1,376 research outputs found
PopNet: a Pop Culture Knowledge Association Network for Supporting Creative Connections
Full text linkPop culture is a pervasive and important aspect of communication and
self-expression. When people wish to communicate using pop culture references,
they need to find connections between their message and the things, people,
location and actions of a movie, tv series, or other pop culture domain.
However, finding an appropriate match from memory is challenging and search
engines are not specific enough to the task. Often domain-specific knowledge
graphs provide the structure, specificity and search capabilities that people
need. We introduce PopNet - a Pop Culture Knowledge Association Network
automatically created from plain text using state-of-the art NLP methods to
extract entities and actions from text summaries of movies and tv shows. The
interface allows people to browse and search the entries to find connections.
We conduct a study showing that this system is accurate and helpful for finding
multiple connections between a message and a pop culture domain
UniNeXt: Exploring A Unified Architecture for Vision Recognition
Full text linkVision Transformers have shown great potential in computer vision tasks. Most
recent works have focused on elaborating the spatial token mixer for
performance gains. However, we observe that a well-designed general
architecture can significantly improve the performance of the entire backbone,
regardless of which spatial token mixer is equipped. In this paper, we propose
UniNeXt, an improved general architecture for the vision backbone. To verify
its effectiveness, we instantiate the spatial token mixer with various typical
and modern designs, including both convolution and attention modules. Compared
with the architecture in which they are first proposed, our UniNeXt
architecture can steadily boost the performance of all the spatial token
mixers, and narrows the performance gap among them. Surprisingly, our UniNeXt
equipped with naive local window attention even outperforms the previous
state-of-the-art. Interestingly, the ranking of these spatial token mixers also
changes under our UniNeXt, suggesting that an excellent spatial token mixer may
be stifled due to a suboptimal general architecture, which further shows the
importance of the study on the general architecture of vision backbone. All
models and codes will be publicly available
Acute suppurative thyroiditis with thyroid metastasis from oesophageal cancer
Get PDFNot required for Clinical Vignette
Sun sensor design and test of a micro satellite
Get PDFAccording to the requirement of small satellite, this paper designed a digital sun sensor which diaphragm is a V-shaped cross-section structure. Using Position Sensitive Detector (PSD) as the light detector, we designed the V-shaped cross-section structure based on the pinhole imaging principle. The sun sensor realized the accurate calculation for two axis sun angle of the sun sensor. The mechanical test, thermal test and testing of the sun sensor are designed and carried out. The mechanical test and thermal test results verify the stability of the sun sensor. Testing result shows that the detection angle can reach (120°)×(120°), and the attitude determination accuracy is better than 6” in the entire viewing field. The mass, volume and power consumption of the sun sensor is 0.177 kg, 78 mm×77 mm×21 mm and 0.25 W. The sun sensor has low power consumption, large viewing angle and high precision characteristics, which realized the sun sensor the miniaturization and meet the requirements of the micro satellite. Its performance has been verified in orbit
RegionBLIP: A Unified Multi-modal Pre-training Framework for Holistic and Regional Comprehension
Full text linkIn this work, we investigate extending the comprehension of Multi-modal Large
Language Models (MLLMs) to regional objects. To this end, we propose to extract
features corresponding to regional objects as soft prompts for LLM, which
provides a straightforward and scalable approach and eliminates the need for
LLM fine-tuning. To effectively extract regional features from regular image
features and irregular point cloud features, we present a novel and unified
position-assisted feature extraction module. Furthermore, training an MLLM from
scratch is highly time-consuming. Thus, we propose incrementally extending
existing pre-trained MLLMs to comprehend more modalities and the regional
objects of those modalities. Specifically, we freeze the Q-Former from BLIP-2,
an impressive MLLM, and optimize the modality-specific Lora parameters in
Q-Former and LLM for each newly introduced modality. The freezing of the
Q-Former eliminates the need for extensive pre-training on massive image-text
data. The freezed Q-Former pre-trained from massive image-text data is also
beneficial for the pre-training on image-region-text data. We name our
framework RegionBLIP. We pre-train RegionBLIP on image-region-text,
point-cloud-text, and point-cloud-region-text data. Experimental results verify
that \Ours{} can preserve the image comprehension capability of BILP-2 and
further gain a comprehension of the newly introduced point cloud modality and
regional objects. The Data, Code, and Pre-trained models will be available at
https://github.com/mightyzau/RegionBLIP
UNDERSTANDING INVESTMENT INTENTION TOWARDS P2P LENDING: AN EMPIRICAL STUDY
Get PDFP2P lending is an innovation of micro-financial operation pattern, which is mainly used to meet the petty loan and investment demands of small and micro businesses and individuals. Given the rapid development of P2P market, there is a pressing need to understand lenders’ initial investment intentions in P2P platform. Although there are some studies exploring the factors explaining P2P lenders’ investment intentions, none of research has been reported from the perspective of the platform. This study extended technology acceptance model with perceived risk and initial trust as a theoretical framework to examine the roles of individual factors and platform factors in determining P2P lenders’ initial investment intentions. This study suggests that risk appetite, trust propensity, perceived ease of use, perceived security assurance, perceived privacy protection, perceived reputation, third-party certification, perceived risk and initial trust together provide a strong explanation for initial investment intention in P2P lending. The finding of this research provided a theoretical foundation for future academic studies as well as practical guidance for rapid development of P2P platform
