arXiv subjects

Amir Jafari

Publications and source records attributed to Amir Jafari.

At least 37 records · Page 2Linked to original sources

arXiv2022

Comparison Study Between Token Classification and Sequence Classification In Text Classification

Unsupervised Machine Learning techniques have been applied to Natural Language Processing tasks and surpasses the benchmarks such as GLUE with great success. Building language models approach achieves good results in one language and it can be applied to multiple NLP task such as classification, summarization, generation and etc as an out of box model. Among all the of the classical approaches used in NLP, the masked language modeling is the most used. In general, the only requirement to build a language model is presence of the large corpus of textual data. Text classification engines uses a variety of models from classical and state of art transformer models to classify texts for in order to save costs. Sequence Classifiers are mostly used in the domain of text classification. However Token classifiers also are viable candidate models as well. Sequence Classifiers and Token Classifier both tend to improve the classification predictions due to the capturing the context information differently. This work aims to compare the performance of Sequence Classifier and Token Classifiers and evaluate each model on the same set of data. In this work, we are using a pre-trained model as the base model and Token Classifier and Sequence Classier heads results of these two scoring paradigms with be compared..

cs.CL↗

arXiv2022

A Deep Learning Anomaly Detection Method in Textual Data

In this article, we propose using deep learning and transformer architectures combined with classical machine learning algorithms to detect and identify text anomalies in texts. Deep learning model provides a very crucial context information about the textual data which all textual context are converted to a numerical representation. We used multiple machine learning methods such as Sentence Transformers, Auto Encoders, Logistic Regression and Distance calculation methods to predict anomalies. The method are tested on the texts data and we used syntactic data from different source injected into the original text as anomalies or use them as target. Different methods and algorithm are explained in the field of outlier detection and the results of the best technique is presented. These results suggest that our algorithm could potentially reduce false positive rates compared with other anomaly detection methods that we are testing.

cs.CL↗

arXiv2022

The Kraichnan Model and Non-Equilibrium Statistical Physics of Diffusive Mixing

We discuss application of methods from the Kraichnan model of turbulent advection to the study of non-equilibrium concentration fluctuations arising during diffusion in liquid mixtures at high Schmidt numbers. This approach treats nonlinear advection of concentration fluctuations exactly, without linearization. Remarkably, we find that static and dynamic structure functions obtained by this method reproduce precisely the predictions of linearized fluctuating hydrodynamics. It is argued that this agreement is an analogue of anomaly non-renormalization which does not, however, protect higher-order multi-point correlations. The latter should thus yield non-vanishing cumulants, unlike those for the Gaussian concentration fluctuations predicted by linearized theory.

cond-mat.stat-mech↗

arXiv2022

High Schmidt-Number Turbulent Advection and Giant Concentration Fluctuations

We consider the effects of thermal noise on the Batchelor-Kraichnan theory of high Schmidt-number mixing in the viscous-dissipation range of turbulent flows. Using fluctuating hydrodynamics for a binary fluid mixture at low Mach numbers, we justify linearization around the deterministic Navier-Stokes solution in the dissipation range. For the latter solution we adopt the standard Kraichnan model and derive asymptotic high-Schmidt limiting equations for the concentration field, in which the thermal velocity fluctuations are exactly represented by a Gaussian random velocity which is white in time. We obtain the exact solution for the concentration spectrum in this high-Schmidt limiting model, showing that the Batchelor prediction in the viscous-convective range is unaltered. Thermal noise dramatically renormalizes the bare diffusivity in this range, but the effect is the same as in laminar flow and thus hidden phenomenologically. However, in the viscous-diffusive range at scales below the Batchelor length (typically micron scales) the predictions based on deterministic Navier-Stokes equations are drastically altered by thermal noise. Whereas the classical theories predict rapidly decaying spectra in the viscous-diffusive range, we obtain a $k^{-2}$ power-law spectrum starting just below the Batchelor length. This spectrum corresponds to non-equilibrium giant concentration fluctuations, due to the imposed concentration variations advected by thermal velocity fluctuations which are experimentally well-observed in quiescent fluids. At higher wavenumbers, the concentration spectrum instead goes to a $k^2$ equipartition spectrum due to equilibrium molecular fluctuations. We work out detailed predictions for water-glycerol and water-fluorescein mixtures. Finally, we discuss broad implications for turbulent flows and novel applications of our methods to experimentally accessible laminar flows.

