| viXra.org > Artificial Intelligence > 2103 |
 |
 |
| viXra.org > Artificial Intelligence > 2103 |
|
|
[9] viXra:2103.0194 [pdf] replaced on 2021-04-01 01:50:20
Authors: Tong Geng, Ang Li, Tianqi Wang, Chunshu Wu, Yanfei Li, Antonino Tumeo, Shuai Che, Steve Reinhardt, Martin Herbordt
Comments: 13 Pages.
In this paper, we propose an architecture design called Ultra-Workload-Balanced-GCN (UWB-GCN) to accelerate graph convolutional network inference. To tackle the major performance bottleneck of workload imbalance, we propose two techniques: dynamic local sharing and dynamic remote switching, both of which rely on hardware flexibility to achieve performance auto-tuning with negligible area or delay overhead. Specifically, UWB-GCN is able to effectively profile the sparse graph pattern while continuously adjusting the workload distribution among parallel processing elements (PEs). After converging, the ideal configuration is reused for the remaining iterations. To the best of our knowledge, this is the first accelerator design targeted to GCNs and the first work that auto-tunes workload balance in accelerator at runtime through hardware, rather than software, approaches. Our methods can achieve near-ideal workload balance in processing sparse matrices. Experimental results show that UWB-GCN can finish the inference of the Nell graph (66K vertices, 266K edges) in 8.1ms, corresponding to 199x, 16x, and 7.5x respectively, compared to the CPU, GPU, and the baseline GCN design without workload rebalancing.
Category: Artificial Intelligence
[8] viXra:2103.0185 [pdf] submitted on 2021-03-29 02:32:14
Authors: Lifeng Gu
Comments: 5 Pages.
Most existing metric learning methods focus on
learning a similarity or distance measure relying
on similar and dissimilar relations between sample pairs. However, pairs of samples cannot be
simply identified as similar or dissimilar in many
real-world applications, e.g., multi-label learning,
label distribution learning. To this end, relation
alignment metric learning (RAML) framework is
proposed to handle the metric learning problem in
those scenarios. But RAML learn a linear metric,
which can’t model complex datasets. Combining with deep learning and RAML framework,
we propose a hierarchical relationship alignment
metric leaning model HRAML, which uses the
concept of relationship alignment to model metric
learning problems under multiple learning tasks,
and makes full use of the consistency between
the sample pair relationship in the feature space
and the sample pair relationship in the label space.
Further we organize several experiment divided
by learning tasks, and verified the better performance of HRAML against many popular methods
and RAML framework.
Category: Artificial Intelligence
[7] viXra:2103.0184 [pdf] submitted on 2021-03-29 02:37:54
Authors: Lifeng Gu
Comments: 9 Pages.
In recent years, representation learning has become the research focus of the machine learning community. Large-scale pre-training neural
networks have become the first step to realize
general intelligence. The key to the success of
neural networks lies in their abstract representation capabilities for data. Several learning fields
are actually discussing how to learn representations and there lacks a unified perspective. We
convert the representation learning problem under multiple tasks into a ranking problem, taking the ranking problem as a unified perspective,
the representation learning under different tasks
is solved by optimizing the approximate NDCG
loss. Experiments under different learning tasks
like classification, retrieval, multi-label learning,
regression, self-supervised learning prove the superiority of approximate NDCG loss. Further, under the self-supervised learning task, the training
data is transformed by data augmentation method
to improve the performance of the approximate
NDCG loss, which proves that the approximate
NDCG loss can make full use of the information
of the unsupervised training data.
Category: Artificial Intelligence
[6] viXra:2103.0174 [pdf] submitted on 2021-03-28 21:30:36
Authors: Lifeng Gu
Comments: 11 Pages.
Science is used to discover the law of world. Machine learning can be used to discover the law of data. In recent years, there are more and more research about interpretability in machine learning community. We hope the machine learning methodsaresafe,interpretable,andtheycanhelpusto find meaningful pattern in data. In this paper, we focus on interpretability of deep representation. We propose a interpretable method of representation based on mutual information, which summarizes the interpretation of representation into three types of information between input data and representation. We further proposed MI-LR module, which can be inserted into the model to estimate the amount of information to explain the model’s representation. Finally, we verify the method through the visualization of the prototype network.
Category: Artificial Intelligence
[5] viXra:2103.0148 [pdf] submitted on 2021-03-23 06:29:02
Authors: Yuanpeng He, Yong Deng
Comments: 32 Pages.
In real life, occurrences of a series of things are supposed to come in an order. Therefore, it is necessary to regard sequence as a crucial factor in managing different kinds of things in fuzzy environment. However, few related researches have been made to provided a reasonable solution to this demand. Therefore, how to measure degree of uncertainty of ordinal fuzzy sets is still an open issue. To address this issue, a novel ordinal relative fuzzy entropy is proposed in this paper taking orders of propositions into consideration in measuring level of uncertainty in fuzzy environment. Compared with previously proposed entropies, effects on degrees of fuzzy uncertainty brought by sequences of sequential propositions are embodied in values of measurement using proposed method in this article. Moreover, some numerical examples are offered to verify the correctness and validity of the proposed entropy.