A high-resolution map of reactive nitrogen inputs to China
Get PDFTo feed an increasingly affluent population, reactive nitrogen (Nr) inputs to China’s lands and waters have substantially increased over the past century. Today, China’s Nr emissions account for over one third of global total emissions, leading to serious environmental pollution and health damages. Quantifying the spatial variability of Nr inputs is crucial for the identification of intervention points to mitigate Nr pollution, which, however, is not well known. Here, we present a database describing Nr inputs to China for the year 2017 with a 1 km × 1 km resolution, considering land use and Nr sources, compiled by using the CHANS model. Results show that the North China Plain, the Sichuan Basin and the Middle-Lower Yangtze River Plain are hotspots of Nr inputs, where per hectare Nr input is an order of magnitude higher than that in other regions. Cropland and surface water bodies receive much higher Nr inputs than other land use types. This unique database will provide basic data for research on environmental health and global change modelling
Data Pruning via Moving-one-Sample-out
Full text linkIn this paper, we propose a novel data-pruning approach called
moving-one-sample-out (MoSo), which aims to identify and remove the least
informative samples from the training set. The core insight behind MoSo is to
determine the importance of each sample by assessing its impact on the optimal
empirical risk. This is achieved by measuring the extent to which the empirical
risk changes when a particular sample is excluded from the training set.
Instead of using the computationally expensive leaving-one-out-retraining
procedure, we propose an efficient first-order approximator that only requires
gradient information from different training stages. The key idea behind our
approximation is that samples with gradients that are consistently aligned with
the average gradient of the training set are more informative and should
receive higher scores, which could be intuitively understood as follows: if the
gradient from a specific sample is consistent with the average gradient vector,
it implies that optimizing the network using the sample will yield a similar
effect on all remaining samples. Experimental results demonstrate that MoSo
effectively mitigates severe performance degradation at high pruning ratios and
achieves satisfactory performance across various settings.Comment: Accepted by the Thirty-seventh Conference on Neural Information
Processing Systems (NeurIPS 2023
The relationship between BSP mRNA expression and 25(OH)D/OPG in peripheral blood of newly diagnosed T2DM patients with different bone mass
Get PDFIntroduction: The objective of the study was to detect the levels of osteoprotegerin (OPG) and 25-hydroxyvitamin D [25(OH)D], as well as the expression of bone sialoprotein (BSP) mRNA, in the peripheral blood of patients with newly diagnosed type 2 diabetes mellitus (T2DM) under different bone mass conditions, and to explore its role and significance in the development process of T2DM combined with osteoporosis (OP). Material and methods: A total of 225 patients hospitalised in the Endocrinology Department and General Department from May 2017 to May 2018 were enrolled and categorised into five groups: the pure T2DM group (group A, 45 patients), the bone mass reduction group (group B, 45 patients), the T2DM + bone mass reduction group (group C, 45 patients), the OP group (group D, 45 patients), and the T2DM + OP group (group E, 45 patients); meanwhile, age-matched healthy subjects undergoing physical examination in our hospital were collected as the normal control group (group NC, 45 cases). Logistic regression analysis was used to analyse the influencing factors of bone mass in patients with T2DM. Results: Compared with group B, the expression levels of glycated haemoglobin (HbA1c), 25(OH)D, N-terminal propeptide of type I procollagen (PINP), fasting plasma glucose (FPG), fasting plasma insulin (FINS), high-density lipoprotein cholesterol (HDL-C), and BSP mRNA were significantly increased while OPG and b-collagen degradation products (b-CTX) were significantly decreased in group A. Conclusion: The expression of BSP mRNA and the decrease of 25(OH)D and OPG in peripheral blood may participate in the development of diabetes and osteoporosis
PodReels: Human-AI Co-Creation of Video Podcast Teasers
Full text linkVideo podcast teasers are short videos that can be shared on social media
platforms to capture interest in the full episodes of a video podcast. These
teasers enable long-form podcasters to reach new audiences and gain new
followers. However, creating a compelling teaser from an hour-long episode is
challenging. Selecting interesting clips requires significant mental effort;
editing the chosen clips into a cohesive, well-produced teaser is
time-consuming. To support the creation of video podcast teasers, we first
investigate what makes a good teaser. We combine insights from both audience
comments and creator interviews to determine a set of essential ingredients. We
also identify a common workflow shared by creators during the process. Based on
these findings, we introduce a human-AI co-creative tool called PodReels to
assist video podcasters in creating teasers. Our user study shows that PodReels
significantly reduces creators' mental demand and improves their efficiency in
producing video podcast teasers
- …
