The sum of all harmonic components of a waveform relative to the fundamental component of waveform is termed as total harmonic distortion (THD). In this paper we have compared THD of photovoltaic (PV) systems connected with a grid for four different MPPT algorithms which includes artificial neural network (ANN), incremental conductance (INC), perturb and observe (P&O), and fuzzy logic control (FLC). The simulation results clearly present the difference in THD among all four MPPT algorithms. We have also designed a three phase LCL filters to filter out the harmonics in the output signal of the system. These are the specially designed filters to eliminate the harmonics with improved performance as well as it is cost effective and are smaller in size because of lesser values of inductor and capacitors in it.
{"title":"MPPT Algorithms with LCL Filter for Grid Connected PV System","authors":"Shivam Dutt Jha, Siddharth, Siddharth Chowdhary, Kuldeep Singh","doi":"10.1109/CONIT55038.2022.9847960","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9847960","url":null,"abstract":"The sum of all harmonic components of a waveform relative to the fundamental component of waveform is termed as total harmonic distortion (THD). In this paper we have compared THD of photovoltaic (PV) systems connected with a grid for four different MPPT algorithms which includes artificial neural network (ANN), incremental conductance (INC), perturb and observe (P&O), and fuzzy logic control (FLC). The simulation results clearly present the difference in THD among all four MPPT algorithms. We have also designed a three phase LCL filters to filter out the harmonics in the output signal of the system. These are the specially designed filters to eliminate the harmonics with improved performance as well as it is cost effective and are smaller in size because of lesser values of inductor and capacitors in it.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133053937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9847821
N. Prathap, Akash Suresh, P. G., T. Manjunath
The corona virus, otherwise known as the ‘Covid-19’ is a pandemic that struck the world in December of 2019 and continues on till this day as of writing this research article. It's a virus that targets & affects an individual's immune system. Its most common symptoms include fever, dry cough & tiredness. The most commonly used method used to detect the presence of the COVID-10 virus is the Reverse Transcription Polymerase Chain Reaction Test also known as the RT-PCR test. It is an invasive biomedical procedure that utilizes a nasal swab for the sample collection and provides results in about 24 hours after testing. The research work presented in this paper makes use of parameters such as the breathing patterns, smoking and drinking habits, etc. to detect the likelihood of an individual being proned to the Covid-19 virus. This is achieved by making use of a data set which will be used to train the various machine learning and deep learning algorithms.
{"title":"Prediction of COVID-19 by analysis of Breathing Patterns using the Concepts of Machine Learning and Deep Learning Techniques","authors":"N. Prathap, Akash Suresh, P. G., T. Manjunath","doi":"10.1109/CONIT55038.2022.9847821","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9847821","url":null,"abstract":"The corona virus, otherwise known as the ‘Covid-19’ is a pandemic that struck the world in December of 2019 and continues on till this day as of writing this research article. It's a virus that targets & affects an individual's immune system. Its most common symptoms include fever, dry cough & tiredness. The most commonly used method used to detect the presence of the COVID-10 virus is the Reverse Transcription Polymerase Chain Reaction Test also known as the RT-PCR test. It is an invasive biomedical procedure that utilizes a nasal swab for the sample collection and provides results in about 24 hours after testing. The research work presented in this paper makes use of parameters such as the breathing patterns, smoking and drinking habits, etc. to detect the likelihood of an individual being proned to the Covid-19 virus. This is achieved by making use of a data set which will be used to train the various machine learning and deep learning algorithms.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121172442","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9847708
K. Sudharson, A. M. Sermakani, V. Parthipan, D. Dhinakaran, G. Eswari Petchiammal, N. Usha
Image classification is among the most important responsibilities in medical visual assessment and is typically the first and foremost basic progression in numerous medical purposes. MRI Image division is used in brain research regularly for analyzing and visualizing anatomical structures, collapsing brain alterations, showing compulsive places, and for careful organization and image-directed therapy. We emphasize disparities between them and discuss about its strengths, reference points, and constraints. To tackle the complexities and difficulty of the brain MRI partition problem, we primarily introduce the core notions of image separation. At that time, we detail varied MRI pre - processing techniques covering image enlisting, predisposed field restoration, and removal of non brain tissue. This system examines items using a controlled division technique based on Convolution Neural Networks (CNN). Because there are fewer strains in the machine, using minor parts allows for more in-depth architecture and a good outcome against additional matching. In addition, we investigated the use of strength in normalization as a preprocessing phase in Hybrid CNN-based partition techniques, which is beneficial for brainstem tumor partitions in MRI image scans when combined with knowledge enlargement.
