Recovering the CMB Signal with Machine Learning

One of the key steps in Cosmic Microwave Background (CMB) data analysis is component separation to recover the CMB signal from multi-frequency observations contaminated by foreground emissions. Needlet Internal Linear Combination (NILC) is one of the successful methods that applies the minimum variance estimation technique to a set of needlet-filtered frequency maps to recover CMB. In this work, we develop a deep convolutional neural network (CNN) model to recover CMB temperature map from needlet-filtered frequency maps over the full sky. The network operates on a multi-resolution representation of spherical data, capturing localised features in both pixel and harmonic space, and is designed to preserve the rotational invariance of the CMB signal. The network model is trained on realistic simulations at Planck frequencies, which include CMB temperature maps generated using cosmological parameters sampled within a 2σ standard deviation around the Planck best-fit values. We demonstrate the network performance for simulations that exhibit different foreground complexities. The recovered CMB temperature map closely follows the true signal with some residual leakage near the Galactic plane. The TT power spectrum is accurately reconstructed up to multipoles of approximately ℓ ∼ 1100. A minor residual systematics remain at smaller scales. Compared to the NILC method, the network shows reduced residual foreground contamination in the recovered CMB map. Once validated on the simulations, the network is applied to Planck PR3 intensity data. The resulting CMB map is consistent with the CMB maps from the Planck legacy products, including those produced using the NILC and SMICA pipelines. This work demonstrates a powerful component separation method to clean spherical signal data from multi-resolution wavelet-filtered maps.

One of the major goals of future cosmic microwave background (CMB) B-mode polarization experiments is the detection of primordial gravitational waves through an unbiased measurement of the tensor-to-scalar ratio r. Robust detection of this signal will require mitigating all possible contamination to the B-mode polarization from astrophysical origins. One such extragalactic contamination arises from the patchiness in the electron density during the reionization epoch. Along with the signature on CMB polarization, the patchy reionization can source secondary anisotropies on the CMB temperature through the kinetic Sunyaev–Zeldovich (kSZ) effect. In order to study the impact of this foreground for the upcoming CMB missions, we present a self-consistent framework to compute the CMB anisotropies based on a physically motivated model of reionization. We show that the value of r can bias towards a higher value if the secondary contribution from reionization is neglected. However, combining small-scale kSZ signal, large-scale E-mode polarization, and B-mode polarization measurements, we can put constraints on the patchiness in electron density during reionization and can mitigate its impact on the value of r. CMB missions such as CMB-S4 and PICO may experience a bias of >0.17σ which can go as high as ∼0.73σ for extreme reionization models allowed by the Planck and SPT CMB measurements. As future experiments target to measure r at 5σ, this is likely to affect the measurement significance and hence possibly affect the claim of detection of r, if not mitigated properly by using joint estimations of different reionization observables.

This research aims to implement deep learning techniques to determine fact and fake news in Indonesian language. The methods used are Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM). The stages of the research consisted of collecting data, labeling data, preprocessing data, word embedding, splitting data, forming CNN and LSTM models, evaluating, testing new input data and comparing evaluations of the established CNN and LSTM models. The Data are collected from a fact and fake news provider site that is valid, namely TurnbackHoax.id. There are 1786 news used in this study, with 802 fact and 984 fake news. The results indicate that the CNN and LSTM methods were successfully applied to determine fact and fake news in Indonesian language properly. CNN has an accuracy test, precision and recall value of 0.88, while the LSTM model has an accuracy test and precision value of 0.84 and a recall of 0.83. In testing the new data input, all of the predictions obtained by CNN are correct, while the prediction results obtained by LSTM have 1 wrong prediction. Based on the evaluation results and the results of testing the new data input, the model produced by the CNN method is better than the model produced by the LSTM method.

