This study explores the integration of concept bottleneck models (CBMs) with knowledge distillation (KD) while preserving locality characteristics CBM. Although KD proves effective in model compression, compressed often lack interpretability their decision-making process. We enhance comprehensive explainability by maintaining CBMs’ inherent through our novel approach to distillation. introduce visual (VICO-KD), which transfers both explicit and implicit concepts from teacher student local...

ResearchAndMarkets wrote in their report on May 15, 2018, that up to 1.2 Trillion USD 2017 of products arecounterfeited goods. The estimated this damage globally at 1.82 2020 (RESEARCH ANDMARKETS, 2018). This paper does

Graph convolutional neural networks (GCNNs) have been successfully applied to a wide range of problems, including low-dimensional Euclidean structural domains representing images, videos, and speech high-dimensional non-Euclidean domains, such as social chemical molecular structures. However, in computer vision, the existing GCNNs are not provided with positional information distinguish between graphs new structures; therefore, performance image classification domain represented by arbitrary...

Data augmentation (DA) is a universal technique to reduce overfitting and improve the robustness of machine learning models by increasing quantity variety training dataset. Although data essential in vision tasks, it rarely applied text datasets since less straightforward. Some studies have concerned augmentation, but most them are for majority languages, such as English or French. There been only few on minority e.g., Korean. This study fills gap demonstrating several common methods Korean...

The study aims to examine the determinants affecting adoption of e government services in Vietnam. An extended version unified theory acceptance and use technology model is employed build research model. investigated through asurvey 433 responses from small medium enterprise (SMEs) owners three Southeast provinces Vietnam, analysed using amultinomial logit findings are expected not only assist SMEs their business registration process but contribute quality improvement promotion e-government

ResearchAndMarkets wrote in their report on May 15, 2018, that up to 1.2 Trillion USD 2017 of products are counterfeited goods. The estimated this damage globally at 1.82 2020. This paper does not consider copyright infringement, digital piracy, counterfeiting or fraudulent documents, but rather examines the prevention a technological basis. presence counterfeit European and US markets increase, intervention inspection bodies authorities alone is obviously sufficient, consumers could make...

It has been widely known that 3D shape models are comprehensively parameterized using point cloud and meshes. The particularly is much simpler to handle compared with meshes, it also contains the information of a model. In this paper, we would like introduce our new method generating from set crucial measurements shapes importance positions. order find correspondence between measurements, introduced representing data called slice structure. A Neural Networks-based hierarchical learning model...

Utilizing pre-trained models involves fully or partially using parameters as initialization. In general, configuring a model demands practitioners’ knowledge about problems an exhaustive trial–error experiment according to given task. this paper, we propose tuning trainable layers genetic algorithm on that is fine-tuned single-channel image datasets for classification The dataset comprises images from grayscale and preprocessed audio signals transformed into log-Mel spectrogram. Four...

This study presents the first attempt to integrate Concept Bottleneck Model (CBM) with knowledge distillation (KD) train lightweight and interpretable models. KD is a promising technique for compressing models, but decision-making process of compressed models somewhat inexplainable. By focusing on inherent interpretability CBMs integrating them KD, we provide complete explainability. We propose Visual Concepts Knowledge Distillation (VICO-KD), which transfers both explicit implicit visual...

This study marks the inaugural endeavors to amalgamate concept bottleneck model (CBM) with knowledge distillation (KD) cultivate lightweight and interpretable models. While KD holds promise in compressing models, decision-making process of compressed models often lacks explication. We furnish comprehensive explainability by accentuating inherent interpretability CBM synergizing them KD. introduce visual concepts (VICO-KD), which transmits explicit implicit from teacher student model,...

Multivariate time series anomaly detection is a crucial technology to prevent unexpected errors from causing critical impacts. Effective in such data requires accurately capturing temporal patterns and ensuring the availability of adequate data. This study proposes patch-wise framework for detection. The proposed approach comprises four key components: (i) maintaining continuous features through patching, (ii) incorporating various information by learning channel dependencies adding relative...

