Search

Machine learning models find extensive application across diverse domains, with their performance heavily reliant on the data quality employed during the learning process. However, real-world datasets include some missing data due to limitations and errors in data collection methods, incomplete or inconsistent data-gathering processes, and human errors during processing. Consequently, effective handling of missing values becomes imperative to ensure optimal model performance. A common way to deal with missing data is to either delete the data containing the missing values or to impute them appropriately. Deletion is straightforward, but at the cost of information loss. Imputation, on the other hand, can result in a loss of variability in the dataset and skewed correlations between variables. The proposed scheme reduces dimensionality by utilizing variables without missing values and employs the outcomes to estimate the missing values. Experimental validations affirm that the proposed scheme mitigates the performance degradation of various machine learning models compared to existing methods. [ABSTRACT FROM AUTHOR]

Since single cell RNA sequencing provides the expression profiles of individual cells, it provides higher cellular differential resolution than traditional bulk RNA sequencing. Using these single cell RNA sequencing data, clustering analysis is generally conducted to find cell types and understand high level biological processes. In order to effectively process the high-dimensional single cell RNA sequencing data fir the clustering analysis, this paper uses a variational autoencoder to transform a high dimensional data space into a lower dimensional latent space, expecting to produce a latent space that can give more accurate clustering results. By clustering the features in the transformed latent space, we compare the performance of various classical clustering methods for single cell RNA sequencing data. Experimental results demonstrate that the proposed framework outperforms many state-of-the-art methods under various clustering performance metrics. [ABSTRACT FROM AUTHOR]

In this paper, a compensation scheme of frequency selective IQ mismatch for high-performance radar systems based on commercial RFIC’s is proposed. Besides, an optimization model and its solution based on the dimension reduction scheme using singular value decomposition are also proposed to design the optimal IQ mismatch compensation digital filter with complex coefficients. The performance of the proposed method had been analyzed through experiments using the IQ mismatch measurement and compensation system implemented on an FPGA board with a target RFIC and compared with the previous method. The experiment result showed a performance improvement of the proposed method over the existing one without noticeable increments in complexities. These performance analysis results showed that the limitation of using commercial RFIC’s in high-performance radar systems due to the undesirable maximum SNR cap caused by their IQ mismatches could be overcome by employing the proposed method. [ABSTRACT FROM AUTHOR]