论文标题
RFI检测和删除射电天文学的自我学习神经网络方法
A Self-Learning Neural Network Approach for RFI Detection and Removal in Radio Astronomy
论文作者
论文摘要
我们提出了一种新型的神经网络（NN）方法，用于从典型射电天文学实验的信号处理链中的原始数字化信号中检测和去除射频干扰（RFI）。我们方法的主要优点是它不需要培训集。取而代之的是，我们的方法依赖于这样一个事实，即来自天文来源的真正感兴趣信号是热的，因此被描述为高斯随机过程，无法压缩。我们采用差异编码器/解码器网络来在数据式流中找到可压缩信息，这些信息可以解释自由度最少的最大差异。我们在Baryon映射实验（BMX）原型的一组玩具问题和存储的环形袋中进行了证明。我们发现RFI的减法在清洁模拟的时间播放方面有效：虽然我们发现，NN的RFI清洗的时间播放的功率光谱受到与添加噪声一致的额外信号的额外信号，但我们发现，它通常在频段中的百分比级别，即使在RFI的频谱中，RFI的频谱也比RFI频谱均高于rfi of Signal nitive of Signal nitive of Signal signal of Signal signal signal signal signal signal signal insime of Signal nitive。我们讨论了这种方法的优势和局限性，并在未来无线电实验的前端中实现了可能的实现。
We present a novel neural network (NN) method for the detection and removal of Radio Frequency Interference (RFI) from the raw digitized signal in the signal processing chain of a typical radio astronomy experiment. The main advantage of our method is that it does not require a training set. Instead, our method relies on the fact that the true signal of interest coming from astronomical sources is thermal and therefore described as a Gaussian random process, which cannot be compressed. We employ a variational encoder/decoder network to find the compressible information in the datastream that can explain the most variance with the fewest degrees of freedom. We demonstrate it on a set of toy problems and stored ringbuffers from the Baryon Mapping eXperiment (BMX) prototype. We find that the RFI subtraction is effective at cleaning simulated timestreams: while we find that the power spectra of the RFI-cleaned timestreams output by the NN suffer from extra signal consistent with additive noise, we find that it is generally around percent level across the band and sub 10 percent in contaminated spectral channels even when RFI power is an order of magnitude larger than the signal. We discuss advantages and limitations of this method and possible implementation in the front-end of future radio experiments.