Category: Artificial Intelligence
[4] viXra:2103.0135 [pdf] submitted on 2021-03-20 20:03:20
Authors: Narayanan Arvind, Saravanan Mugund, Avinash Kumar Singh
Comments: 6 Pages. Presented at Samudramanthan 2021, Indian Institute of Technology Kharagpur
Maritime digital KYC processes are susceptible to various face spoofing attacks. When any
unauthorized person tries to enter in the authentication system by presenting a fraud image
and/or video, it is termed as a spoofing attack. Face anti-spoofing attacks have been typically
approached from texture based models (e.g. Local Binary patterns) combined with machine
learning (e.g. KNN) approaches. The aim of this study is to build a robust face anti-spoofing
system using deep convolutional neural networks for maritime digital KYC processes. The
research is based on analyzing the features of genuine and fake images. We use the freely
available NUAA photograph imposter database for our face anti-spoofing study. The database
has respectively 7500 and 5100 labelled imposter and client face images. We split the dataset
into train and test sets with an 80%-20% split ratio using stratified sampling. 2D convolutional
layers combined with 2D MaxPooling layers followed by Flattening and Dense layers are employed for our deep network architecture. The research is carried out using scikit-learn and keras open-source libraries for python. The training accuracy of the reported model is 100% and the testing accuracy is 99.92%. The accuracy of our present deep learning approach surpasses the accuracy of all the models available in literature.
Category: Artificial Intelligence
[3] viXra:2103.0095 [pdf] submitted on 2021-03-15 20:31:15
Authors: Tanvir Rahman
Comments: 3 Pages.
Pneumonia is a life-threatening infectious disease affecting one or both lungs in humans commonly caused by bacteria called Streptococcus pneumonia. The present study aimed to examine the risk factors for death due to pneumonia in young children. One or more in three deaths in Asia is caused due to pneumonia as reported by World Health Organization (WHO). Chest X-Rays which are used to diagnose pneumonia need expert radiotherapists for evaluation. Thus, developing an automatic system for detecting pneumonia would be beneficial and it can save lots of peoples life and help stopping and curing and controll for treating the disease without any delay particularly in remote areas. Due to the success of deep learning algorithms in analyzing medical images, Convolutional Neural Networks (CNNs) have gained much attention for disease classification. In addition, features learned by pre-trained CNN models on large-scale datasets are much useful in image classification tasks. In this work, we appraise the functionality of pre-trained CNN models utilized as feature-extractors followed by different classifiers for the classification of abnormal and normal chest X-Rays. We analytically determine the optimal CNN model for the purpose. Statistical results obtained demonstrates that pretrained CNN models employed along with supervised classifier algorithms can be very beneficial in analyzing chest X-ray images, specifically to detect
Category: Artificial Intelligence
[2] viXra:2103.0056 [pdf] submitted on 2021-03-11 16:49:40
Authors: Khosnur Alam, Rima Akter
Comments: 8 Pages.
December 31, 2019, a new virus starts spreading in Uhan of China. Nowadays April 2020 the world has seen the worst Pandemic of the century. World health organization tells everybody to test and test but the test is very rare and costly for 3rd world countries. A cheap and easier testing method is now badly required for countries like Bangladesh. So we want to develop a computer-based detection system that can identify Covid-19 patients in a fast and easy way. The chest X-ray image of Covid-19 patients is similar to pneumonia patients. This proposed system can separate Covid-19 X-ray images from pneumonia. The main objective of this research is to develop a system that can detect covid-19 and pneumonia from X-ray images using a deep learning approach.
Category: Artificial Intelligence
[1] viXra:2103.0045 [pdf] submitted on 2021-03-06 21:17:03
Authors: Chandan Maloo, Akhil Kaza
Comments: 4 Pages.
The popularity, cost-effectiveness and ease of buying and selling that marketplaces like Craigslist, Offerup offer to users has been plagued with the rising number of unsolicited spam listings, fraudulent transactions and in some extreme cases law enforcement also needs to be involved. Driven by the need to protect Offerup users from this growing menace, research in spam, fraud listing filtering/detection systems has been increasingly active in the last decade. However, the adaptive nature of Scammers and Fraudsters has often rendered most of these systems ineffective. While several spam detection models have been reported in literature, the reported performance on an out of sample test data shows the room for more improvement. Presented in this research is an improved spam detection model based on Locality Sensitive Hashing algorithm which to the best of our knowledge has received little attention in spam/fraud detection problems. Experimental results show that the proposed model outperforms earlier approaches across a wide range of evaluation metrics inside Offerup.
Category: Artificial Intelligence
| Third-Party Links: |
|