{"title":"Hybrid Deep Learning Neural System for Brain Tumor Detection","authors":"K. Sudharson, A. M. Sermakani, V. Parthipan, D. Dhinakaran, G. Eswari Petchiammal, N. Usha","doi":"10.1109/CONIT55038.2022.9847708","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9847708","url":null,"abstract":"Image classification is among the most important responsibilities in medical visual assessment and is typically the first and foremost basic progression in numerous medical purposes. MRI Image division is used in brain research regularly for analyzing and visualizing anatomical structures, collapsing brain alterations, showing compulsive places, and for careful organization and image-directed therapy. We emphasize disparities between them and discuss about its strengths, reference points, and constraints. To tackle the complexities and difficulty of the brain MRI partition problem, we primarily introduce the core notions of image separation. At that time, we detail varied MRI pre - processing techniques covering image enlisting, predisposed field restoration, and removal of non brain tissue. This system examines items using a controlled division technique based on Convolution Neural Networks (CNN). Because there are fewer strains in the machine, using minor parts allows for more in-depth architecture and a good outcome against additional matching. In addition, we investigated the use of strength in normalization as a preprocessing phase in Hybrid CNN-based partition techniques, which is beneficial for brainstem tumor partitions in MRI image scans when combined with knowledge enlargement.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129355747","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9847977
N. Kulkarni, Nidhi Singh, Yamini Joshi, Nikhil Hasabi, S. Meena, Uday Kulkarni, Sunil V. Gurlahosur
Deep Neural Networks are known for their applications in the domains like computer vision, natural language processing, speech recognition, pattern recognition etc. Though these models are incredibly powerful they consume a considerable amount of memory bandwidth, storage and other computational resources. These heavy models can be successfully executed on machines with CPU/GPU/TPU support. It becomes difficult for the embedded devices to execute them as they are computationally constrained. In order to ease the deployment of these models onto the embedded devices we need to optimize them. Optimization of the model refers to the decrease in model size without compromising with the performance such as model accuracy, number of flops, and model parameters. We present a hybrid optimisation method to address this problem. Hybrid optimization is a 2-phase technique, pruning followed by quantization. Pruning is the process of eliminating inessential weights and connections in order to reduce the model size. Once the unnecessary parameters are removed, the weights of the remaining parameters are converted into 8-bit integer value and is termed quantization of the model. We verify and validate the performance of this hybrid optimization technique for image classification task on the CIFAR-10 dataset. We performed hybrid optimization process for 3 heavy weight models in this work namely ResNet56, ResNet110 and GoogleNet. On an average, the difference in number of flops and parameters is 40%. The reduction in number of parameters and flops has negligible effect on model performance and the variation in accuracy is less than 2%. Further, the optimized model is deployed on edge devices and embedded platform, NVIDIA Jetson TX2 Module.
{"title":"Hybrid Optimization for DNN Model Compression and Inference Acceleration","authors":"N. Kulkarni, Nidhi Singh, Yamini Joshi, Nikhil Hasabi, S. Meena, Uday Kulkarni, Sunil V. Gurlahosur","doi":"10.1109/CONIT55038.2022.9847977","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9847977","url":null,"abstract":"Deep Neural Networks are known for their applications in the domains like computer vision, natural language processing, speech recognition, pattern recognition etc. Though these models are incredibly powerful they consume a considerable amount of memory bandwidth, storage and other computational resources. These heavy models can be successfully executed on machines with CPU/GPU/TPU support. It becomes difficult for the embedded devices to execute them as they are computationally constrained. In order to ease the deployment of these models onto the embedded devices we need to optimize them. Optimization of the model refers to the decrease in model size without compromising with the performance such as model accuracy, number of flops, and model parameters. We present a hybrid optimisation method to address this problem. Hybrid optimization is a 2-phase technique, pruning followed by quantization. Pruning is the process of eliminating inessential weights and connections in order to reduce the model size. Once the unnecessary parameters are removed, the weights of the remaining parameters are converted into 8-bit integer value and is termed quantization of the model. We verify and validate the performance of this hybrid optimization technique for image classification task on the CIFAR-10 dataset. We performed hybrid optimization process for 3 heavy weight models in this work namely ResNet56, ResNet110 and GoogleNet. On an average, the difference in number of flops and parameters is 40%. The reduction in number of parameters and flops has negligible effect on model performance and the variation in accuracy is less than 2%. Further, the optimized model is deployed on edge devices and embedded platform, NVIDIA Jetson TX2 Module.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128772512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9848072
A. Chopde, Dhanesh Manwani, Kshitij Kadam
The accuracy of the model parameters of a solar cell model is crucial for forecasting power generation, fault diagnosis and system optimization in solar power generation systems. In this paper, we present a comparative study of analytical methods and optimization techniques (specifically, fminsearch and curve fitting in MATLAB) used to determine the unknown parameters of a solar cell model. In the proposed technique, we found the initial guess of these parameters using the analytical methods, reported in the literature. These initial values were refined using fminsearch optimizer in MATLAB and compared with the results obtained by fitting the curves. Comparison results are presented to validate the effectiveness of the proposed technique for parameter extraction.
