination treatment of si-ANRIL and DDP led to a lower viability, a more DNA strand breaks damage and a higher comet tail length compared with any single treatment, whereas let-7a inhibitor abolished these effects. Furthermore, depletion of ANRIL exacerbated DDP-induced cytotoxicity in NPC cells in vivo. Taken together, these data indicated that knockdown of ANRIL represses tumorigenicity and enhances DDP-induced cytotoxicity via regulating microRNA let-7a in NPC cells, providing a promising therapeutic strategy for NPC patients.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.