We demonstrate that, for the baseline design of the CORE satellite mission, the polarized foregrounds can be controlled at the level required to allow the detection of the primordial cosmic microwave background (CMB) B-mode polarization with the desired accuracy at both reionization and recombination scales, for tensor-to-scalar ratio values of r≳ 5× 10−3. We consider detailed sky simulations based on state-of-the-art CMB observations that consist of CMB polarization with τ=0.055 and tensor-to-scalar values ranging from r=10−2 to 10−3, Galactic synchrotron, and thermal dust polarization with variable spectral indices over the sky, polarized anomalous microwave emission, polarized infrared and radio sources, and gravitational lensing effects. Using both parametric and blind approaches, we perform full component separation and likelihood analysis of the simulations, allowing us to quantify both uncertainties and biases on the reconstructed primordial B-modes. Under the assumption of perfect control of lensing effects, CORE would measure an unbiased estimate of r=(5 ± 0.4)× 10−3 after foreground cleaning. In the presence of both gravitational lensing effects and astrophysical foregrounds, the significance of the detection is lowered, with CORE achieving a 4σ-measurement of r=5× 10−3 after foreground cleaning and 60% delensing. For lower tensor-to-scalar ratios (r=10−3) the overall uncertainty on r is dominated by foreground residuals, not by the 40% residual of lensing cosmic variance. Moreover, the residual contribution of unprocessed polarized point-sources can be the dominant foreground contamination to primordial B-modes at this r level, even on relatively large angular scales, ℓ ∼ 50. Finally, we report two sources of potential bias for the detection of the primordial B-modes by future CMB experiments: (i) the use of incorrect foreground models, e.g. a modelling error of Δβs = 0.02 on the synchrotron spectral indices may result in an excess in the recovered reionization peak corresponding to an effective Δ r > 10−3; (ii) the average of the foreground line-of-sight spectral indices by the combined effects of pixelization and beam convolution, which adds an effective curvature to the foreground spectral energy distribution and may cause spectral degeneracies with the CMB in the frequency range probed by the experiment.

Due to the continuous growth and advancement of technology, an enormous volume of text data is generated daily across various sources including social media platforms, websites, search engines, healthcare records, and news articles. Extracting meaningful patterns from text data, such as viewpoints, related theories, journal distribution, facts, and the development of online news text, is a challenging task due to the varying lengths of the texts. One issue arises from the length of the text data itself, and another challenge lies in extracting valuable features, especially in news articles. In the deep learning models, the convolutional neural networks (CNNs) are capable of capturing local features in text data, but unable to capture the structural information or semantic relationships between words. Consequently, a sole CNN network often yields poor performance in text classification tasks, whereas the Gated Recurrent Unit (GRU) is adept at effectively extracting semantic information and understanding the global structural relationships present in textual data. This paper presents a solution to the problem by introducing a new text classification that integrates the strengths of CNN and GRU. The proposed hybrid models incorporate word vectorization and word dispersion in parallel. Initially, the model trains word vectors using the Word2vec model and then leverages the GRU model to capture semantic information from text sentences. Subsequently, the CNN method is employed to capture crucial semantic features, leading to classification using the SoftMax layer. Experimental findings demonstrated that the proposed hybrid GRU_CNN model outperformed and achieved accuracy 97.73% as compared to individual CNN, LSTM, and GRU models in terms of classification effectiveness and accuracy.

Convolutional Neural Network (CNN) is a deep learning algorithm that is widely used to identify and classify a digital image object. In this study the Convolutional Neural Network (CNN) is used as an algorithm that functions to identify leaf types (certain plants) based on images obtained from a public dataset provider named Daun Jamu Indonesia. The existence of image characteristics causes the assistance process to require a more detailed feature selection process. Therefore the CNN method is used in order to solve the problem. The Convolutional Neural Network (CNN) method is capable of performing image recognition by minimizing feature extraction. CNN is also reliable in processing unstructured data because it uses a multi-layered structure of artificial reasoning networks. The image recognition process is carried out by looking for the shape of the model that matches the processed data in order to get the best results. In this study, the augmentation process was carried out on the training data and validation data so that overfitting does not occur in the Convolutional Neural Network (CNN). The results obtained in this study indicate that the Convolutional Neural Network (CNN) method can identify leaf types with a measured accuracy rate of 92% using the Confusion Matrix evaluation method. It is hoped that this research can be used as a reference for the use of the Convolutional Neural Network (CNN) method for image data, especially plant leaf types.

Some researchers propose using the Convolutional Neural Network (CNN) method to classified batik images. It can extract features automatically without the need to define feature manually from the image. However, CNN’s weakness is that its accuracy is quite low, especially for small-sized datasets, compared to machine learning methods that use handcrafted feature extraction. In this research, an ensemble CNN method is proposed to improve the accuracy of the CNN method in classifying batik images. This method will train several CNN models at once, and then by voting and averaging techniques, the output label will be determined. Test results for two different datasets show this method can improve the accuracy of the CNN method and get an accuracy value of 100%. This method is also proven to extract features faster than the state-of-the-art method like MTCD+SVM, which is included in the hand-crafted feature extraction category.

