In supervised machine learning, feature selection plays a very important role by potentially enhancing explainability and performance as measured by computing time and accuracy-related metrics. In this paper, we investigate a method for feature selection based on the well-known L1 and L2 regularization strategies associated with logistic regression (LR). It is well known that the learned coefficients, which serve as weights, can be used to rank the features. Our approach is to synthesize the findings of L1 and L2 regularization. For our experiment, we chose the CIC-IDS2018 dataset owing partly to its size and also to the existence of two problematic classes that are hard to separate. We report first with the exclusion of one of them and then with its inclusion. We ranked features first with L1 and then with L2, and then compared logistic regression with L1 (LR+L1) against that with L2 (LR+L2) by varying the sizes of the feature sets for each of the two rankings. We found no significant difference in accuracy between the two methods once the feature set is selected. We chose a synthesis, i.e., only those features that were present in both the sets obtained from L1 and that from L2, and experimented with it on more complex models like Decision Tree and Random Forest and observed that the accuracy was very close in spite of the small size of the feature set. Additionally, we also report on the standard metrics: accuracy, precision, recall, and f1-score.

We present high-resolution rotation curves and mass models of 26 dwarf galaxies from LITTLE THINGS. LITTLE THINGS is a high-resolution Very Large Array HI survey for nearby dwarf galaxies in the local volume within 11 Mpc. The rotation curves of the sample galaxies derived in a homogeneous and consistent manner are combined with Spitzer archival 3.6 micron and ancillary optical U, B, and V images to construct mass models of the galaxies. We decompose the rotation curves in terms of the dynamical contributions by baryons and dark matter halos, and compare the latter with those of dwarf galaxies from THINGS as well as Lambda CDM SPH simulations in which the effect of baryonic feedback processes is included. Being generally consistent with THINGS and simulated dwarf galaxies, most of the LITTLE THINGS sample galaxies show a linear increase of the rotation curve in their inner regions, which gives shallower logarithmic inner slopes alpha of their dark matter density profiles. The mean value of the slopes of the 26 LITTLE THINGS dwarf galaxies is alpha =-0.32 +/- 0.24 which is in accordance with the previous results found for low surface brightness galaxies (alpha = -0.2 +/- 0.2) as well as the seven THINGS dwarf galaxies (alpha =-0.29 +/- 0.07). However, this significantly deviates from the cusp-like dark matter distribution predicted by dark-matter-only Lambda CDM simulations. Instead our results are more in line with the shallower slopes found in the Lambda CDM SPH simulations of dwarf galaxies in which the effect of baryonic feedback processes is included. In addition, we discuss the central dark matter distribution of DDO 210 whose stellar mass is relatively low in our sample to examine the scenario of inefficient supernova feedback in low mass dwarf galaxies predicted from recent Lambda SPH simulations of dwarf galaxies where central cusps still remain.

[Abridged] We use the ATLAS3D sample of 260 early-type galaxies (ETGs) to study the apparent kinematic misalignment angle, Psi, defined as the angle between the photometric and kinematic major axis. We find that 71% of nearby ETGs are strictly aligned systems (Psi > 5 deg), an additional 14% have 5 < Psi < 10 deg and 90% of galaxies have Psi < 15 deg. Taking into account measurement uncertainties, 90% of galaxies can be considered aligned to better than 5 deg, suggesting that only a small fraction of early-type galaxies (~10%) are not consistent with axisymmetry within the projected half-light radius. We identify morphological features such as bars and rings (30%), dust structures (16%), blue nuclear colours (6%) and evidence of interactions (8%). We use kinemetry to analyse the mean velocity maps and separate galaxies in two broad types of regular and non-regular rotators. We find 82% of regular rotators and 17% non-regular rotators, with 2 galaxies that we were not able to classify due to data quality. The non-regular rotators are typically found in dense regions and are massive. The majority of galaxies do not have any specific kinemetric features, but we highlight here the frequency of the kinematically distinct cores (7%) and the aligned double peaks in the velocity dispersion maps (4%). Most of the galaxies that are misaligned have complex kinematics and are non-regular rotators. While the trends are weak, there is a tendency that large values of Psi are found in galaxies at intermediate environmental densities and among the most massive galaxies in the sample. We suggest that the most common formation mechanism for ETGs preserves the axisymmetry of the disk progenitors. Less commonly, the formation process results in a triaxial galaxy with much lower net angular momentum.

