A Comprehensive Performance Analysis of Transfer Learning Optimization in Visual Field Defect Classification

Abu, Masyitah; Zahri, Nik Adilah Hanin; Amir, Amiza; Ismail, Muhammad Izham; Yaakub, Azhany; Anwar, Said Amirul

doi:10.3390/diagnostics12051258

Cited by 8 publications

(5 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By modeling the PSRR parameter as the objective function, other specifications as constraint functions, and the geometrical size as design variables, the design problem can be formulated as an optimization problem in canonical form as presented in (18).…”

Section: Circuit Optimization Process a Problem Formulationmentioning

confidence: 99%

See 1 more Smart Citation

Novel Methods for Improved Particle Swarm Optimization in Designing the Bandgap Reference Circuit

Hoang,

Quoc,

Zhang

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Bandgap reference (BGR) circuits play a crucial role as voltage reference generators for other components within a variety of analog integrated circuits. Therefore, their power supply rejection ratio (PSRR), which represents the BGR's ability to maintain a stable output in the presence of supply ripples, has a substantial impact on the overall circuit performance and is consequently worthy of attention. In this work, the design of a 65nm-CMOS BGR circuit and the computational approach to optimize its PSRR value are presented. Our proposed BGR circuit incorporates a modification in the op-amp structure to enhance the PSRR parameter. Detailed explanations of the proposed op-amp's differential gain calculation and the PSRR parameter formula are provided. Additionally, we employ the particle swarm optimization (PSO) algorithm to maximize PSRR while satisfying other specifications. Four distinct approaches for utilizing the inertia weight parameter of the PSO algorithm are explored: two newly developed techniques (PSO -local exploitation orienter (PSO-LEO) and PSO -global exploration orienter (PSO-GEO)) together with two conventional approaches. A comprehensive comparative analysis reveals the superiority of our proposed PSO-GEO technique. Ultimately, our 65 nm CMOS bandgap circuit exhibits exceptional performance, featuring a temperature coefficient of 6.4056 ppm/°C and a PSRR of 99.9896 dB at 1 kHz. In conclusion, this work contributes substantially through op-amp architecture modification, PSRR optimization via the PSO algorithm, and introducing new inertia weight parameter utilization strategies and analyzing them.

show abstract

Section: Circuit Optimization Process a Problem Formulationmentioning

confidence: 99%

“…In (18), x = [w_pbias, w_ndiff, w_m1_pcasc, 1_ndiff, w_pdiff, w_m2_pcasc, w_nbias] is the vector of design variables. The boundary of each variable is chosen based on experience.…”

Section: Circuit Optimization Process a Problem Formulationmentioning

confidence: 99%

Novel Methods for Improved Particle Swarm Optimization in Designing the Bandgap Reference Circuit

Hoang,

Quoc,

Zhang

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The most popular models were AlexNet, VGG, Xception, Inception, MobileNet, DenseNet, ResNet, GoogleLeNet, and YOLOs. In (Abu et al, 2022;Sharma et al, 2021;Zhao, 2017), all suggested considering fine-tuning several hyperparameters (feature map, filter size, activation function, pool size, optimiser, learning rate, batch size, epoch, dropout rate, loss function, and evaluation metric) of the pre-trained model.…”

Section: Transfer Learningmentioning

confidence: 99%

Ten deep learning techniques to address small data problems with remote sensing

Safonova¹,

Ghazaryan²,

Stiller³

et al. 2023

Preprint

View full text Add to dashboard Cite

Researchers and engineers have increasingly used Deep Learning (DL) for a variety of Remote Sensing (RS) tasks. However, data from local observations or via ground truth is often quite limited for training DL models, especially when these models represent key socio-environmental problems, such as the monitoring of extreme, destructive climate events, biodiversity, and sudden changes in ecosystem states. Such cases, also known as small data problems, pose significant methodological challenges. This review summarises these challenges in the RS domain and the possibility of using emerging DL techniques to overcome them. We show that the small data problem is a common challenge across disciplines and scales that results in poor model generalisability and transferability, yet this has not been investigated in a structured way. We first introduce ten emerging DL techniques: transfer learning, self-supervised learning, semi-supervised learning, few-shot learning, zero-shot learning, active learning, weakly supervised learning, multitask learning, process-aware learning, and ensemble learning; we also include a validation technique known as spatial k-fold cross validation. These techniques have shown promising potential in other scientific disciplines, but have been rarely applied in the RS domain. We also provide guidance on which learning technique to use in various cases, which helps to create a more methodologically robust DL application (and a greater number of them) that can be used to tackle socially important problems with limited data.

show abstract

“…On the other hand, MobileNet represents a more lightweight architectural approach with fewer layers that is designed for mobile and embedded vision applications [14]. Despite its reduced complexity when compared to VGG and ResNet, MobileNet remains adept at delivering valuable feature representations for a diverse array of transfer learning tasks, characterized by a heightened level of efficiency [15] [16].…”

Section: Introductionmentioning

confidence: 99%

A Comparative Study of MobileNet Architecture Optimizer for Crowd Prediction

Putra,

Naufal,

Hidayat

2023

JPIT

View full text Add to dashboard Cite

Artificial intelligence technology has grown quickly in recent years. Convolutional neural network (CNN) technology has also been developed as a result of these developments. However, because convolutional neural networks entail several calculations and the optimization of numerous matrices, their application necessitates the utilization of appropriate technology, such as GPUs or other accelerators. Applying transfer learning techniques is one way to get around this resource barrier. MobileNetV2 is an example of a lightweight convolutional neural network architecture that is appropriate for transfer learning. The objective of the research is to compare the performance of SGD and Adam using the MobileNetv2 convolutional neural network architecture. Model training uses a learning rate of 0.0001, batch size of 32, and binary cross-entropy as the loss function. The training process is carried out for 100 epochs with the application of early stop and patience for 10 epochs. Result of this research is both models using Adam's optimizer and SGD show good capability in crowd classification. However, the model with the SGD optimizer has a slightly superior performance even with less accuracy than model with Adam optimizer. Which is model with Adam has accuracy 96%, while the model with SGD has 95% accuracy. This is because in the graphical results model with the SGD optimizer shows better stability than the model with the Adam optimizer. The loss graph and accuracy graph of the SGD model are more consistent and tend to experience lower fluctuations than the Adam model.

show abstract

A Comprehensive Performance Analysis of Transfer Learning Optimization in Visual Field Defect Classification

Cited by 8 publications

References 44 publications

Novel Methods for Improved Particle Swarm Optimization in Designing the Bandgap Reference Circuit

Novel Methods for Improved Particle Swarm Optimization in Designing the Bandgap Reference Circuit

Ten deep learning techniques to address small data problems with remote sensing

A Comparative Study of MobileNet Architecture Optimizer for Crowd Prediction

Contact Info

Product

Resources

About