Fuzzy Convolution Neural Networks for Tabular Data Classification
Authors:
Arun D. Kulkarni
Abstract:
Recently, convolution neural networks (CNNs) have attracted a great deal of attention due to their remarkable performance in various domains, particularly in image and text classification tasks. However, their application to tabular data classification remains underexplored. There are many fields such as bioinformatics, finance, medicine where nonimage data are prevalent. Adaption of CNNs to class…
▽ More
Recently, convolution neural networks (CNNs) have attracted a great deal of attention due to their remarkable performance in various domains, particularly in image and text classification tasks. However, their application to tabular data classification remains underexplored. There are many fields such as bioinformatics, finance, medicine where nonimage data are prevalent. Adaption of CNNs to classify nonimage data remains highly challenging. This paper investigates the efficacy of CNNs for tabular data classification, aiming to bridge the gap between traditional machine learning approaches and deep learning techniques. We propose a novel framework fuzzy convolution neural network (FCNN) tailored specifically for tabular data to capture local patterns within feature vectors. In our approach, we map feature values to fuzzy memberships. The fuzzy membership vectors are converted into images that are used to train the CNN model. The trained CNN model is used to classify unknown feature vectors. To validate our approach, we generated six complex noisy data sets. We used randomly selected seventy percent samples from each data set for training and thirty percent for testing. The data sets were also classified using the state-of-the-art machine learning algorithms such as the decision tree (DT), support vector machine (SVM), fuzzy neural network (FNN), Bayes classifier, and Random Forest (RF). Experimental results demonstrate that our proposed model can effectively learn meaningful representations from tabular data, achieving competitive or superior performance compared to existing methods. Overall, our finding suggests that the proposed FCNN model holds promise as a viable alternative for tabular data classification tasks, offering a fresh prospective and potentially unlocking new opportunities for leveraging deep learning in structured data analysis.
△ Less
Submitted 18 June, 2024; v1 submitted 4 June, 2024;
originally announced June 2024.
An insight into chromatic behaviour of jitter in pulsars and its modelling: A case study of PSR J0437$-$4715
Authors:
A. D. Kulkarni,
R. M. Shannon,
D. J. Reardon,
M. T. Miles,
M. Bailes,
M. Shamohammadi
Abstract:
Pulse-to-pulse profile shape variations introduce correlations in pulsar times of arrival (TOAs) across radio frequency measured at the same observational epoch. This leads to a broadband noise in excess of radiometer noise, which is termed pulse jitter noise. The presence of jitter noise limits the achievable timing precision and decreases the sensitivity of pulsar-timing data sets to signals of…
▽ More
Pulse-to-pulse profile shape variations introduce correlations in pulsar times of arrival (TOAs) across radio frequency measured at the same observational epoch. This leads to a broadband noise in excess of radiometer noise, which is termed pulse jitter noise. The presence of jitter noise limits the achievable timing precision and decreases the sensitivity of pulsar-timing data sets to signals of interest such as nanohertz-frequency gravitational waves. Current white noise models used in pulsar timing analyses attempt to account for this, assuming complete correlation of uncertainties through the arrival times collected in a unique observation and no frequency dependence of jitter (which corresponds to a rank-one covariance matrix). However, previous studies show that the brightest millisecond pulsar at decimetre wavelengths, PSR J0437$-$4715, shows decorrelation and frequency dependence of jitter noise. Here we present a detailed study of the decorrelation of jitter noise in PSR J0437$-$4715 and implement a new technique to model it. We show that the rate of decorrelation due to jitter can be expressed as a power-law in frequency. We analyse the covariance matrix associated with the jitter noise process and find that a higher-rank-approximation is essential to account for the decorrelation and to account for frequency dependence of jitter noise. We show that the use of this novel method significantly improves the estimation of other chromatic noise parameters such as dispersion measure variations. However, we find no significant improvement in errors and estimation of other timing model parameters suggesting that current methods are not biased for other parameters, for this pulsar due to this misspecification. We show that pulse energy variations show a similar decorrelation to the jitter noise, indicating a common origin for both observables.
△ Less
Submitted 8 January, 2024; v1 submitted 7 January, 2024;
originally announced January 2024.
