textit{What should a malicious user write next to fool a detection model?} Identifying malicious users is critical to ensure the safety and integrity of internet platforms. Several deep learning based detection models have been created. However, mali
cious users can evade deep detection models by manipulating their behavior, rendering these models of little use. The vulnerability of such deep detection models against adversarial attacks is unknown. Here we create a novel adversarial attack model against deep user sequence embedding-based classification models, which use the sequence of user posts to generate user embeddings and detect malicious users. In the attack, the adversary generates a new post to fool the classifier. We propose a novel end-to-end Personalized Text Generation Attack model, called texttt{PETGEN}, that simultaneously reduces the efficacy of the detection model and generates posts that have several key desirable properties. Specifically, texttt{PETGEN} generates posts that are personalized to the users writing style, have knowledge about a given target context, are aware of the users historical posts on the target context, and encapsulate the users recent topical interests. We conduct extensive experiments on two real-world datasets (Yelp and Wikipedia, both with ground-truth of malicious users) to show that texttt{PETGEN} significantly reduces the performance of popular deep user sequence embedding-based classification models. texttt{PETGEN} outperforms five attack baselines in terms of text quality and attack efficacy in both white-box and black-box classifier settings. Overall, this work paves the path towards the next generation of adversary-aware sequence classification models.
Intercellular heterogeneity is a major obstacle to successful personalized medicine. Single-cell RNA sequencing (scRNA-seq) technology has enabled in-depth analysis of intercellular heterogeneity in various diseases. However, its full potentials for
personalized medicine are yet to be reached. Towards this, we propose A Single-cell Guided pipeline to Aid Repurposing of Drugs (ASGARD). ASGARD can repurpose single drugs for each cell cluster and for multiple cell clusters at individual patient levels; it can also predict personalized drug combinations to address the intercellular heterogeneity within each patient. We tested ASGARD on three independent datasets, including advanced metastatic breast cancer, acute lymphoblastic leukemia, and coronavirus disease 2019 (COVID-19). On single-drug therapy, ASGARD shows significantly better average accuracy (AUC=0.95) compared to two other single-cell pipelines (AUC 0.69 and 0.57) and two other bulk-cell-based drug repurposing methods (AUC 0.80 and 0.75). The top-ranked drugs, such as fulvestrant and neratinib for breast cancer, tretinoin and vorinostat for leukemia, and chloroquine and enalapril for severe COVID19, are either approved by FDA or in clinical trials treating corresponding diseases. In conclusion, ASGARD is a promising pipeline guided by single-cell RNA-seq data, for repurposing personalized drugs and drug combinations. ASGARD is free for academic use at https://github.com/lanagarmire/ASGARD.
$alpha$-clustering structure is a significant topic in light nuclei. A Bayesian convolutional neural network (BCNN) is applied to classify initial non-clustered and clustered configurations, namely Woods-Saxon distribution and three-$alpha$ triangula
r (four-$alpha$ tetrahedral) structure for $^{12}$C ($^{16}$O), from heavy-ion collision events generated within a multi-phase transport (AMPT) model. Azimuthal angle and transverse momentum distributions of charged pions are taken as inputs to train the classifier. On multiple-event basis, the overall classification accuracy can reach $95%$ for $^{12}$C/$^{16}$O + $^{197}$Au events at $sqrt{S_{NN}} =$ 200 GeV. With proper constructions of samples, the predicted deviations on mixed samples with different proportions of both configurations could be within $5%$. In addition, setting a simple confidence threshold can further improve the predictions on the mixed dataset. Our results indicate promising and extensive possibilities of application of machine-learning-based techniques to real data and some other problems in physics of heavy-ion collisions.
Departing from the weak solution, we prove the uniqueness, smoothing estimates and the global dynamics for the non cutoff spatially homogeneous Boltzmann equation with moderate soft potentials. Our results show that the behavior of the solution(inclu
ding the production of regularity and the longtime behavior) can be {it characterized quantitatively} by the initial data at the large velocities, i.e.(i). initially polynomial decay at the large velocities in $L^1$ space will induce the finite smoothing estimates in weighted Sobolev spaces and the polynomial convergence rate (including the lower and upper bounds) to the equilibrium; (ii). initially the exponential decay at the large velocities in $L^1$ space will induce $C^infty$ regularization effect and the stretched exponential convergence rate. The new ingredients of the proof lie in the development of the localized techniques in phase and frequency spaces and the propagation of the exponential momentum.
