ABSTRACT There are a number of tools and software packages dealing with the problem of efficient ... more ABSTRACT There are a number of tools and software packages dealing with the problem of efficient primer and probe design. Some of these tools are well accepted and very famous among biologists and some tools tend to disappear in oblivion as time passes by. In an ...
Twitter is a massive repository and a gold mine of human thoughts that expresses a person’s insta... more Twitter is a massive repository and a gold mine of human thoughts that expresses a person’s instant feeling. A retrospective review of tweets during the ensuing COVID-19 pandemic may provide valuable insights into the understanding of a person’s feelings. With an enormous amount of data, training a model to understand the precise feeling is a daunting job. There are many developments recorded in the field of deep learning to pave the way forward. The purpose of this paper is to compare the tweet sentiment classification using Google BERT, attention based Bidirectional LSTM and Convolutional Neural Networks (CNNs). The final models are trained on the Twitter dataset SemEval-2016, where the embeddings are fine-tuned again. Such models proved to be highly effective and accurate in the study of emotions as opposed to machine learning techniques.
Advances in intelligent systems and computing, Dec 14, 2018
The paper deals with the application of artificial neural network on financial data. We applied d... more The paper deals with the application of artificial neural network on financial data. We applied different activation functions in hidden layer with regularization to overcome the problem of overfitting. We present a comparative analysis of all combinations of activation functions and regularizations applied on BSE Sensex and Nifty 50 dataset containing the stock indices of last 7 years.
TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON), Dec 7, 2021
In the current time, Indian agriculture is lagging in the use of advanced technological solutions... more In the current time, Indian agriculture is lagging in the use of advanced technological solutions in tackling various farming-related issues such as crop health, weed problems, crop diseases, etc. We intend to bridge this gap by proposing technological solutions to automatically detect insects in Soybean crops. Soybean (Glycine max) is an edible seed from an annual legume in the pea family (Fabaceae). The soybean is the world's most economically important bean, providing vegetable protein to millions of people as well as ingredients for hundreds of chemical goods. Object detection is a computer vision task that involves the identification of object class with its location in the image. We have employed three popular object detection algorithms for insect identification on Soybean crop fields. YOLO v3, v4, and v5 have been trained to detect and demarcate the insect presence on the field. The simulation results revealed that the YOLO v5 delivers the best insect detection accuracy with mean average precision (mAP) of 99.5% followed by YOLO v4 and v3.
Machine learning tasks of interest, such as classification, recognition etc., require labeled dat... more Machine learning tasks of interest, such as classification, recognition etc., require labeled data which is often very difficult to obtain. Unlabeled data is relatively easier to obtain compared to labeled data. Self-taught transfer Learning is one such method in which higher-level features or bases are extracted from the huge amount of unlabeled data in source domain and further these features are used with labeled data in target domain to perform the task of interest. We have introduced novel method, using isomap, for generation of features or bases in source domain, which used in target domain to perform the classification with small amount of labeled data. Proposed method is compared with sparse coding and results are promising and efficient.
Communications in computer and information science, 2018
Neural networks having a large number of parameters are considered as very effective machine lear... more Neural networks having a large number of parameters are considered as very effective machine learning tool. But as the number of parameters becomes large, the network becomes slow to use and the problem of overfitting arises. Various ways to prevent overfitting of model are further discussed here and a comparative study has been done for the same. The effects of various regularization methods on the performance of neural net models are observed.
Multiobjective optimization algorithms generate a set of Pareto optimal points which are all equi... more Multiobjective optimization algorithms generate a set of Pareto optimal points which are all equivalent. However, as per the Decision Maker (DM), a number of these solutions are unacceptable. In this paper, we propose two methods to obtain a subset of the Pareto front on the basis of the amount of loss a DM is ready to bear while compromising for a Pareto solution. For this, we introduce two different methods dBValue and dAValue to handle the scenario. A concept of Relenting Factor (R) is used to incorporate the compromise defined by DM. An elitist archive is implemented which is periodically filtered on the basis of dBValue or dAValue techniques to filter out best individuals for each generation. This semi-automatic filtering guides the search space towards Region Of Interest (ROI) with respect to the DM. The proposed method is evaluated on 3 objectives standard benchmark test problems. Results demonstrate that the proposed method is efficiently able to converge on the true Pareto front while limiting the generated solutions only to the ROI.
