The recent deep cross-modal hashing (DCMH) has achieved superior performance in effective and efficient cross-modal retrieval and thus has drawn increasing attention. Nevertheless, there are still... Show moreThe recent deep cross-modal hashing (DCMH) has achieved superior performance in effective and efficient cross-modal retrieval and thus has drawn increasing attention. Nevertheless, there are still two limitations for most existing DCMH methods: (1) single labels are usually leveraged to measure the semantic similarity of cross-modal pairwise instances while neglecting that many cross-modal datasets contain abundant semantic information among multi-labels. (2) several DCMH methods utilized the multi-labels to supervise the learning of hash functions. Nevertheless, the feature space of multilabels suffers the weakness of sparse, resulting in sub-optimization for the hash functions learning. Thus, this paper proposed a multi-label modality enhanced attention-based self-supervised deep cross-modal hashing (MMACH) framework. Specifically, a multi-label modality enhanced attention module is designed to integrate the significant features from cross-modal data into multi-labels feature representations, aiming to improve its completion. Moreover, a multi-label cross-modal triplet loss is defined based on the criterion that the feature representations of cross-modal pairwise instances with more common categories should preserve higher semantic similarity than other instances. To the best of our knowledge, the multi-label cross-modal triplet loss is the first time designed for cross-modal retrieval. Extensive experiments on four multi-label cross-modal datasets demonstrate the effectiveness and efficiency of our proposed MMACH. Moreover, the MMACH also achieved superior performance and outperformed several state-of-the-art methods on the task of cross-modal retrieval. The source code of MMACH is available at https://github.com/SWU-CS-MediaLab/MMACH. (c) 2021 Elsevier B.V. All rights reserved. Show less
With ICT as an engine of economic growth, the mapping and monitoring of catching-up behaviour has high policy relevance for countries in trying to follow pace. Using a novel way of measuring and... Show moreWith ICT as an engine of economic growth, the mapping and monitoring of catching-up behaviour has high policy relevance for countries in trying to follow pace. Using a novel way of measuring and combining two sequential S-shaped catching-up stages (imitation and self-propagation), we compare the catching-up dynamics of three European and three Asian countries, taking US as the first mover. Based on logistic model, take-off time and growth rate parameters are estimated for the two distinct stages (imitation and self-propagation) whereby the combination of information on both curves reveals the specific catching-up pattern of the considered countries. The result shows that China and Korea start late but grow very fast in both stages.The caveat from such a dynamic is a short growth timer, leading to a lack of incubation and consequently an absence of potential to obtain sustainable growth. Japan’s catching-up pattern is much more similar to that of European countries. In general, Germany, Great Britain, Japan, Korea and China are presenting a healthy catching-up pattern, but they have several ‘unhealthy’ subfields which need more attention to be paid. Show less
Brassica rapa displays enormous morphological diversity, with leafy vegetables, turnips and oil crops. Turnips (Brassica rapa subsp. rapa) represent one of the morphotypes, which form tubers... Show moreBrassica rapa displays enormous morphological diversity, with leafy vegetables, turnips and oil crops. Turnips (Brassica rapa subsp. rapa) represent one of the morphotypes, which form tubers and can be used to study the genetics underlying storage organ formation. In the present study we investigated several characteristics of an extensive turnip collection comprising 56 accessions from both Asia (mainly Japanese origin) and Europe. Population structure was calculated using data from 280 evenly distributed SNP markers over 56 turnip accessions. We studied the anatomy of turnip tubers and measured carbohydrate composition of the mature turnip tubers of a subset of the collection. The variation in 16 leaf traits, 12 tuber traits and flowering time was evaluated in five independent experiments for the entire collection. The effect of vernalization on flowering and tuber formation was also investigated. SNP marker profiling basically divided the turnip accessions into two subpopulations, with admixture, generally corresponding with geographical origin (Europe or Asia). The enlarged turnip tuber consists of both hypocotyl and root tissue, but the proportion of the two tissues differs between accessions. The ratio of sucrose to fructose and glucose differed among accessions, while generally starch content was low. The evaluated traits segregated in both subpopulations, with leaf shape, tuber colour and number of shoots per tuber explaining most variation between the two subpopulations. Vernalization resulted in reduced flowering time and smaller tubers for the Asian turnips whereas the European turnips were less affected by vernalization. Show less
Bachert, C.; Steen, K. van; Zhang, N.; Holtappels, G.; Cattaert, T.; Maus, B.; ... ; Howarth, P. 2012
