ECBC: Efficient Convolution via Blocked Columnizing.

Direct convolution methods are now drawing increasing attention as they eliminate the additional storage demand required by indirect convolution algorithms (i.e., the transformed matrix generated by the im2col convolution algorithm). Nevertheless, the direct methods require special input-output tensor formatting, leading to extra time and memory consumption to get the desired data layout. In this article, we show that indirect convolution, if implemented properly, is able to achieve high computation performance with the help of highly optimized subroutines in matrix multiplication while avoid incurring substantial memory overhead. The proposed algorithm is called efficient convolution via blocked columnizing (ECBC). Inspired by the im2col convolution algorithm and the block algorithm of general matrix-to-matrix multiplication, we propose to conduct the convolution computation blockwisely. As a result, the tensor-to-matrix transformation process (e.g., the im2col operation) can also be done in a blockwise manner so that it only requires a small block of memory as small as the data block. Extensive experiments on various platforms and networks validate the effectiveness of ECBC, as well as the superiority of our proposed method against a set of widely used industrial-level convolution algorithms.

Metabolic Profiles of Carbohydrates in Streptococcus thermophilus During pH-Controlled Batch Fermentation.

Revealing the metabolic profiles of carbohydrates with their regulatory genes and metabolites is conducive to understanding their mechanism of utilization in Streptococcus thermophilus MN-ZLW-002 during pH-controlled batch fermentation. Transcriptomics and metabolomics were used to study carbohydrate metabolism. More than 200 unigenes were involved in carbohydrate transport. Of these unigenes, 55 were involved in the phosphotransferase system (PTS), which had higher expression levels than those involved in ABC protein-dependent systems, permeases, and symporters. The expression levels of the genes involved in the carbohydrate transport systems and phosphate transport system were high at the end-lag and end-exponential growth phases, respectively. In addition, 166 differentially expressed genes (DEGs) associated with carbohydrate metabolism were identified. Most genes had their highest expression levels at the end-lag phase. The pfk, ldh, zwf, and E3.2.1.21 genes involved in the glycolytic pathway had higher expression levels at the end-exponential growth phase than the mid-exponential growth phase. The results showed high expression levels of lacZ and galKTM genes and reabsorption of extracellular galactose. S. thermophilus MN-ZLW-002 can metabolize and utilize galactose. Overall, this comprehensive network of carbohydrate metabolism is useful for further studies of the control of glycolytic pathway during the high-density culture of S. thermophilus.

Characterization of Volatile Compounds by SPME-GC-MS During the Ripening of Kedong Sufu, a Typical Chinese Traditional Bacteria-Fermented Soybean Product.

The purpose of the present study was to evaluate the volatile profile of Kedong sufu, which is a typical bacteria-fermented soybean product in China, using solid phase microextraction coupled to gas chromatography and mass spectrometry and to reveal the evolution and diversity of flavor substances for this specialty. A total of 75 compounds were identified, including 35 esters, 4 alcohols, 4 phenols, 4 aldehydes, 7 acids, 10 ketones, and 11 other compounds from sufu samples during ripening. Some volatile compounds increased with ripening time, especially hexadecenoic acid ethyl ester, methoxy acetic acid pentyl ester, benzene propanoic acid ethyl ester, ethyl 9-hexadecenoate, ethyl oleate, ethanol, 3-methyl-1-butanol, 5-methoxy-1-pentanol, and eugenol; these compounds enriched the flavors and provided the typical savory taste of Kedong sufu. PRACTICAL APPLICATION: This research elucidated the formation of flavor substances in sufu. For traditional fermented foods, this study provides a scientific basis for promoting the generation of typical flavor substances and for the precise determination of maturity time.

Transcriptomic and proteomic profiling revealed global changes in Streptococcus thermophilus during pH-controlled batch fermentations.

Understanding global changes of physiological processes at the molecular level during the growth of Streptococcus thermophilus is essential for the rational design of cultivation media and the optimization of bioprocesses. Transcriptomics and proteomics were combined to investigate the global changes at the transcript and protein level during the growth of S. thermophilus. The expression of 1396 genes (FDR ≤ 0.001) and 876 proteins (P < 0.05) changed significantly over time. The most remarkable growth phase dependent changes occurred in the late-lag phase and were related to heterofermentation, glycolysis, peptidoglycan biosynthesis, conversion between amino acids and stress response. The present results could provide theoretical guidance for high-cell-density culture, help design cultivation media, and help attain a high biomass of S. thermophilus.

The nutrient requirements of Lactobacillus rhamnosus GG and their application to fermented milk.