{"title":"A Comparative Study of Analytical Methods And Optimization Techniques Used for Parameter Extraction of A Solar Cell Model","authors":"A. Chopde, Dhanesh Manwani, Kshitij Kadam","doi":"10.1109/CONIT55038.2022.9848072","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9848072","url":null,"abstract":"The accuracy of the model parameters of a solar cell model is crucial for forecasting power generation, fault diagnosis and system optimization in solar power generation systems. In this paper, we present a comparative study of analytical methods and optimization techniques (specifically, fminsearch and curve fitting in MATLAB) used to determine the unknown parameters of a solar cell model. In the proposed technique, we found the initial guess of these parameters using the analytical methods, reported in the literature. These initial values were refined using fminsearch optimizer in MATLAB and compared with the results obtained by fitting the curves. Comparison results are presented to validate the effectiveness of the proposed technique for parameter extraction.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"106 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116426399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9848262
Sumeet Balwade, Deepak Mali, Sagar V. Mahajan, Birudev Yele, Nilesh P. Sable
When it comes to recognizing someone, the most significant feature is their face. Face recognition aids in verifying any person's identification by using his particular traits because it acts as an individual identity for everyone. The whole technique for authenticating any face data is separated into two stages. Face Recognition system is used in the first step. is done rapidly unless in circumstances when the item is put relatively far away, and then the second phase begins in which the face is identified as a person. The entire process is then repeated, assisting in the development of a face recognition model, which is regarded to be one of the most meticulously planned biometric technologies. The photographs of people's faces are collected by people, and also the images are processed immediately by the identification equipment. As a result, the paper offers pertinent facial recognition research from a variety of perspectives. The study outlines the developmental stages and the technology associated with facial recognition. We offer face detection and recognition analysis research for real-world settings, as well as universal Face recognition databases and assessment criteria We take a look at face recognition in advance. Face recognition has emerged as a viable future growth path with a variety of applications.
{"title":"Real Time Identity Identification using Deep Learning","authors":"Sumeet Balwade, Deepak Mali, Sagar V. Mahajan, Birudev Yele, Nilesh P. Sable","doi":"10.1109/CONIT55038.2022.9848262","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9848262","url":null,"abstract":"When it comes to recognizing someone, the most significant feature is their face. Face recognition aids in verifying any person's identification by using his particular traits because it acts as an individual identity for everyone. The whole technique for authenticating any face data is separated into two stages. Face Recognition system is used in the first step. is done rapidly unless in circumstances when the item is put relatively far away, and then the second phase begins in which the face is identified as a person. The entire process is then repeated, assisting in the development of a face recognition model, which is regarded to be one of the most meticulously planned biometric technologies. The photographs of people's faces are collected by people, and also the images are processed immediately by the identification equipment. As a result, the paper offers pertinent facial recognition research from a variety of perspectives. The study outlines the developmental stages and the technology associated with facial recognition. We offer face detection and recognition analysis research for real-world settings, as well as universal Face recognition databases and assessment criteria We take a look at face recognition in advance. Face recognition has emerged as a viable future growth path with a variety of applications.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116683206","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9848046
Subhasri Kar, Sumit Banerjee, C. K. Chanda
Partial shading is a common phenomenon in the photovoltaic (PV) system and reduces maximum generated output power. In the efficient solar module shadowing effect, a relatively hot spot hinders improving the overall PV system. Bypass diode (BPD) and blocking diodes (BD) are two essential components to protect the PV module or array from unwanted heat loss. Light intensity is an essential parameter for PV power generation, and various levels of sun irradiance levels give the theoretical understanding of partial shading in the solar cell. In this paper, the current-voltage (I-V) and power-voltage (P-V) characteristics have been analysed in the case of shadowing cells with unshaded cells in Matlab software. Also, the output characteristics have been shown with and without BPD.
