Deep Learning with Loss Ensembles for Solar Power Prediction in Smart Cities

Hajiabadi, M Rahimi; Farhadi, Mahdi; Babaiyan, Vahide; Estebsari, Abouzar

doi:10.3390/smartcities3030043

Cited by 11 publications

(3 citation statements)

References 25 publications

(25 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To clarify, the loss function is a technique for measuring the algorithm's success in modelling the dataset by computing the gradients associated with the neural network's error prediction. In terms of error prediction, the loss function calculates the error for a single training sample by comparing the predicted output to the intended output [17]. While this is occurring, gradients are leveraged to adjust the weights to suit the training data.…”

Section: Hyperparameter Selectionmentioning

confidence: 99%

Improving a Deep Neural Network Generative-Based Chatbot Model

Wan Ahmad,

Jambli

2024

AEJ

View full text Add to dashboard Cite

A chatbot is an application that is developed in the field of machine learning, which has become a hot topic of research in recent years. The majority of today's chatbots integrate the Artificial Neural Network (ANN) approach with a Deep Learning environment, which results in a new generation chatbot known as a Generative-Based Chatbot. The current chatbot application mostly fails to recognize the optimum capacity of the network environment due to its complex nature resulting in low accuracy and loss rate. In this paper, we aim to conduct an experiment in evaluating the performance of chatbot model when manipulating the selected hyperparameters that can greatly contribute to the well-performed model without modifying any major structures and algorithms in the model. The experiment involves training two models, which are the Attentive Sequence-to-Sequence model (baseline model), and Attentive Seq2Sequence with Hyperparametric Optimization. The result was observed by training two models on Cornell Movie-Dialogue Corpus, run by using 10 epochs. The comparison shows that after optimization, the model’s accuracy and loss rate were 87% and 0.51%, respectively, compared to the results before optimizing the network (79% accuracy and 1.05% loss)

show abstract

Section: Hyperparameter Selectionmentioning

confidence: 99%

Improving a Deep Neural Network Generative-Based Chatbot Model

Wan Ahmad,

Jambli

2024

AEJ

View full text Add to dashboard Cite

show abstract

“…The performance of the PV array is also affected by the change in irradiance from panel to panel due to shadows of passing clouds, trees, nearby buildings, changes in climate, disposition of dust, the aging effect of the panel, etc [84,85]. Therefore, partial shading conditions (PSC) can occur in the PV system due to any of the above reasons.…”

Section: Solar Pv System Under Partial Shadingmentioning

confidence: 99%

Enhanced Maximum Power Point Techniques for Solar Photovoltaic System under Uniform Insolation and Partial Shading Conditions: A Review

Bhukya¹,

Kedika²,

Salkuti

2022

Algorithms

View full text Add to dashboard Cite

In the recent past, the solar photovoltaic (PV) system has emerged as the most promising source of alternative energy. This solar PV system suffers from an unavoidable phenomenon due to the fluctuating environmental conditions. It has nonlinearity in I-V curves, which reduces the output efficiency. Hence, the optimum maximum power point (MPP) extraction of the PV system is difficult to achieve. Therefore, for maximizing the power output of PV systems, a maximum power point tracking (MPPT) mechanism, which is a control algorithm that can constantly track the MPP during operation, is required. However, choosing a suitable MPPT technique might be confusing because each method has its own set of advantages and disadvantages. Hence, a proper review of these methods is essential. In this paper, a state-of-the-art review on various MPPT techniques based on their classifications, such as offline, online, and hybrid techniques under uniform and nonuniform irradiances, is presented. In comparison to offline and online MPPT methods, intelligent MPPT techniques have better tracking accuracy and tracking efficiency with less steady state oscillations. Unlike online and offline techniques, intelligent methods track the global MPP under partial shade conditions. This review paper will be a useful resource for researchers, as well as practicing engineers, to pave the way for additional research and development in the MPPT field.

show abstract

“…Classification functions also divided into binary entropy loss/log loss and hinge loss. During the execution of AReLU the first function was used [15]. AReLU is applied on MNIST dataset that contains 70000 images of black and white handwritten digits divided into 60000 images for training and 10000 images for Testing [17].…”

Section: Introductionmentioning

confidence: 99%

Adaptive Rectified Linear Unit (Arelu) for Classification Problems to Solve Dying Problem in Deep Learning

Atoum¹

2023

IJACSA

View full text Add to dashboard Cite

A convolutional neural network (CNN) is a subset of machine learning as well as one of the different types of artificial neural networks that are used for different applications and data types. Activation functions (AFs) are used in this type of network to determine whether or not its neurons are activated. One nonlinear AF named as Rectified Linear Units (ReLU) which involves a simple mathematical operations and it gives better performance. It avoids rectifying vanishing gradient problem that inherents older AFs like tanh and sigmoid. Additionally, it has less computational cost. Despite these advantages, it suffers from a problem called Dying problem. Several modifications have been appeared to address this problem, for example; Leaky ReLU (LReLU). The main concept of our algorithm is to improve the current LReLU activation functions in mitigating the dying problem on deep learning by using the readjustment of values (changing and decreasing value) of the loss function or cost function while number of epochs are increased. The model was trained on the MNIST dataset with 20 epochs and achieved lowest misclassification rate by 1.2%. While optimizing our proposed methods, we received comparatively better results in terms of simplicity, low computational cost, and with no hyperparameters.

show abstract

Deep Learning with Loss Ensembles for Solar Power Prediction in Smart Cities

Cited by 11 publications

References 25 publications

Improving a Deep Neural Network Generative-Based Chatbot Model

Improving a Deep Neural Network Generative-Based Chatbot Model

Enhanced Maximum Power Point Techniques for Solar Photovoltaic System under Uniform Insolation and Partial Shading Conditions: A Review

Adaptive Rectified Linear Unit (Arelu) for Classification Problems to Solve Dying Problem in Deep Learning

Contact Info

Product

Resources

About