Stellar feedback drives multiphase gas outflows from starburst galaxies, but the interpretation of dust emission in these winds remains uncertain. To investigate this, we analyze new JWST mid-infrared images tracing polycyclic aromatic hydrocarbon (PAH) emission at 7.7 and 11.3~$\mu$m from the outflow of the prototypical starburst M82 out to $3.2$ kpc. We find that PAH emission shows significant correlations with CO, H$\alpha$, and X-ray emission within the outflow, though the strengths and behaviors of these correlations vary with gas phase and distance from the starburst. PAH emission correlates strongly with cold molecular gas, with PAH--CO scaling relations in the wind nearly identical to those in galaxy disks despite the very different conditions. The H$\alpha$--PAH correlation indicates that H$\alpha$ traces the surfaces of PAH-bearing clouds, consistent with arising from ionized layers produced by shocks. Meanwhile the PAH--X-ray correlation disappears once distance effects are controlled for past 2~kpc, suggesting that PAHs are decoupled from the hot gas and the global correlation merely reflects the large-scale structure of the outflow. The PAH-to-neutral gas ratio remains nearly flat to 2~kpc, with variations following changes in the radiation field. This implies that the product of PAH abundance and dust-to-gas ratio does not change significantly over the inner portion of the outflow. Together, these results demonstrate that PAHs robustly trace the cold phase of M82's wind, surviving well beyond the starburst and providing a powerful, high-resolution proxy for mapping the life cycle of entrained cold material in galactic outflows.

The pattern speeds of NGC 3031, NGC 2366, and DDO 154 are measured using a solution of the Tremaine-Weinberg equations derived in a previous paper. Four different data sets of NGC 3031 produce consistent results despite differences in angular resolution, spectral resolution, and sensitivities to structures on different scales. The results for NGC 3031 show that the pattern speed is more similar to the material speed than it is to the speed of a rigidly rotating pattern, and that there are no clear indications of unique corotation or Lindblad resonances. Unlike NGC 3031, the results for NGC 2366 and DDO 154 show clear departures from the material speed. The results for NGC 2366 and DDO 154 also show that the solution method can produce meaningful results that are simple to interpret even if there is not a coherent or well-defined pattern in the data. The angular resolution of a data set has the greatest affect on the results, especially for determining the radial behavior of the pattern speed, and whether there is a single, global pattern speed.

Knowledge distillation is a popular machine learning technique that aims to transfer knowledge from a large 'teacher' network to a smaller 'student' network and improve the student's performance by training it to emulate the teacher. In recent years, there has been significant progress in novel distillation techniques that push performance frontiers across multiple problems and benchmarks. Most of the reported work focuses on achieving state-of-the-art results on the specific problem. However, there has been a significant gap in understanding the process and how it behaves under certain training scenarios. Similarly, transfer learning (TL) is an effective technique in training neural networks on a limited dataset faster by reusing representations learned from a different but related problem. Despite its effectiveness and popularity, there has not been much exploration of knowledge distillation on transfer learning. In this thesis, we propose a machine learning architecture we call TL+KD that combines knowledge distillation with transfer learning; we then present a quantitative and qualitative comparison of TL+KD with TL in the domain of image classification. Through this work, we show that using guidance and knowledge from a larger teacher network during fine-tuning, we can improve the student network to achieve better validation performances like accuracy. We characterize the improvement in the validation performance of the model using a variety of metrics beyond just accuracy scores, and study its performance in scenarios such as input degradation.

Evidence has emerged for a stochastic signal correlated among 67 pulsars within the 15-year pulsar-timing data set compiled by the NANOGrav collaboration. Similar signals have been found in data from the European, Indian, Parkes, and Chinese PTAs. This signal has been interpreted as indicative of the presence of a nanohertz stochastic gravitational wave background. To explore the internal consistency of this result we investigate how the recovered signal strength changes as we remove the pulsars one by one from the data set. We calculate the signal strength using the (noise-marginalized) optimal statistic, a frequentist metric designed to measure correlated excess power in the residuals of the arrival times of the radio pulses. We identify several features emerging from this analysis that were initially unexpected. The significance of these features, however, can only be assessed by comparing the real data to synthetic data sets. After conducting identical analyses on simulated data sets, we do not find anything inconsistent with the presence of a stochastic gravitational wave background in the NANOGrav 15-year data. The methodologies developed here can offer additional tools for application to future, more sensitive data sets. While this analysis provides an internal consistency check of the NANOGrav results, it does not eliminate the necessity for additional investigations that could identify potential systematics or uncover unmodeled physical phenomena in the data.

