DAda-NC: A Decoupled Adaptive Online Training Algorithm for Deep Learning Under Non-convex Conditions

2022 ◽  
pp. 76-88
Author(s):  
Yangfan Zhou ◽  
Cheng Cheng ◽  
Jiang Li ◽  
Yafei Ji ◽  
Haoyuan Wang ◽  
...  
Keyword(s):  
Deep Learning ◽  
Online Training ◽  
Training Algorithm

scholarly journals One-Stage Logo Detection Framework using AdaBoost Resnet50 Backbone

2019 ◽  
Vol 8 (3) ◽  
pp. 451-457
Keyword(s):  
Deep Learning ◽  
Gradient Descent ◽  
Training Algorithm ◽  
The Past ◽  
One Stage ◽  
Input Dataset ◽  
Potential Applications ◽  
Or Organization ◽  
Deep Learning Model ◽  
Wide Potential

Logo is an important asset as it is designed to express identity or character of the company or organization that owns the logo. The advent of deep learning methods and proliferated of logo images sample dataset in the past decade has made automated logo detection from digital images or video an interesting computer vision problem with wide potential applications. This paper presents a novel one-stage logo detector framework in which the backbone of the proposed logo detector is a deep learning model which is trained supervisedly using gradient descent training algorithm and the target logo classes as input dataset. The experiment results showed that AdaBoost Resnet50 (0.58 MAP) as the logo detector backbone outperforms Resnet50 (0.56 MAP), VGG19 (0.32 MAP), and AdaBoost VGG19 (0.56 MAP).

scholarly journals OTNEL: A Distributed Online Deep Learning Semantic Annotation Methodology

2020 ◽  
Vol 4 (4) ◽  
pp. 31
Author(s):  
Christos Makris ◽  
Michael Angelos Simos
Keyword(s):  
Deep Learning ◽  
Semantic Annotation ◽  
Online Training ◽  
Graph Representation ◽  
Entity Linking ◽  
Architecture Model ◽  
Named Entity ◽  
Unstructured Text ◽  
Entity Disambiguation ◽  
Named Entity Disambiguation

Semantic representation of unstructured text is crucial in modern artificial intelligence and information retrieval applications. The semantic information extraction process from an unstructured text fragment to a corresponding representation from a concept ontology is known as named entity disambiguation. In this work, we introduce a distributed, supervised deep learning methodology employing a long short-term memory-based deep learning architecture model for entity linking with Wikipedia. In the context of a frequently changing online world, we introduce and study the domain of online training named entity disambiguation, featuring on-the-fly adaptation to underlying knowledge changes. Our novel methodology evaluates polysemous anchor mentions with sense compatibility based on thematic segmentation of the Wikipedia knowledge graph representation. We aim at both robust performance and high entity-linking accuracy results. The introduced modeling process efficiently addresses conceptualization, formalization, and computational challenges for the online training entity-linking task. The novel online training concept can be exploited for wider adoption, as it is considerably beneficial for targeted topic, online global context consensus for entity disambiguation.

scholarly journals A Hybrid Cascade Neuro–Fuzzy Network with Pools of Extended Neo–Fuzzy Neurons and its Deep Learning

2019 ◽  
Vol 29 (3) ◽  
pp. 477-488 ◽  
Author(s):  
Yevgeniy V. Bodyanskiy ◽  
Oleksii K. Tyshchenko
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Data Streams ◽  
Online Training ◽  
Training Procedure ◽  
Neuro Fuzzy ◽  
Stationary Objects ◽  
Set Up ◽  
Cascade Neural Network ◽  
Fuzzy Network

Abstract This research contribution instantiates a framework of a hybrid cascade neural network based on the application of a specific sort of neo-fuzzy elements and a new peculiar adaptive training rule. The main trait of the offered system is its competence to continue intensifying its cascades until the required accuracy is gained. A distinctive rapid training procedure is also covered for this case that offers the possibility to operate with non-stationary data streams in an attempt to provide online training of multiple parametric variables. A new training criterion is examined for handling non-stationary objects. Additionally, there is always an occasion to set up (increase) the inference order and the number of membership relations inside the extended neo-fuzzy neuron.

A novel path planning method for biomimetic robot based on deep learning

2016 ◽  
Vol 36 (2) ◽  
pp. 186-191 ◽  
Author(s):  
Yang Lu ◽  
Shujuan Yi ◽  
Yurong Liu ◽  
Yuling Ji
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Path Planning ◽  
Training Algorithm ◽  
Robot Path Planning ◽  
Content Type ◽  
Max Pooling ◽  
Biomimetic Robot ◽  
Planning Task ◽  
Robot Path

Purpose This paper aims to design a multi-layer convolutional neural network (CNN) to solve biomimetic robot path planning problem. Design/methodology/approach At first, the convolution kernel with different scales can be obtained by using the sparse auto encoder training algorithm; the parameter of the hidden layer is a series of convolutional kernel, and the authors use these kernels to extract first-layer features. Then, the authors get the second-layer features through the max-pooling operators, which improve the invariance of the features. Finally, the authors use fully connected layers of neural networks to accomplish the path planning task. Findings The NAO biomimetic robot respond quickly and correctly to the dynamic environment. The simulation experiments show that the deep neural network outperforms in dynamic and static environment than the conventional method. Originality/value A new method of deep learning based biomimetic robot path planning is proposed. The authors designed a multi-layer CNN which includes max-pooling layer and convolutional kernel. Then, the first and second layers features can be extracted by these kernels. Finally, the authors use the sparse auto encoder training algorithm to train the CNN so as to accomplish the path planning task of NAO robot.

scholarly journals On Intra-Class Variance for Deep Learning of Classifiers

2019 ◽  
Vol 44 (3) ◽  
pp. 285-301
Author(s):  
Rafał Pilarczyk ◽  
Władysław Skarbek
Keyword(s):  
Deep Learning ◽  
Stochastic Gradient ◽  
Training Algorithm ◽  
Class Membership ◽  
Target Class ◽  
Membership Probability ◽  
Novel Technique ◽  
Training Stage ◽  
Comparable Accuracy ◽  
Trained Network

Abstract A novel technique for deep learning of image classifiers is presented. The learned CNN models higher offer better separation of deep features (also known as embedded vectors) measured by Euclidean proximity and also no deterioration of the classification results by class membership probability. The latter feature can be used for enhancing image classifiers having the classes at the model’s exploiting stage different from from classes during the training stage. While the Shannon information of SoftMax probability for target class is extended for mini-batch by the intra-class variance, the trained network itself is extended by the Hadamard layer with the parameters representing the class centers. Contrary to the existing solutions, this extra neural layer enables interfacing of the training algorithm to the standard stochastic gradient optimizers, e.g. AdaM algorithm. Moreover, this approach makes the computed centroids immediately adapting to the updating embedded vectors and finally getting the comparable accuracy in less epochs.

Sign in / Sign up

Export Citation Format

Share Document