{"title":"Elimination of Hot-Spot in a Photovoltaic Module using Protection Diode","authors":"Subhasri Kar, Sumit Banerjee, C. K. Chanda","doi":"10.1109/CONIT55038.2022.9848046","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9848046","url":null,"abstract":"Partial shading is a common phenomenon in the photovoltaic (PV) system and reduces maximum generated output power. In the efficient solar module shadowing effect, a relatively hot spot hinders improving the overall PV system. Bypass diode (BPD) and blocking diodes (BD) are two essential components to protect the PV module or array from unwanted heat loss. Light intensity is an essential parameter for PV power generation, and various levels of sun irradiance levels give the theoretical understanding of partial shading in the solar cell. In this paper, the current-voltage (I-V) and power-voltage (P-V) characteristics have been analysed in the case of shadowing cells with unshaded cells in Matlab software. Also, the output characteristics have been shown with and without BPD.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116817852","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9847667
Aditi Karvekar, P. Joshi
This paper aims at implementing a smart., light weight and economical DC to DC converter topology in closed loop mode in order to design an uninterrupted power supply to the electronic and auxiliary loads in a More Electric Aircraft with high efficiency and good transient and steady state response. The proposed topology uses a non-isolated cascaded bidirectional DC to DC converter which can be used in buck as well as boost mode. The aircraft load which is to be supplied with DC power is so connected that it can take power from AC generator mounted on the engine shaft as well as from an auxiliary battery which gets connected to the load in case the engine gets overloaded due to takeoff., landing or turbulence conditions. The changeover between these two sources happens automatically and appropriate gate signals are provided to the semiconductor switches in the DC to DC converter with the help of closed loop control consisting of PI voltage controller hysteresis current controller. The system performance is tested under randomly varying load conditions and the load voltage and current waveforms are compared against their respective reference values. The system transient response is evaluated in terms of overshoot and voltage regulation across the load. PI controllers can be replaced with more advanced controllers like sliding mode controller or fuzzy controller in order to get even better system response in terms of load current overshoot under changing load conditions.
{"title":"A Simplified Approach to Closed Loop Control of A Non-Isolated Bidirectional DC To DC Converter","authors":"Aditi Karvekar, P. Joshi","doi":"10.1109/CONIT55038.2022.9847667","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9847667","url":null,"abstract":"This paper aims at implementing a smart., light weight and economical DC to DC converter topology in closed loop mode in order to design an uninterrupted power supply to the electronic and auxiliary loads in a More Electric Aircraft with high efficiency and good transient and steady state response. The proposed topology uses a non-isolated cascaded bidirectional DC to DC converter which can be used in buck as well as boost mode. The aircraft load which is to be supplied with DC power is so connected that it can take power from AC generator mounted on the engine shaft as well as from an auxiliary battery which gets connected to the load in case the engine gets overloaded due to takeoff., landing or turbulence conditions. The changeover between these two sources happens automatically and appropriate gate signals are provided to the semiconductor switches in the DC to DC converter with the help of closed loop control consisting of PI voltage controller hysteresis current controller. The system performance is tested under randomly varying load conditions and the load voltage and current waveforms are compared against their respective reference values. The system transient response is evaluated in terms of overshoot and voltage regulation across the load. PI controllers can be replaced with more advanced controllers like sliding mode controller or fuzzy controller in order to get even better system response in terms of load current overshoot under changing load conditions.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115524196","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9848388
Kunal Agarwal, S. Vadhera
Being over reliant on fossils fuels has resulted in massive increase of pollution levels causing the average global temperature to rise. Keeping in mind that, power extraction from renewable energy sources have been of great interest for all nations. Extracting energy from wind is a popular and sustainable source of energy. Since wind speeds are intermittent in nature, prediction of wind speeds is an important aspect in power generation through wind turbines. This work focuses on wind speed prediction along the coastal line of peninsular India taking four time-related parameters and eight meteorological parameters wherein past wind speeds are also used as an input of twenty-seven sites. The data has been collected from Indian Meteorological Department for a span of five years (2016 - 2020), three-hour average. Time-series prediction has been done using Levenberg-Marquardt (LM), Bayesian Regularization (BR) and Scaled Conjugate Gradient (SCG) algorithms in MATLAB and a comparative study has been done while altering the training, validation and testing percentages along with number of hidden layers in the neural network to identify the best algorithm with the help of linear regression and mean square error. Further sensitivity analysis is done amongst all the seven meteorological parameters in order to identify the most and least wind speed affecting factors.