Evidence for a low-frequency stochastic gravitational wave background has recently been reported based on analyses of pulsar timing array data. The most likely source of such a background is a population of supermassive black hole binaries, the loudest of which may be individually detected in these datasets. Here we present the search for individual supermassive black hole binaries in the NANOGrav 15-year dataset. We introduce several new techniques, which enhance the efficiency and modeling accuracy of the analysis. The search uncovered weak evidence for two candidate signals, one with a gravitational-wave frequency of $\sim$4 nHz, and another at $\sim$170 nHz. The significance of the low-frequency candidate was greatly diminished when Hellings-Downs correlations were included in the background model. The high-frequency candidate was discounted due to the lack of a plausible host galaxy, the unlikely astrophysical prior odds of finding such a source, and since most of its support comes from a single pulsar with a commensurate binary period. Finding no compelling evidence for signals from individual binary systems, we place upper limits on the strain amplitude of gravitational waves emitted by such systems.

Varying dynamics pose a fundamental difficulty when deploying safe control laws in the real world. Safety Index Synthesis (SIS) deeply relies on the system dynamics and once the dynamics change, the previously synthesized safety index becomes invalid. In this work, we show the real-time efficacy of Safety Index Adaptation (SIA) in varying dynamics. SIA enables real-time adaptation to the changing dynamics so that the adapted safe control law can still guarantee 1) forward invariance within a safe region and 2) finite time convergence to that safe region. This work employs SIA on a package-carrying quadruped robot, where the payload weight changes in real-time. SIA updates the safety index when the dynamics change, e.g., a change in payload weight, so that the quadruped can avoid obstacles while achieving its performance objectives. Numerical study provides theoretical guarantees for SIA and a series of hardware experiments demonstrate the effectiveness of SIA in real-world deployment in avoiding obstacles under varying dynamics.

The polar regions of Jupiter host a myriad of dynamically interesting phenomena including vortex configurations, folded-filamentary regions (FFRs), and chaotic flows. Juno observations have provided unprecedented views of the high latitudes, allowing for more constraints to be placed upon the troposphere and the overall atmospheric energy cycle. Moist convective events are believed to be the primary drivers of energetic storm behavior as observed on the planet. Here, we introduce a novel single layer shallow water model to investigate the effects of polar moist convective events at high resolution, the presence of dynamical instabilities over long timescales, and the emergence of FFRs at high latitudes. We use a flexible, highly parallelizable, finite-difference hydrodynamic code to explore the parameter space set up by previous models. We study the long term effects of deformation length (Ld), injected pulse size, and injected geopotential. We find that models with Ld beyond 1500 km (planetary Burger number, Bu$=4.4\times10^{-4}$) tend to homogenize their potential vorticity (PV) in the form of dominant stable polar cyclones, while lower Ld cases tend to show less stability with regards to Arnol'd-type flows. We also find that large turbulent forcing scales consistently lead to the formation of high latitude FFRs. Our findings support the idea that moist convection, occurring at high latitudes, may be sufficient to produce the dynamical variety seen at the Jovian poles. Additionally, derived values of localized horizontal shear and Ld may constrain FFR formation and evolution.

We present new observations of the central 1 kpc of the M 82 starburst obtained with the James Webb Space Telescope (JWST) near-infrared camera (NIRCam) instrument at a resolution ~0.05"-0.1" (~1-2 pc). The data comprises images in three mostly continuum filters (F140M, F250M, and F360M), and filters that contain [FeII] (F164N), H2 v=1-0 (F212N), and the 3.3 um PAH feature (F335M). We find prominent plumes of PAH emission extending outward from the central starburst region, together with a network of complex filamentary substructure and edge-brightened bubble-like features. The structure of the PAH emission closely resembles that of the ionized gas, as revealed in Paschen alpha and free-free radio emission. We discuss the origin of the structure, and suggest the PAHs are embedded in a combination of neutral, molecular, and photoionized gas.