physics.flu-dyn↗

arXiv2021

On ternary quadratic forms over the rational numbers

In this note, we give an elementary proof of the following classical fact. Any positive definite ternary quadratic form over the rational numbers fails to represent infinitely many positive integers. For any ternary quadratic form (positive definite or indefinite), our method constructs certain congruence classes whose elements, up to a square factor, are the only elements not represented over the rational numbers by that form. In the case of a positive definite ternary form, we show that these classes are non-empty. This shows that the minimum number of variables in a positive definite quadratic form representing all positive integers is four. Our proof is very elementary and only uses quadratic reciprocity of Gauss.

math.HO↗

arXiv2021

On Semi-Invariants of a Matrix

For an algebraically closed field $K$ of characteristic zero and a non-singular matrix $A\in \mbox{GL}_n(K)$, a semi-invariant polynomial of $A$ is defined to be a polynomial $p(x)=p(x_1,\dots,x_n)$ with coefficients in $K$ such that $p(xA)=λp(x)$ for some $λ\in K$. In this article, we classify all semi-invariant polynomials of $A$ in terms of a canonically constructed basis that will be made precise in the text.

math.AC↗

arXiv2021

The effects of data size on Automated Essay Scoring engines

We study the effects of data size and quality on the performance on Automated Essay Scoring (AES) engines that are designed in accordance with three different paradigms; A frequency and hand-crafted feature-based model, a recurrent neural network model, and a pretrained transformer-based language model that is fine-tuned for classification. We expect that each type of model benefits from the size and the quality of the training data in very different ways. Standard practices for developing training data for AES engines were established with feature-based methods in mind, however, since neural networks are increasingly being considered in a production setting, this work seeks to inform us as to how to establish better training data for neural networks that will be used in production.

cs.CL↗

arXiv2021

Automated essay scoring using efficient transformer-based language models

Automated Essay Scoring (AES) is a cross-disciplinary effort involving Education, Linguistics, and Natural Language Processing (NLP). The efficacy of an NLP model in AES tests it ability to evaluate long-term dependencies and extrapolate meaning even when text is poorly written. Large pretrained transformer-based language models have dominated the current state-of-the-art in many NLP tasks, however, the computational requirements of these models make them expensive to deploy in practice. The goal of this paper is to challenge the paradigm in NLP that bigger is better when it comes to AES. To do this, we evaluate the performance of several fine-tuned pretrained NLP models with a modest number of parameters on an AES dataset. By ensembling our models, we achieve excellent results with fewer parameters than most pretrained transformer-based models.

cs.CL↗

arXiv2021

Topological theory of physical fields

We study the topology associated with physical vector and scalar fields. A mathematical object, e.g., a ball, can be continuously deformed, without tearing or gluing, to make other topologically equivalent objects, e.g., a cube or a solid disk. If tearing or gluing get involved, i.e., the deformation is not continuous anymore, the initial topology will consequently change giving rise to a topologically distinct object, e.g., a torus. This simple concept in general topology may be employed in the study of physical systems described by fields. Instead of continuously deforming objects, we can take a continuously evolving field, with an appropriately defined topology, such that the topology remains unchanged in time unless the system undergoes an important physical change, e.g., a transition to a different energy state. For instance, a sudden change in the magnetic topology in an energetically relaxing plasma, a process called reconnection, strongly affects the dynamics, e.g., it is involved in launching solar flares and generating large scale magnetic fields in astrophysical objects. In this topological formalism, the magnetic topology in a plasma can spontaneously change due to the presence of dissipative terms in the induction equation which break its time symmetry. We define a topology for the vector field $\bf F$ in the phase space $(\bf x, F)$. As for scalar fields represented by a perfect fluid, e.g., the inhomogeneous inflaton or Higgs fields, the fluid velocity $\bf u$ defines the corresponding topology. The vector field topology in its corresponding phase space $(\bf x, F)$ will be preserved in time if certain conditions including time reversal invariance are satisfied by the field and its governing differential equation.