The combination and cross-correlation of the upcoming Euclid data with cosmic microwave background (CMB) measurements is a source of great expectation since it will provide the largest lever arm of epochs, ranging from recombination to structure formation across the entire past light cone. In this work, we present forecasts for the joint analysis of Euclid and CMB data on the cosmological parameters of the standard cosmological model and some of its extensions. This work expands and complements the recently published forecasts based on Euclid-specific probes, namely galaxy clustering, weak lensing, and their cross-correlation. With some assumptions on the specifications of current and future CMB experiments, the predicted constraints are obtained from both a standard Fisher formalism and a posterior-fitting approach based on actual CMB data. Compared to a Euclid-only analysis, the addition of CMB data leads to a substantial impact on constraints for all cosmological parameters of the standard Λ-cold-dark-matter model, with improvements reaching up to a factor of ten. For the parameters of extended models, which include a redshift-dependent dark energy equation of state, non-zero curvature, and a phenomenological modification of gravity, improvements can be of the order of two to three, reaching higher than ten in some cases. The results highlight the crucial importance for cosmological constraints of the combination and cross-correlation of Euclid probes with CMB data.

The presence of matter in the path of relic photons causes distortions in the angular pattern of the cosmic microwave background (CMB) temperature fluctuations, modifying their properties in a slight but measurable way. Recently, the Planck Collaboration released the estimated convergence map, an integrated measure of the large-scale matter distribution that produced the weak gravitational lensing (WL) phenomenon observed in Planck CMB data. We perform exhaustive analyses of this convergence map calculating the variance in small and large regions of the sky, but excluding the area masked due to galactic contaminations, and compare them with the features expected in the set of simulated convergence maps, also released by the Planck collaboration. Our goal is to search for sky directions or regions where the WL imprints anomalous signatures to the variance estimator revealed through a $\chi^2$ analyses at a statistically significant level. In the local analysis of the Planck convergence map we identified 8 patches of the sky in disagreement, in more than 2$\sigma$, with what is observed in the average of the simulations. In contrast, in the large regions analysis we found no statistically significant discrepancies, but, interestingly, the regions with the highest $\chi^2$ values are surrounding the ecliptic poles. Thus, our results show a good agreement with the features expected by the $\Lambda$CDM concordance model, as given by the simulations. Yet, the outliers regions found here could suggest that the data still contain residual contamination, like noise, due to over- or under-estimation of systematic effects in the simulation data set.

We discuss the kinetic Sunyaev-Zel'dovich (SZ) contribution to cosmic microwave background (CMB) temperature fluctuations due to coherent rotational velocity component of electrons within halos. This effect produces a distinct dipole-like temperature distribution across the the cluster, and provides a promising way to measure the angular momentum distribution of gas inside clusters. Information obtained from such a measurement may provide new insights to the origin and evolution of angular momentum in hierarchical structure formation theory. For typical, well relaxed, clusters of mass a few times 10$^{14}$ M$_{\sun}$, the peak fluctuation is of the order a few $\mu$K, depending on the rotational velocity and the inclination angle of the rotational axis. For clusters which had underwent a recent merger, the contribution to temperature fluctuations could be even larger. This dipole signature is similar to the one produced by lensed CMB towards galaxy clusters, though the lensing contribution spans a larger angular extent than the one due to rotational scattering as the former depends on the gradient of the cluster potential. Since the lensing contribution towards clusters are aligned with the large scale CMB gradient, when higher resolution observations towards clusters are combined with a wide field CMB map, these two effects can be separated. An additional, but less important, source of confusion is the dipolar pattern produced by the moving-lens effect involving, again, the gradient of the cluster potential and the transverse velocity. The angular power spectrum of temperature anisotropies produced by the halo rotation is expected to be smaller than those due to the thermal SZ and peculiar velocity kinetic SZ effects.