Transformer-based time series forecasting models use patch tokens for temporal patterns and variate to learn covariates’ dependencies. While inherently facilitate self-supervised learning, are more suitable linear forecasters as they help mitigate distribution drift. However, the of prohibits masked model pretraining, masking an entire is absurd. To close this gap, we propose LSPatch-T (Long–Short Patch Transfer), a framework that transfers knowledge from short-length into full-length...

This work discusses the challenges of multi-label image classification and presents a novel Efficient Shuffle Net (EffShuffNet) based on convolutional neural network (CNN) architecture to address these challenges. Multi-label is difficult as complexity prediction increases with number labels classes, current multi-model approaches require optimized deep learning models which increase computational costs. The EffShuff block divides input feature map into two parts processes them differently,...

Capsule networks exhibit the potential to enhance computer vision tasks through their utilization of equivariance for capturing spatial relationships. However, broader adoption these has been impeded by computational complexity routing mechanism and shallow backbone model. To address challenges, this paper introduces an innovative hybrid architecture that seamlessly integrates a pretrained model with task-specific capsule head (CapsHead). Our methodology is extensively evaluated across range...

In the face of increasing irregular temperature patterns and climate shifts, need for accurate power consumption prediction is becoming increasingly important to ensure a steady supply electricity. Existing deep learning models have sought improve accuracy but commonly require greater computational demands. this research, on other hand, we introduce DelayNet, lightweight model that maintains efficiency while accommodating extended time sequences. Our DelayNet designed based observation...

This study introduces "shortcut routing," a novel routing mechanism in capsule networks that addresses computational inefficiencies by directly activating global capsules from local capsules, eliminating intermediate layers. An attention-based approach with fuzzy coefficients is also explored for improved efficiency. Experimental results on Mnist, smallnorb, and affNist datasets show comparable classification performance, achieving accuracies of 99.52%, 93.91%, 89.02% respectively. The...

Forecasting, commonly used in econometrics, meteorology, or energy consumption prediction, is the field of study that deals with time series data to predict future trends. Former studies have revealed both traditional statistical models and recent deep learning-based approaches achieved good performance forecasting. In particular, temporal convolutional networks (TCNs) proved their effectiveness several benchmarks. However, presented TCN are too heavy deploy on resource-constrained systems,...

본 논문에서는 분류 목적으로 훈련 된 심층신경망에서 서술자을 추출하여 컨텐츠 기반 이미지 검색의 수행능력을 연구하였다. 잡초 분류를 위해 학습된 VGG 심층신경망 모델을 미세 조정하여 사용한다. 해당 모델의 2개의 전결합층과 전역 평균 풀링으로 부터 이미지의 서술자인 특징벡터를 얻는다. 특징벡터의 차원을 줄이기 위하여 주성분 분석을 적용하고, 오토인코더 네트워크를 개발하여 32, 64, 128, 256차원으로 줄였다. 실험은 전남대학교 데이터 세트에서 ‘종’ 검색을 진행 하였다. 실험에 따르면 훈련된 수집된 특징은 검색에서 우수한 성능을 보인다. 차원 축소 기법 없이 평균정밀도의 평균값은 0.97693을 달성한다. 반면 오토인코더를 통한 설명자 줄이면, 256차원에서 정밀도의 0.97719를

Recent empirical works reveal that visual representation learned by deep neural networks can be successfully used as descriptors for image retrieval. A common technique is to leverage pre-trained models learn ranking losses and fine-tuning with labeled data. However, retrieval systems’ performance significantly decreases when querying images of lower resolution than the training images. This study considered a contrastive learning framework fine-tuned on features extracted from network...

To understand how a Convolutional Neural Network (CNN) model captures the features of pattern to determine which class it belongs to, in this paper, we use Gradient-weighted Class Activation Mapping (Grad-CAM) visualize and analyze well CNN behave on CNU weeds dataset. We apply technique Resnet figure out specific class, what makes get correct/wrong classification, those wrong label images can cause negative effect during training process. In experiment, Grad-CAM highlights important regions...