The aerosol formation is associated with the rupture of the liquid plug during the pulmonary airway reopening. The fluid dynamics of this process is difficult to predict because the rupture involved complex liquid-gas transition. Equation of state (E
OS) plays a key role in the thermodynamic process of liquid-gas transition. Here, we propose an EOS-based multiphase lattice Boltzmann model, in which the nonideal force is directly evaluated by EOSs. This multiphase model is used to model the pulmonary airway reopening and study aerosol formation during exhalation. The numerical model is first validated with the simulations of Fujioka et al.(2008). and the result is in reasonable agreement with their study. Furthermore, two rupture cases with and without aerosol formation are contrasted and analyzed. It is found that the injury on the epithelium in the case with aerosol formation is essentially the same that of without aerosol formation even while the pressure drop in airway increases by about 67%. Then extensive simulations are performed to investigate the effects of pressure drop, thickness of liquid plug and film on aerosol size and the mechanical stresses. The results show that aerosol size and the mechanical stresses increase as the pressure drop enlarges and thickness of liquid plug become thicken, while aerosol size and the mechanical stresses decrease as thickness of liquid film is thicken. The present multiphase model can be extended to study the generation and transmission of bioaerosols which can carry the bioparticles of influenza or coronavirus.
Fact checking by professionals is viewed as a vital defense in the fight against misinformation.While fact checking is important and its impact has been significant, fact checks could have limited visibility and may not reach the intended audience, s
uch as those deeply embedded in polarized communities. Concerned citizens (i.e., the crowd), who are users of the platforms where misinformation appears, can play a crucial role in disseminating fact-checking information and in countering the spread of misinformation. To explore if this is the case, we conduct a data-driven study of misinformation on the Twitter platform, focusing on tweets related to the COVID-19 pandemic, analyzing the spread of misinformation, professional fact checks, and the crowd response to popular misleading claims about COVID-19. In this work, we curate a dataset of false claims and statements that seek to challenge or refute them. We train a classifier to create a novel dataset of 155,468 COVID-19-related tweets, containing 33,237 false claims and 33,413 refuting arguments.Our findings show that professional fact-checking tweets have limited volume and reach. In contrast, we observe that the surge in misinformation tweets results in a quick response and a corresponding increase in tweets that refute such misinformation. More importantly, we find contrasting differences in the way the crowd refutes tweets, some tweets appear to be opinions, while others contain concrete evidence, such as a link to a reputed source. Our work provides insights into how misinformation is organically countered in social platforms by some of their users and the role they play in amplifying professional fact checks.These insights could lead to development of tools and mechanisms that can empower concerned citizens in combating misinformation. The code and data can be found in http://claws.cc.gatech.edu/covid_counter_misinformation.html.
We describe a time-dependent functional involving the relative entropy and the $dot{H}^1$ seminorm, which decreases along solutions to the spatially homogeneous Landau equation with Coulomb potential. The study of this monotone functionial sheds ligh
t on the competition between the dissipation and the nonlinearity for this equation. It enables to obtain new results concerning regularity/blowup issues for the Landau equation with Coulomb potential.
Applying the theory of elliptic functions we establish two Jacobi theta function identities. From these identities we confirm two q-trigonometric identities conjectured by Gosper. As an application, we give a new and simple proof of a Pi_{q}-identity of Gosper.
It is expected in physics that the homogeneous quantum Boltzmann equation with Fermi-Dirac or Bose-Einstein statistics and with Maxwell-Boltzmann operator (neglecting effect of the statistics) for the weak coupled gases will converge to the homogeneo
us Fokker-Planck-Landau equation as the Planck constant $hbar$ tends to zero. In this paper and the upcoming work cite{HLP2}, we will provide a mathematical justification on this semi-classical limit. Key ingredients into the proofs are the new framework to catch the {it weak projection gradient}, which is motivated by Villani cite{V1} to identify the $H$-solution for Fokker-Planck-Landau equation, and the symmetric structure inside the cubic terms of the collision operators.
The well-known Rutherford differential cross section, denoted by $ dOmega/dsigma$, corresponds to a two body interaction with Coulomb potential. It leads to the logarithmically divergence of the momentum transfer (or the transport cross section) whic
h is described by $$int_{{mathbb S}^2} (1-costheta) frac{dOmega}{dsigma} dsigmasim int_0^{pi} theta^{-1}dtheta. $$ Here $theta$ is the deviation angle in the scattering event. Due to screening effect, physically one can assume that $theta_{min}$ is the order of magnitude of the smallest angles for which the scattering can still be regarded as Coulomb scattering. Under ad hoc cutoff $theta geq theta_{min}$ on the deviation angle, L. D. Landau derived a new equation in cite{landau1936transport} for the weakly interacting gas which is now referred to as the Fokker-Planck-Landau or Landau equation. In the present work, we establish a unified framework to justify Landaus formal derivation in cite{landau1936transport} and the so-called Landau approximation problem proposed in cite{alexandre2004landau} in the close-to-equilibrium regime. Precisely, (i). we prove global well-posedness of the Boltzmann equation with cutoff Rutherford cross section which is perhaps the most singular kernel both in relative velocity and deviation angle. (ii). we prove a global-in-time error estimate between solutions to Boltzmann and Landau equations with logarithm accuracy, which is consistent with the famous Coulomb logarithm. Key ingredients into the proofs of these results include a complete coercivity estimate of the linearized Boltzmann collision operator, a uniform spectral gap estimate and a novel linear-quasilinear method.