astro-ph.HE↗

arXiv2021

Magnetic topology in fluids

We study the evolution of turbulent magnetic fields from a topological point of view, invoking commonplace mathematical tools from general topology and dynamical systems theory which connect magnetic field evolution to time reversal invariance, entropy increase and the second law of thermodynamics. We show that in fact magnetic topology is well-defined only in the phase space corresponding to a dynamical system governed by the induction equation. Hence the field's topology and stochasticity can be studied in terms of the corresponding phase space trajectories rather than the field lines in real Euclidean space. In fact, our results suggest that magnetic field lines should not be taken too literally because their existence and uniqueness and more importantly continuity in time require strong mathematical conditions, hardly satisfied in astrophysical systems. As for magnetic topology change, it is shown that the phase space topology is preserved in time for a magnetic field which, besides satisfying few continuity conditions, solves a time reversal invariant induction equation. What breaks the time symmetry in the induction equation is the presence of non-ideal plasma effects at small scales such as resistivity, which results from random collisions between diffusing electrons and other particles. The small scale, stochastic disturbances produced thereby are super-linearly amplified by Richardson diffusion in the turbulent cascade, which are eventually manifested as large scale reconnection events, somehow similar to stretching quantum fluctuations during inflation to seed large scale cosmological structures. This suggests that reconnection is rooted in the second law of thermodynamics that dictates entropy increase which in turn breaks the time symmetry.

astro-ph.HE↗

arXiv2020

On the chromatic number of almost stable general Kneser hypergraphs

Let $n\ge 1$ and $s\ge 1$ be integers. An almost $s$-stable subset $A$ of $[n]=\{1,\dots,n\}$ is a subset such that for any two distinct elements $i, j\in A$, one has $|i-j|\ge s$. For a family $\cal F$ of non-empty subsets of $[n]$ and an integer $r\ge 2$, the chromatic number of the $r$-uniform Kneser hypergraph $\mbox{KG}^r({\cal F})$, whose vertex set is $\cal F$ and whose edge set is the set of $\{A_1,\dots, A_r\}$ of pairwise disjoint elements in $\cal F$, has been studied extensively in the literature and Abyazi Sani and Alishahi were able to give a lower bound for it in terms of the equatable $r$-colorability defect, $\mbox{ecd}^r({\cal F})$. In this article, the methods of Chen for the special family of all $k$-subsets of $[n]$, are modified to give lower bounds for the chromatic number of almost stable general Kneser hypergraph $\mbox{KG}^r({\cal F}_s)$ in terms of $\mbox{ecd}^s({\cal F})$. Here ${\cal F}_s$ is the collection of almost $s$-stable elements of $\cal F$. We also propose a generalization of a conjecture of Meunier.

math.CO↗

arXiv2020

Lower bounds for the chromatic number of certain Kneser-type hypergraphs

Let $n\ge 1$, $r\ge 2$, and $s\ge 0$ be integers and ${\cal P}=\{P_1,\dots, P_l\}$ be a partition of $[n]=\{1,\dots, n\}$ with $|P_i|\le r$ for $i=1,\dots, l$. Also, let $\cal F$ be a family of non-empty subsets of $[n]$. The $r$-uniform Kneser-type hypergraph $\mbox{KG}^r({\cal F}, {\cal P},s)$ is the hypergraph with the vertex set of all $\cal P$-admissible elements $F\in {\cal F}$, that is $|F\cap P_i|\le 1$ for $i=1,\dots, l$ and the edge set of all $r$-subsets $\{F_1,\dots, F_r\}$ of the vertex set that $|F_i\cap F_j|\le s$ for all $1\le i<j\le r$. In this article, we extend the equitable $r$-colorability defect $\mbox{ecd}^r({\cal F})$ of Abyazi Sani and Alishahi to the case when one allows intersection among the vertices of an edge. It will be denoted by $\mbox{ecd}^r({\cal F},s)$. We then, give (under certain assumptions) lower bounds for the chromatic number of $\mbox{KG}^r({\cal F}, {\cal P},s)$ and some of its variants in terms of $\mbox{ecd}^r({\cal F},\lfloor s/2\rfloor)$. This work generalizes many existing results in the literature of the Kneser hypergraphs. It generalizes the previous results of the current authors from the special family of all $k$-subsets of $[n]$ to a general family $\cal F$ of subsets.

math.CO↗

arXiv2020

On some topological and combinatorial lower bounds on chromatic number of Kneser type hyper graphs