Lactobacillus rhamnosus GG (LGG) performs many physiological functions, but the fermentation time is long when fermented milk is prepared using LGG alone. To shorten the fermentation time, we analyzed the nutrient requirement profiles of LGG. Based on nutrient requirement profiles, we evaluated the effects on the fermentation time, quality, and sensory properties of unmodified cow's milk fermented by LGG alone. According to the consumption and necessary patterns of amino acids and those of purine, pyrimidine, vitamins, metal ions, and nutrients essential to LGG, we selected Cys, Ser, Arg, Pro, Asp, Glu, guanine, uracil, and xanthine with which to supplement milk. Compared with fermented milk prepared using LGG alone in unmodified milk, the fermentation time of supplemented milk was shortened by 5 h. Viable cell counts, titratable acidity, and water-retaining capability of the fermented milk were improved by addition of nutrient supplements. Supplementation with nutrients did not obviously change the sensory and textural characteristics of fermented milk.

Profiles of Streptococcus thermophilus MN-ZLW-002 nutrient requirements in controlled pH batch fermentations.

This study aimed to evaluate the profiles of Streptococcus thermophilus nutrient requirements to guide the design of media for high cell density culturing. The growth kinetics, physiological state, and nutrient requirement profiles of S. thermophilus were analyzed in chemically defined media. The results showed that the intracellular ATP concentration, H+ -ATPase activity, NADH/NAD+ , and NH3 concentrations varied with intracellular pH. The nutrient components with the highest amounts required were Leu and Asp; ascorbic acid and p-amino benzoic acid; K+ and PO43- ; and guanine and uracil. The nutrient components with the largest required ratios were Arg, His, and Met; folic acid, cyanocobalamine, biotin, and nicotinic acid; Ca2+ and Mg2+ ; and guanine and uracil. In this study, different nutrient components were primarily used at different phase. Trp, Tyr, calcium pantothenate, thiamine, guanine, and Mg2+ were mainly used from late-lag to midexponential phase. Met, Pro, Phe, Ala, Gly, nicotinic acid, and riboflavin were mainly used from midexponential to late-exponential phase. The highest bioavailabilities of nutrient components were also found at diverse phase. Met, Leu, Ile, Asn, Glu, Lys, Pro, Gly, riboflavin, nicotinic acid, adenine, uracil, inosine, and Ca2+ had the highest bioavailability from late-lag to midexponential phase. Lactose, Glu, Asp, His, Trp, Cys, Val, Arg, Phe, Ala, Ser, Thr, Tyr, folate and cobalamin, calcium pantothenate, ascorbic acid, thiamine, biotin, p-amino benzoic acid, vitamin B6 , K+ , Mg2+ , guanine, xanthine, and PO43- had the highest bioavailability from midexponential to late-exponential phase. This study elucidated the nutrient requirement profiles with culture time and biomass at various average growth rates during the growth of S. thermophilus. The present results will help to formulate complex media for high cell density cultivation and provide the theoretical basis for S. thermophilus feeding strategies.

Metabolic profiles of cysteine, methionine, glutamate, glutamine, arginine, aspartate, asparagine, alanine and glutathione in Streptococcus thermophilus during pH-controlled batch fermentations.

Elucidating the amino acid (AA) metabolism patterns of Streptococcus thermophilus has important effects on the precise design of nitrogen sources for high-cell-density culture. Transcriptomics and metabolomics were combined to reveal the cysteine, methionine, glutamate, glutamine, arginine, aspartate, asparagine and alanine metabolic pathways in S. thermophilus MN-ZLW-002, including glutathione. The changes in the synthesis, consumption and concentration of AAs and their metabolites, as well as regulatory genes with time were revealed. The metabolism of L-cysteine, L-glutamate, L-aspartate and L-alanine generated some potential functional metabolites. The metabolism of methionine and glutamate generated potential harmful metabolites. S. thermophilus MN-ZLW-002 can synthesize glutathione. Some potential functional metabolites have similar biological functions, indicating that S. thermophilus can resist environmental stresses through multiple mechanisms. The expression of some key genes in synthesis pathway of AA indicated that cysteine, methionine, asparagine, aspartate, arginine and lysine were insufficient or imbalance between nutrient components. The accumulation of large amounts of AA metabolites might be the primary cause of the overconsumption of AAs and influence the growth of S. thermophilus. The present study revealed the metabolic profiles of abovementioned AAs as well as those of regulatory genes and metabolites. These results were beneficial to the precise design of nitrogen sources and regulation of functional metabolites for the high-cell-density culture of S. thermophilus.

Quantized CNN: A Unified Approach to Accelerate and Compress Convolutional Networks.

We are witnessing an explosive development and widespread application of deep neural networks (DNNs) in various fields. However, DNN models, especially a convolutional neural network (CNN), usually involve massive parameters and are computationally expensive, making them extremely dependent on high-performance hardware. This prohibits their further extensions, e.g., applications on mobile devices. In this paper, we present a quantized CNN, a unified approach to accelerate and compress convolutional networks. Guided by minimizing the approximation error of individual layer's response, both fully connected and convolutional layers are carefully quantized. The inference computation can be effectively carried out on the quantized network, with much lower memory and storage consumption. Quantitative evaluation on two publicly available benchmarks demonstrates the promising performance of our approach: with comparable classification accuracy, it achieves 4 to $6 \times $ acceleration and 15 to $20\times $ compression. With our method, accurate image classification can even be directly carried out on mobile devices within 1 s.