{"title":"Short-term Wind Speed Forecasting of Coastal Line of Peninsular India Using NARX Models","authors":"Kunal Agarwal, S. Vadhera","doi":"10.1109/CONIT55038.2022.9848388","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9848388","url":null,"abstract":"Being over reliant on fossils fuels has resulted in massive increase of pollution levels causing the average global temperature to rise. Keeping in mind that, power extraction from renewable energy sources have been of great interest for all nations. Extracting energy from wind is a popular and sustainable source of energy. Since wind speeds are intermittent in nature, prediction of wind speeds is an important aspect in power generation through wind turbines. This work focuses on wind speed prediction along the coastal line of peninsular India taking four time-related parameters and eight meteorological parameters wherein past wind speeds are also used as an input of twenty-seven sites. The data has been collected from Indian Meteorological Department for a span of five years (2016 - 2020), three-hour average. Time-series prediction has been done using Levenberg-Marquardt (LM), Bayesian Regularization (BR) and Scaled Conjugate Gradient (SCG) algorithms in MATLAB and a comparative study has been done while altering the training, validation and testing percentages along with number of hidden layers in the neural network to identify the best algorithm with the help of linear regression and mean square error. Further sensitivity analysis is done amongst all the seven meteorological parameters in order to identify the most and least wind speed affecting factors.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115481139","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-24DOI: 10.1109/CONIT55038.2022.9848088
Bhavana Kumbar, Ankita Mane, Varsha Chalageri, Shashidhara B. Vyakaranal, S. Meena, Sunil V. Gurlahosur, Uday Kulkarni
Deep Neural Networks (DNNs) have exemplified exceptional success in solving various complicated tasks that were difficult to solve in the past using conventional machine learning methods. Deep learning has become an inevitable part of several applications in the present scenarios. However., the latest works have found that the DNNs are unfortified against the prevailing adversarial attacks. The addition of imperceptible perturbations to the inputs causes the neural networks to fail and predict incorrect outputs. In practice., adversarial attacks create a significant challenge to the success of deep learning as they aim to deteriorate the performance of the classifiers by fooling the deep learning algorithms. This paper provides a comprehensive comparative study on the common adversarial attacks and countermeasures against them and also analyzes their behavior on standard datasets such as MNIST and CIFAR10 and also on a custom dataset that spans over 1000 images consisting of 5 classes. To mitigate the adversarial effects on deep learning models., we provide solutions against the conventional adversarial attacks that reduce 70% accuracy. It results in making the deep learning models more resilient against adversaries.
{"title":"A Comparative Study on Adversarial Attacks and Defense Mechanisms","authors":"Bhavana Kumbar, Ankita Mane, Varsha Chalageri, Shashidhara B. Vyakaranal, S. Meena, Sunil V. Gurlahosur, Uday Kulkarni","doi":"10.1109/CONIT55038.2022.9848088","DOIUrl":"https://doi.org/10.1109/CONIT55038.2022.9848088","url":null,"abstract":"Deep Neural Networks (DNNs) have exemplified exceptional success in solving various complicated tasks that were difficult to solve in the past using conventional machine learning methods. Deep learning has become an inevitable part of several applications in the present scenarios. However., the latest works have found that the DNNs are unfortified against the prevailing adversarial attacks. The addition of imperceptible perturbations to the inputs causes the neural networks to fail and predict incorrect outputs. In practice., adversarial attacks create a significant challenge to the success of deep learning as they aim to deteriorate the performance of the classifiers by fooling the deep learning algorithms. This paper provides a comprehensive comparative study on the common adversarial attacks and countermeasures against them and also analyzes their behavior on standard datasets such as MNIST and CIFAR10 and also on a custom dataset that spans over 1000 images consisting of 5 classes. To mitigate the adversarial effects on deep learning models., we provide solutions against the conventional adversarial attacks that reduce 70% accuracy. It results in making the deep learning models more resilient against adversaries.","PeriodicalId":270445,"journal":{"name":"2022 2nd International Conference on Intelligent Technologies (CONIT)","volume":"330 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115604832","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}