A Strategic Weight Refinement Maneuver for Convolutional Neural Networks DOI

Patrick Allan Sharma,

Adarsh Sharma,

Dinesh Kumar

et al.

2022 International Joint Conference on Neural Networks (IJCNN), Journal Year: 2021, Volume and Issue: unknown, P. 1 - 7

Published: July 18, 2021

Stochastic Gradient Descent algorithms (SGD) remain a popular optimizer for deep learning networks and have been increasingly used in applications involving large datasets producing promising results. SGD approximates the gradient on small subset of training examples, randomly selected every iteration during network training. This randomness leads to selection an inconsistent order examples resulting ambiguous values solve cost function. paper applies Guided (GSGD) - variant neural networks. GSGD minimizes loss maximizes classification accuracy by overcoming data SGDs. It temporarily bypasses instances computation weight update, leading better convergence at rate $O(\frac{1}{\rho T-})$ . Previously, has only shallow like logistic regression. We try incorporate Convolutional Neural Networks (CNNs) evaluate comparison with same trained test our approach benchmark image datasets. Our baseline results show improves up 3% standard CNNs.

Language: Английский

SMOTified-GAN for Class Imbalanced Pattern Classification Problems DOI Creative Commons
Anurag Sharma, Prabhat Kumar Singh, Rohitash Chandra

et al.

IEEE Access, Journal Year: 2022, Volume and Issue: 10, P. 30655 - 30665

Published: Jan. 1, 2022

Class imbalance in a dataset is major problem for classifiers that results poor prediction with high true positive rate (TPR) but low negative (TNR) majority training dataset. Generally, the pre-processing technique of oversampling minority class(es) are used to overcome this deficiency. Our focus on using hybridization Generative Adversarial Network (GAN) and Synthetic Minority Over-Sampling Technique (SMOTE) address class imbalanced problems. We propose novel two-phase approach involving knowledge transfer has synergy SMOTE GAN. The unrealistic or overgeneralized samples transformed into realistic distribution data by GAN where there not enough available process them itself effectively. named it SMOTified-GAN as works pre-sampled produced rather than randomly generating itself. experimental prove sample quality been improved variety tested benchmark datasets. Its performance up 9\% from next best algorithm F1-score measurements. time complexity also reasonable which around $O(N^2d^2T)$ sequential algorithm.

Language: Английский

Citations

73
Classification and Analysis of Optimization Techniques for Integrated Energy Systems Utilizing Renewable Energy Sources: A Review for CHP and CCHP Systems DOI Open Access
Mohammad Ali Bagherian, Kamyar Mehranzamir, Amin Beiranvand Pour

et al.

Processes, Journal Year: 2021, Volume and Issue: 9(2), P. 339 - 339

Published: Feb. 12, 2021

Energy generation and its utilization is bound to increase in the following years resulting accelerating depletion of fossil fuels, consequently, undeniable damages our environment. Over past decade, despite significant efforts renewable energy realization developments for electricity generation, carbon dioxide emissions have been increasing rapidly. This due fact that there a need go beyond power sector target an integrated manner. In this regard, systems integration concept looks into how different systems, or forms, can connect together order provide value consumers producers. Cogeneration trigeneration are two most well established technologies capable producing three forms simultaneously within single system. Integrated make very strong proposition since it results saving, fuel diversification, supply cleaner energy. Optimization such be carried out using several techniques with regards objective functions. study, variety optimization methods provides possibility performance improvements, without presence constraints, demonstrated, pinpointing characteristics each method along detailed statistical reports. context, classified primary groups including unconstrained constrained techniques. Further, potential applications evolutionary computing Systems (IESs), particularly Combined Heat Power (CHP) Cooling, Heating, (CCHP), utilizing sources grasped reviewed thoroughly. It was illustrated employment classical fading out, replacing Amongst modern heuristic algorithms, has contributed more certain application; while Genetic Algorithm (GA) favored thermoeconomic optimization, Particle Swarm (PSO) mostly applied economic improvements. Given mathematical nature constraint satisfaction property Mixed-Integer Linear Programming (MILP), gaining prominence scheduling systems.

Language: Английский

Citations

49
A Modified High-Order Neural Network with Smoothing L1 Regularization and Momentum Terms DOI
Khidir Shaib Mohamed,

Ibrhim. M. A. Suliman,

Mahmoud I. Alfeel

et al.

