Cross-modal hashing has received widespread attentions on cross-modal retrieval task due to its superior efficiency and low storage cost. However, most existing methods learn binary codes directly from multimedia data, which cannot fully utilize the semantic knowledge of data. Furthermore, they ranking based similarity relevance data points with multi-label. And usually use a relax constraint hash code causes non-negligible quantization loss in optimization. In this paper, method called Deep...

With the rapid development of deep neural networks, cross-modal hashing has made great progress. However, information different types data is asymmetrical, that to say, if resolution an image high enough, it can reproduce almost 100% real-world scenes. text usually carries personal emotion and not objective so we generally think will be much richer than text. Although most existing methods unify semantic feature extraction hash function learning modules for end-to-end learning, they ignore...

QRS complex detection is usually the most important step for automated electrocardiogram (ECG) analysis. In this paper, we present a new approach of detection. The Mexican-hat wavelet and Morlet are used to transform ECG signal, according trait that modulus maxima two coefficients above correspond with R peaks detector unit waves which based on jump sequence in coefficient proposed. Traditional methods employs only one kind perform transformation ECG, whereas proposed method use at same...

There is a huge demand on multilingual tourism information of Japan because the increasing number tourists from foreign countries. Most them may expect typical and stereotyped culture, nature, modern society Japan. However, people different backgrounds, cultures, languages might aspects Japan, as well. In this paper, we analyze these kinds differences cultural preference for We propose machine‐learning‐based method to figure out countries based comparing access logs site in languages. focus...

Email classification methods based on the content general use Vector Space Model. The model is constructed frequency of every independent word appearing in content. Frequency VSM does not take context environment into account, thus feature vectors can accurately represent content, which will result inaccurate classification. This paper presents a new approach to Concept Model using WordNet. In our approach, WordNet we extract high-level information categories during training process by...

Image processing on surfaces has drawn significant interest in recent years, particularly the context of denoising. Salt-and-pepper noise is a special type which randomly sets portion image pixels to minimum or maximum intensity while keeping others unaffected. In this paper, We propose L$_p$TV models triangle meshes recover images corrupted by salt-and-pepper surfaces. establish lower bound for data fitting term recovered image. Motivated property, we corresponding algorithm based proximal...

<title>Abstract</title> The thermal error samples of the machine tool feed system are limited and highly nonlinear, making it difficult for a single prediction model to accurately predict errors under complex operating conditions. To improve accuracy, this paper proposes modeling method based on fusion MCNN-LSTM-SVM tailored small samples, combining deep learning with traditional algorithms. A Multi-Scale Convolutional Neural Network (MCNN) Long Short-Term Memory (LSTM) work in parallel...

With the rapid growth in scale and complexity of large language models (LLMs), costs training inference have risen substantially. Model compression has emerged as a mainstream solution to reduce memory usage computational overhead. This paper presents Group Quantization Sparse Acceleration (\textbf{GQSA}), novel technique tailored for LLMs. Traditional methods typically focus exclusively on either quantization or sparsification, but relying single strategy often results significant...

<p indent=0mm>Cross-modal Hashing has received a lot of attentions in the field cross-modal retrieval due to its high efficiency and low storage cost. Most existing methods learn Hash codes directly from multimodal data cannot fully utilize semantic information data, so distribution consistency low-dimensional features across modalities be guaranteed. To this end, adversarial projection learning based for (APLH) is proposed, which uses training different ensure modalities. On basis, matching...

The image captioning task is about to generate suitable descriptions from images. For this there can be several challenges such as accuracy, fluency and diversity. However are few metrics that cover all these properties while evaluating results of models.In paper we first conduct a comprehensive investigation on contemporary metrics. Motivated by the auto-encoder mechanism research advances word embeddings propose learning based for captioning, which call Intrinsic Image Captioning...

Image registration has become a hot topic in compute vision and shown its variety applications medical areas. The best alignment of images could be estimated by maximization cost function using an iterative optimization, thus optimization is one the most important parts image method. Local optimizations, such as Powell’s method, frequently failed due to existence local minima function, therefore, global methods are required. In this paper, new evolutionary algorithm, group search optimizer...

Automatically evaluating the quality of image captions can be very challenging since human language is quite flexible that there various expressions for same meaning. Most current captioning metrics rely on token level matching between candidate caption and ground truth label sentences. It usually neglects sentence-level information. Motivated by auto-encoder mechanism contrastive representation learning advances, we propose a learning-based metric captioning, which call Intrinsic Image...