In this paper, we prove a generalization of a conjecture of Erdös, about the chromatic number of certain Kneser-type hypergraphs. For integers $n,k,r,s$ with $n\ge rk$ and $2\le s\le r$, the $r$-uniform general Kneser hypergraph $\mbox{KG}^r_s(n,k)$, has all $k$-subsets of $\{1,\dots,n\}$ as the vertex set and all multi-sets $\{A_1,\dots, A_r\}$ of $k$-subsets with $s$-wise empty intersections as the edge set. The case $r=s=2$, was considers by Kneser \cite{K} in 1955, where he conjectured that its chromatic number is $n-2(k-1)$. This was finally proved by Lovász \cite{L} in 1978. The case $r>2$ and $s=2$, was considered by Erdös in 1973, and he conjectured that its chromatic number is $\left\lceil\frac{n-r(k-1)}{r-1}\right\rceil$. This conjecture was proved by Alon, Frankl and Lovász \cite{AFL} in 1986. The case where $s>2$, was considered by Sarkaria \cite{S} in 1990, where he claimed to prove a lower bound for its chromatic number which generalized all previous results. Unfortunately, an error was found by Lange and Ziegler \cite{Z'} in 2006 in the induction method of Sarkaria on the number of prime factors of $r$, and Sarkaria's proof only worked when $s$ is less than the smallest prime factor of $r$ or $s=2$. In this paper, by applying the $\mathbb Z_p$-Tucker lemma of Ziegler \cite{Z} and Meunier \cite{M}, we finally prove the general Erdös conjecture and prove the claimed result of Sarkaria for any $2\le s\le r$. We also provide another proof of a special case of this result, using methods similar to those of Alon, Frankl, and Lovász \cite{AFL} and compute the connectivity of certain simplicial complexes that might be of interest in their own right.

math.CO↗

arXiv2020

Electromagnetic Helicity in Classical Physics

This pedagogical note revisits the concept of electromagnetic helicity in classical systems. In particular, magnetic helicity and its role in mean field dynamo theories is briefly discussed highlighting the major mathematical inconsistency in most of these theories---violation of magnetic helicity conservation. A short review of kinematic dynamo theory and its classic theorems is also presented in the Appendix.

astro-ph.HE↗

arXiv2020

Nanoflare Theory Revisited

Local magnetic reversals are an inseparable part of magnetohydrodynamic (MHD) turbulence whose collective outcome on an arbitrary scale in the inertial range may lead to a global stochastic reconnection event with a rate independent of small scale physics. We show that this picture is intimately related to the nanoflare theory of the solar corona. First, we argue that due to stochastic flux freezing, a generalized version of flux freezing in turbulence, the magnetic field follows the turbulent flow in a statistical sense. Bending and stretching an initially smooth field, therefore, the turbulence generally increases the magnetic spatial complexity. Strong magnetic shears associated with such a highly tangled field can trigger local reversals and field annihilations that convert magnetic energy into kinetic and thermal energy respectively. The former maintains the turbulence, which incidentally continues to entangle the field completing the cycle, while the latter enhances the heat generation in the dissipative range. We support this theoretical picture invoking recent analytical and numerical studies which suggest a correlation between magnetic complexity and magnetic energy dissipation. The amplification of multiple local, in-phase reversals by super-linear Richardson diffusion may initiate a global reconnection at larger scales, however, even in the absence of such a global stochastic reconnection, the small scale reversals will continue to interact with the turbulence. We employ conventional scaling laws of MHD turbulence to illustrate that these local events are indeed efficient in both enhancing the turbulence and generating heat. Finally, using an MHD numerical simulation, we show that the time evolution of the magnetic complexity is statistically correlated with the kinetic energy injection rate and/or magnetic-to-thermal energy conversion rate.