Signal Image and Video Processing, Journal Year: 2025, Volume and Issue: 19(5)

Published: March 24, 2025

Language: Английский

Citations

0
Dynamic mode selection and resource allocation approach for 5G-vehicle-to-everything (V2X) communication using asynchronous federated deep reinforcement learning method DOI
Iftikhar Rasheed

Vehicular Communications, Journal Year: 2022, Volume and Issue: 38, P. 100532 - 100532

Published: Oct. 1, 2022

Language: Английский

Citations

12
SGD method for entropy error function with smoothing $$l_0$$ regularization for neural networks DOI

Trong-Tuan Nguyen,

Van-Dat Thang,

Nguyen Van Thin

et al.

Applied Intelligence, Journal Year: 2024, Volume and Issue: 54(13-14), P. 7213 - 7228

Published: June 3, 2024

Language: Английский

Citations

2
3D ground penetrating radar road underground target identification algorithm using time-frequency statistical features of data DOI
Fanruo Li, Feng Yang, Xu Qiao

et al.

NDT & E International, Journal Year: 2023, Volume and Issue: 137, P. 102860 - 102860

Published: April 28, 2023

Language: Английский

Citations

6
A Guided Neural Network Approach to Predict Early Readmission of Diabetic Patients DOI Creative Commons
Avishek Anishkar Ram, Zain Ali,

Vandana Krishna

et al.

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 47527 - 47538

Published: Jan. 1, 2023

Diabetes is a major chronic health problem affecting millions globally. Effective diabetes management can reduce the risk of hospital readmission and associated financial losses for both healthcare system insurance companies. Hospital high-priority quality measure that reflects inadequacies in also increase costs negatively influence hospitals' reputation. Predicting readmissions early stages prompts great attention to patients with high readmission. There has been some attempt applying machine learning predictive models such as ensemble Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM) Artificial Neural Networks (ANN) correctly identify if happen within 30 days (< days) or it may never happens after ( $\ge 30$ days). We are proposing new method applied ANN guide through its gradient descent optimizers by realizing consistent vs inconsistent data every batch. Our results show there up 1.5% improvement classification accuracies 2-class 3-class variations experimented benchmark dataset when using guided optimizer train opposed standard optimizer. Guided able achieve better error convergence than ANN.

Language: Английский

Citations

6
Using an innovative method for breast cancer diagnosis based on Extreme Gradient Boost optimized by Simplified Memory Bounded A* DOI
Tao Chen,

Xichao Zhou,

Guan Wang

et al.

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 87, P. 105450 - 105450

Published: Oct. 1, 2023

Language: Английский

Citations

6
A Propagation Study of LoRa P2P Links for IoT Applications: The Case of Near-Surface Measurements over Semitropical Rivers DOI Creative Commons
A. Gutiérrez-Gómez, Víctor Rangel, R.M. Edwards

et al.

Sensors, Journal Year: 2021, Volume and Issue: 21(20), P. 6872 - 6872

Published: Oct. 16, 2021

Internet of Things (IoT) radio networks are becoming popular in several scenarios for short-range applications (e.g., wearables and home security) medium-range shipping container tracking autonomous farming). They have also been proposed water monitoring flood warning systems. IoT communications may use long range (LoRa) radios working the 915 MHz industrial, scientific medical (ISM) band. In this research, we study propagation characteristics LoRa chirp signals close to over a tropical meadow region. We as case Colima River Mexico. develop novel point-to-point measurement sounding system that does not require decoding propriety bursts provides accurate power versus distance profiles along riparian zone steeply dropping mountain river. used obtain measurements reported work, which analyzed modeled. The results show signal exhibits log-normal distribution. As result processing, two new experimental path loss models presented. considerable degradation received within vegetation less at antenna heights closer surface.

Language: Английский

Citations

13
Point‐of‐interest detection from Weibo data for map updating DOI
Xue Yang, Jie Gao, Jianbo Yu

et al.

Transactions in GIS, Journal Year: 2022, Volume and Issue: 26(6), P. 2716 - 2738

Published: Sept. 1, 2022

Abstract Points‐of‐interest (POIs) geographic information system data are increasingly important for supporting map generation and navigation services, although updating their semantic location still largely depends on manual labor. In this study, we propose a novel method to automatically detect the changes in POIs from Chinese text check‐in position provided by social media platform, Weibo. The proposed includes three steps: (1) POI name recognition; (2) confirmation; (3) change detection. First, recognizing POI's Weibo using improved conditional random field algorithm. Then, of each named integrating address with position. detected recognized extracting status words three‐level word database. To verify effectiveness method, examine Wuhan as case commercial real‐world collected January September 2020. Based validation common platforms, investigation 55 samples, identification accuracies newly added POIs, unchanged expired approximately 100, 95.8, 91.7%, respectively.

Language: Английский

Citations

7