We look at the morphology of fast and slow rotator early-type galaxies. Edge-on fast rotators are lenticular galaxies. They appear like spiral galaxies with the gas and dust removed, and in some cases are flat ellipticals with disky isophotes. Fast rotators are often barred and span the same full range of bulge fractions as spiral galaxies. The slow rotators are rounder and are generally consistent with being genuine, namely spheroidal-like, elliptical galaxies. We propose a revision to the tuning-fork diagram by Hubble as it gives a misleading description of ETGs. We study for the first time the kinematic morphology-density T-Sigma relation using fast and slow rotators to replace lenticulars and ellipticals. We find that our relation is cleaner than using classic morphology. Slow rotators are nearly absent at the lowest density environments [f(SR)<2%] and generally constitute a small fraction [f(SR)~4%] of the total galaxy population in the environments explored by our survey, with the exception of the densest core of the Virgo cluster [f(SR)~20%]. We find a clean log-linear relation between the fraction f(Sp) of spiral galaxies and the local galaxy surface density. The existence of a smooth kinematic T-Sigma relation in the field excludes processes related to the cluster environment as main contributors to the apparent conversion of spirals into fast-rotators in low-density environments. It shows that the segregation is driven by local effects at the small-group scale. Only at the largest densities in the Virgo core does the f(Sp) relation break down and steepens sharply, while the fraction of slow-rotators starts to significantly increase. This suggests that a different mechanism is at work there. (Abridged)

Previous observations of dark vortices in Neptune's atmosphere, such as Voyager-2's Great Dark Spot, have been made in only a few, broad-wavelength channels, which has hampered efforts to pinpoint their pressure level and what makes them dark. Here, we present Very Large Telescope (Chile) MUSE spectrometer observations of Hubble Space Telescope's NDS-2018 dark spot, made in 2019. These medium-resolution 475 - 933 nm reflection spectra allow us to show that dark spots are caused by a darkening at short wavelengths (< 700 nm) of a deep ~5-bar aerosol layer, which we suggest is the H$_2$S condensation layer. A deep bright spot, named DBS-2019, is also visible on the edge of NDS-2018, whose spectral signature is consistent with a brightening of the same 5-bar layer at longer wavelengths (> 700 nm). This bright feature is much deeper than previously studied dark spot companion clouds and may be connected with the circulation that generates and sustains such spots.

The Atlas3D project is a multi-wavelength survey combined with a theoretical modeling effort. The observations span from the radio to the millimeter and optical, and provide multi-colour imaging, two-dimensional kinematics of the atomic (HI), molecular (CO) and ionized gas (Hbeta, [OIII] and [NI]), together with the kinematics and population of the stars (Hbeta, Fe5015 and Mgb), for a carefully selected, volume-limited (1.16*10^5 Mpc^3) sample of 260 early-type (elliptical E and lenticular S0) galaxies (ETGs). The models include semi-analytic, N-body binary mergers and cosmological simulations of galaxy formation. Here we present the science goals for the project and introduce the galaxy sample and the selection criteria. The sample consists of nearby (D<42 Mpc) morphologically-selected ETGs extracted from a parent sample of 871 galaxies (8% E, 22% S0 and 70% spirals) brighter than M_K<-21.5 mag (stellar mass M_Star>6*10^9 M_Sun). We analyze possible selection biases and we conclude that the parent sample is essentially complete and statistically representative of the nearby galaxy population. We present the size-luminosity relation for the spirals and ETGs and show that the ETGs in the Atlas3D sample define a tight red sequence in a colour-magnitude diagram, with few objects in the transition from the blue cloud. We describe the strategy of the SAURON integral-field observations and the extraction of the stellar kinematics with the pPXF method. We give an overview of the characteristics of the other main datasets already available for our sample and of the ongoing modelling projects.

We present the first catalog of targeted searches for continuous gravitational waves (CWs) from 114 active galactic nuclei (AGN) that may host supermassive black hole binaries (SMBHBs), using the NANOGrav 15 yr data set. By incorporating electromagnetic priors on sky location, distance, redshift, and CW frequency, our strain and chirp mass upper limits are on average 2.6$\times$ more constraining than sky-averaged limits. Bayesian model comparisons against a common uncorrelated red noise for the gravitational wave background (GWB) disfavor a CW signal for almost all targets, yielding a mean Bayes factor of $0.87 \pm 0.31$. There are two notable exceptions: SDSS J153636.22+044127.0, ``Rohan'' with $\mathrm{BF} = 3.37(5)$, and SDSS J072908.71+400836.6, ``Gondor'' with $\mathrm{BF} = 2.44(3)$. These Bayes factors correspond to p-values of $0.01$--$0.03$ ($1.9\sigma$--$2.3\sigma$) and $0.05$--$0.08$ ($1.4\sigma$--$1.6\sigma$), respectively, depending on the empirical null distribution. We outline the beginnings of a detection protocol by identifying and carrying out a battery of tests on Rohan and Gondor to verify their binary nature. Notably, when replacing the common uncorrelated red noise model with a Hellings--Downs correlated GWB, Rohan's Bayes factor drops to $1.25(7)$, while Gondor's increases to $3.2(1)$. Both have rich electromagnetic datasets, including optical and infrared variability and spectroscopic features that support their classification as SMBHB candidates, though this was discovered after the targeted searches were complete. Our results suggest more simulations are needed to confirm or refute the nature of these and future SMBHB candidates, while creating a roadmap for targeted CW detection.