astro-ph.HE↗

arXiv2020

Power and spatial complexity in stochastic reconnection

The level of spatial complexity associated with a given vector field on an arbitrary range of scales \iffalse ${\bf F(x}, t)$ can be quantified by a simple, time-dependent function $S(t)={1\over 2}(1-\hat{\bf F}_l.\hat{\bf F}_L)_{rms}$ with ${\bf F}_l$ (${\bf F}_L$) being the average field in a volume of scale $l$ ($L>l$) and the unit vector defined as $\hat{\bf F}={\bf F}/|{\bf F}|$. Thus,\fi can be quantified by a simple, scale-dependent function of time; $0\leq S(t)\leq 1$. Previous work has invoked kinetic and magnetic complexities, associated with velocity and magnetic fields ${\bf u(x}, t)$ and ${\bf B(x}, t)$, to study magnetic reconnection and diffusion in turbulent and magnetized fluids. In this paper, using the coarse-grained momentum equation, we argue that the fluid jets associated with magnetic reconnection events at an arbitrary scale $l$ in the turbulence inertial range are predominantly driven by the Lorentz force ${\bf{N}}_l=({\bf j\times B})_l-{\bf j}_l\times {\bf B}_l$. This force, is induced by the subscale currents and is analogous to the turbulent electromotive force ${\cal E}_l=({\bf u\times B})_l-{\bf u}_l\times {\bf B}_l$ in dynamo theories. Typically, high (low) magnetic complexities during reconnection imply large (small) spatial gradients for the magnetic field, i.e., strong (weak) Lorentz forces ${\bf N}_l$. Reconnection launches jets of fluid, hence the rate of change of kinetic complexity is expected to strongly correlate with the power injected by the Lorentz force ${\bf N}_l$. We test this prediction using an incompressible, homogeneous magnetohydrodynamic (MHD) simulation and associate it with previous results. It follows that the stronger (weaker) the turbulence, the more (less) complex the magnetic field and the stronger (weaker) the reconnection field and thus the ensuing reconnection.

astro-ph.HE↗

arXiv2020

Statistical Analysis of Stochastic Magnetic Fields

Previous work has introduced scale-split energy density ψfor a given vector field B in order to quantify the field stochasticity S(t). Application to turbulent magnetic fields leads to the prediction that tangling magnetic field by turbulence increases magnetic stochasticity. An increasing stochasticity in turn leads to disalignments of the coarse-grained fields at smaller scales thus they average to weaker fields at larger scales upon coarse-graining. The field's resistance against tanglement by the turbulence may lead at some point to its sudden slippage through the fluid, decreasing the stochasticity and increasing the energy density. Thus the maxima (minima) of magnetic stochasticity are expected to approximately coincide with the minima (maxima) of energy density, occurrence of which corresponds to slippage of the magnetic field through the fluid. Field-fluid slippage, on the other hand, has been already found to be intimately related to magnetic reconnection. In this paper, we test these theoretical predictions numerically using a homogeneous, incompressible magnetohydrodynamic (MHD) simulation. Apart from expected small scale deviations, possibly due to e.g., intermittency and strong field annihilation, the theoretically predicted global relationship between stochasticity and magnetic energy is observed in different sub-volumes of the simulation box. This may indicate ubiquitous local field-fluid slippage and small scale reconnection events in MHD turbulence. We also show that the maximum magnetic stochasticity, i.e., \partial_t S(t)=0 & \partial^2_t S(t)<0, leads to sudden increases in kinetic stochasticity level which may correspond to fluid jets driven by the reconnecting field lines, i.e., reconnection. This suggests a new mathematical approach to the reconnection problem.

astro-ph.HE↗

arXiv2020

3D Turbulent Reconnection: Theory, Tests and Astrophysical Implications

Magnetic reconnection, topological change in magnetic fields, is a fundamental process in magnetized plasmas. It is associated with energy release in regions of magnetic field annihilation, but this is only one facet of this process. Astrophysical flows normally have very large Reynolds numbers and are expected to be turbulent, in agreement with observations. In strong turbulence magnetic lines constantly reconnect everywhere at all scales, making magnetic reconnection an intrinsic part of turbulent cascade. We note that this is inconsistent with the usual practice of regarding magnetic lines as persistent dynamical elements. A number of theoretical, numerical, and observational studies, starting with Lazarian & Vishniac (1999), demonstrated that 3D turbulence makes magnetic reconnection fast and that these two processes are intrinsically connected. We discuss the dramatic violation of the textbook concept of magnetic flux-freezing in the presence of turbulence and demonstrate that in the presence of turbulence the plasma effects are subdominant to turbulence as far as the magnetic reconnection is concerned. This justifies an MHD-like treatment of magnetic reconnection at scales much larger than the relevant plasma scales. We discuss numerical and observational evidences supporting the turbulent reconnection model. In particular, we show that tearing reconnection is suppressed in 3D and, unlike the 2D case, the 3D reconnection induces turbulence that makes reconnection independent of resistivity. We show that turbulent reconnection dramatically affects the key astrophysical processes, e.g., star formation, turbulent dynamo, acceleration of cosmic rays. We provide criticism of the concept of "reconnection-mediated turbulence" and explain why turbulent reconnection is very different from enhanced turbulent resistivity and hyper-resistivity, and why the latter has fatal conceptual flaws.

astro-ph.HE↗