The measurement of the large-scale B-mode polarization in the cosmic microwave background (CMB) is a fundamental goal of future CMB experiments. However, because of unprecedented sensitivity, future CMB experiments will be much more sensitive to any imperfect modelling of the Galactic foreground polarization in the reconstruction of the primordial B-mode signal. We compare the sensitivity to B-modes of different concepts of CMB satellite missions (LiteBIRD, COrE, COrE+, PRISM, EPIC, PIXIE) in the presence of Galactic foregrounds. In particular, we quantify the impact on the tensor-to-scalar parameter of incorrect foreground modelling in the component separation process. Using Bayesian fitting and Gibbs sampling, we perform the separation of the CMB and Galactic foreground B-modes. The recovered CMB B-mode power spectrum is used to compute the likelihood distribution of the tensor-to-scalar ratio. We focus the analysis to the very large angular scales that can be probed only by CMB space missions, i.e. the Reionization bump, where primordial B-modes dominate over spurious B-modes induced by gravitational lensing. We find that fitting a single modified blackbody component for thermal dust where the "real" sky consists of two dust components strongly bias the estimation of the tensor-to-scalar ratio by more than 5{\sigma} for the most sensitive experiments. Neglecting in the parametric model the curvature of the synchrotron spectral index may bias the estimated tensor-to-scalar ratio by more than 1{\sigma}. For sensitive CMB experiments, omitting in the foreground modelling a 1% polarized spinning dust component may induce a non-negligible bias in the estimated tensor-to-scalar ratio.

Abstract: This research aims to develop a method for detecting leaf spot disease in oil palm seedlings using Convolutional Neural Network (CNN). Leaf spot disease in oil palm seedlings can hinder growth and production. CNN has proven effective in image processing and classification, particularly in plant disease detection. In this study, we utilized a dataset of images containing oil palm seedling leaves infected with leaf spot disease and healthy leaves. We performed data processing, built a CNN model, and conducted hyperparameter tuning. The test results demonstrate that the developed CNN model achieves high accuracy in recognizing and distinguishing between oil palm seedling leaves infected with leaf spot disease and healthy ones. This research contributes to the development of plant disease detection technology that can support economic growth in the oil palm plantation sector. Keywords: Convolutional Neural Network, image processing, leaf spot disease detection, oil palm seedlings. Abstrak: Penelitian ini bertujuan untuk mengembangkan metode deteksi penyakit bercak pada bibit kelapa sawit menggunakan Convolutional Neural Network (CNN). Bibit kelapa sawit yang terinfeksi penyakit bercak dapat menghambat pertumbuhan dan produksi kelapa sawit. Metode CNN telah terbukti efektif dalam pengolahan citra dan klasifikasi, khususnya dalam deteksi penyakit pada tanaman. Dalam penelitian ini, kami menggunakan dataset citra daun bibit kelapa sawit yang terinfeksi penyakit bercak dan yang normal. Kami melakukan processing data, membangun model CNN, dan melakukan tuning hyperparameter. Hasil pengujian menunjukkan bahwa model CNN yang dikembangkan memiliki akurasi yang tinggi dalam mengenali dan membedakan citra daun bibit kelapa sawit yang terinfeksi penyakit bercak dan yang normal. Penelitian ini memberikan kontribusi dalam pengembangan teknologi deteksi penyakit tanaman yang dapat mendukung pertumbuhan ekonomi di sektor perkebunan kelapa sawit. Kata kunci: bibit kelapa sawit, Convolutional Neural Network, deteksi penyakit bercak, pengolahan citra.

One of the main obstacles for extracting the Cosmic Microwave Background (CMB) from mm/submm observations is the pollution from the main Galactic components: synchrotron, free‐free and thermal dust emission. The feasibility of using simple neural networks to extract CMB has been demonstrated on both temperature and polarization data obtained by the WMAP satellite. The main goal of this paper is to demonstrate the feasibility of neural networks for extracting the CMB signal from the Planck polarization data with high precision. Both auto‐correlation and cross‐correlation power spectra within a mask covering about 63 % of the sky have been used together with a “high pass filter” in order to minimize the influence of the remaining systematic errors in the Planck Q and U maps. Using the Planck 2015 released polarization maps, a BB power spectrum have been extracted by Multilayer Perceptron neural networks. This spectrum contains a bright feature with signal to noise ratios 4.5 within 200 ≪ l ≪ 250. The spectrum is significantly brighter than the BICEP2 2015 spectrum, with a spectral behaviour quite different from the “canonical” models (weak lensing plus B‐modes spectra with different tensor to scalar ratios). The feasibility of the neural network to remove the residual systematics from the available Planck polarization data to a high level has been demonstrated. (© 2016 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)

Understanding the learning mechanism of convolutional neural networks in spectral analysis

Temperature fluctuations of the cosmic microwave background radiation: A case of non-extensivity?