We present observations and timing analyses of 68 millisecond pulsars (MSPs) comprising the 15-year data set of the North American Nanohertz Observatory for Gravitational Waves (NANOGrav). NANOGrav is a pulsar timing array (PTA) experiment that is sensitive to low-frequency gravitational waves. This is NANOGrav's fifth public data release, including both "narrowband" and "wideband" time-of-arrival (TOA) measurements and corresponding pulsar timing models. We have added 21 MSPs and extended our timing baselines by three years, now spanning nearly 16 years for some of our sources. The data were collected using the Arecibo Observatory, the Green Bank Telescope, and the Very Large Array between frequencies of 327 MHz and 3 GHz, with most sources observed approximately monthly. A number of notable methodological and procedural changes were made compared to our previous data sets. These improve the overall quality of the TOA data set and are part of the transition to new pulsar timing and PTA analysis software packages. For the first time, our data products are accompanied by a full suite of software to reproduce data reduction, analysis, and results. Our timing models include a variety of newly detected astrometric and binary pulsar parameters, including several significant improvements to pulsar mass constraints. We find that the time series of 23 pulsars contain detectable levels of red noise, 10 of which are new measurements. In this data set, we find evidence for a stochastic gravitational-wave background.

Anomaly-based cyber threat detection using deep learning is on a constant growth in popularity for novel cyber-attack detection and forensics. A robust, efficient, and real-time threat detector in a large-scale operational enterprise network requires high accuracy, high fidelity, and a high throughput model to detect malicious activities. Traditional anomaly-based detection models, however, suffer from high computational overhead and low detection accuracy, making them unsuitable for real-time threat detection. In this work, we propose LogSHIELD, a highly effective graph-based anomaly detection model in host data. We present a real-time threat detection approach using frequency-domain analysis of provenance graphs. To demonstrate the significance of graph-based frequency analysis we proposed two approaches. Approach-I uses a Graph Neural Network (GNN) LogGNN and approach-II performs frequency domain analysis on graph node samples for graph embedding. Both approaches use a statistical clustering algorithm for anomaly detection. The proposed models are evaluated using a large host log dataset consisting of 774M benign logs and 375K malware logs. LogSHIELD explores the provenance graph to extract contextual and causal relationships among logs, exposing abnormal activities. It can detect stealthy and sophisticated attacks with over 98% average AUC and F1 scores. It significantly improves throughput, achieves an average detection latency of 0.13 seconds, and outperforms state-of-the-art models in detection time.

AGN feedback is invoked as one of the most relevant mechanisms that shape the evolution of galaxies. Our goal is to understand the interplay between AGN feedback and the interstellar medium in M51, a nearby spiral galaxy with a modest AGN and a kpc-scale radio jet expanding through the disc of the galaxy. For that purpose, we combine molecular gas observations in the CO(1-0) and HCN(1-0) lines from the Plateau de Bure interferometer with archival radio, X-ray, and optical data. We show that there is a significant scarcity of CO emission in the ionisation cone, while molecular gas emission tends to accumulate towards the edges of the cone. The distribution and kinematics of CO and HCN line emission reveal AGN feedback effects out to r~500pc, covering the whole extent of the radio jet, with complex kinematics in the molecular gas which displays strong local variations. We propose that this is the result of the almost coplanar jet pushing on molecular gas in different directions as it expands; the effects are more pronounced in HCN than in CO emission, probably as the result of radiative shocks. Following previous interpretation of the redshifted molecular line in the central 5" as caused by a molecular outflow, we estimate the outflow rates to be Mdot_H2~0.9Msun/yr and Mdot_dense~0.6Msun/yr, which are comparable to the molecular inflow rates (~1Msun/yr); gas inflow and AGN feedback could be mutually regulated processes. The agreement with findings in other nearby radio galaxies suggests that this is not an isolated case, and probably the paradigm of AGN feedback through radio jets, at least for galaxies hosting low-luminosity active nuclei.