ABSTRACT There are a number of tools and software packages dealing with the problem of efficient ... more ABSTRACT There are a number of tools and software packages dealing with the problem of efficient primer and probe design. Some of these tools are well accepted and very famous among biologists and some tools tend to disappear in oblivion as time passes by. In an ...
Twitter is a massive repository and a gold mine of human thoughts that expresses a person’s insta... more Twitter is a massive repository and a gold mine of human thoughts that expresses a person’s instant feeling. A retrospective review of tweets during the ensuing COVID-19 pandemic may provide valuable insights into the understanding of a person’s feelings. With an enormous amount of data, training a model to understand the precise feeling is a daunting job. There are many developments recorded in the field of deep learning to pave the way forward. The purpose of this paper is to compare the tweet sentiment classification using Google BERT, attention based Bidirectional LSTM and Convolutional Neural Networks (CNNs). The final models are trained on the Twitter dataset SemEval-2016, where the embeddings are fine-tuned again. Such models proved to be highly effective and accurate in the study of emotions as opposed to machine learning techniques.
Advances in intelligent systems and computing, Dec 14, 2018
The paper deals with the application of artificial neural network on financial data. We applied d... more The paper deals with the application of artificial neural network on financial data. We applied different activation functions in hidden layer with regularization to overcome the problem of overfitting. We present a comparative analysis of all combinations of activation functions and regularizations applied on BSE Sensex and Nifty 50 dataset containing the stock indices of last 7 years.
TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON), Dec 7, 2021
In the current time, Indian agriculture is lagging in the use of advanced technological solutions... more In the current time, Indian agriculture is lagging in the use of advanced technological solutions in tackling various farming-related issues such as crop health, weed problems, crop diseases, etc. We intend to bridge this gap by proposing technological solutions to automatically detect insects in Soybean crops. Soybean (Glycine max) is an edible seed from an annual legume in the pea family (Fabaceae). The soybean is the world's most economically important bean, providing vegetable protein to millions of people as well as ingredients for hundreds of chemical goods. Object detection is a computer vision task that involves the identification of object class with its location in the image. We have employed three popular object detection algorithms for insect identification on Soybean crop fields. YOLO v3, v4, and v5 have been trained to detect and demarcate the insect presence on the field. The simulation results revealed that the YOLO v5 delivers the best insect detection accuracy with mean average precision (mAP) of 99.5% followed by YOLO v4 and v3.
Machine learning tasks of interest, such as classification, recognition etc., require labeled dat... more Machine learning tasks of interest, such as classification, recognition etc., require labeled data which is often very difficult to obtain. Unlabeled data is relatively easier to obtain compared to labeled data. Self-taught transfer Learning is one such method in which higher-level features or bases are extracted from the huge amount of unlabeled data in source domain and further these features are used with labeled data in target domain to perform the task of interest. We have introduced novel method, using isomap, for generation of features or bases in source domain, which used in target domain to perform the classification with small amount of labeled data. Proposed method is compared with sparse coding and results are promising and efficient.
Communications in computer and information science, 2018
Neural networks having a large number of parameters are considered as very effective machine lear... more Neural networks having a large number of parameters are considered as very effective machine learning tool. But as the number of parameters becomes large, the network becomes slow to use and the problem of overfitting arises. Various ways to prevent overfitting of model are further discussed here and a comparative study has been done for the same. The effects of various regularization methods on the performance of neural net models are observed.
Multiobjective optimization algorithms generate a set of Pareto optimal points which are all equi... more Multiobjective optimization algorithms generate a set of Pareto optimal points which are all equivalent. However, as per the Decision Maker (DM), a number of these solutions are unacceptable. In this paper, we propose two methods to obtain a subset of the Pareto front on the basis of the amount of loss a DM is ready to bear while compromising for a Pareto solution. For this, we introduce two different methods dBValue and dAValue to handle the scenario. A concept of Relenting Factor (R) is used to incorporate the compromise defined by DM. An elitist archive is implemented which is periodically filtered on the basis of dBValue or dAValue techniques to filter out best individuals for each generation. This semi-automatic filtering guides the search space towards Region Of Interest (ROI) with respect to the DM. The proposed method is evaluated on 3 objectives standard benchmark test problems. Results demonstrate that the proposed method is efficiently able to converge on the true Pareto front while limiting the generated solutions only to the ROI.
Uploads