arXiv subjects

Negin Karisani

Publications and source records attributed to Negin Karisani.

10 recordsLinked to original sources

arXiv2022

Multi-View Active Learning for Short Text Classification in User-Generated Data

Mining user-generated data often suffers from the lack of enough labeled data, short document lengths, and the informal user language. In this paper, we propose a novel active learning model to overcome these obstacles in the tasks tailored for query phrases--e.g., detecting positive reports of natural disasters. Our model has three novelties: 1) It is the first approach to employ multi-view active learning in this domain. 2) It uses the Parzen-Rosenblatt window method to integrate the representativeness measure into multi-view active learning. 3) It employs a query-by-committee strategy, based on the agreement between predictors, to address the usually noisy language of the documents in this domain. We evaluate our model in four publicly available Twitter datasets with distinctly different applications. We also compare our model with a wide range of baselines including those with multiple classifiers. The experiments testify that our model is highly consistent and outperforms existing models.

cs.CL↗

arXiv2022

Efficient simplicial replacement of semi-algebraic sets

We prove that for any $\ell \geq 0$, there exists an algorithm which takes as input a description of a semi-algebraic subset $S \subset \mathbb{R}^k$ given by a quantifier-free first order formula $ϕ$ in the language of the reals, and produces as output a simplicial complex $Δ$, whose geometric realization, $|Δ|$ is $\ell$-equivalent to $S$. The complexity of our algorithm is bounded by $(sd)^{k^{O(\ell)}}$, where $s$ is the number of polynomials appearing in the formula $ϕ$, and $d$ a bound on their degrees. For fixed $\ell$, this bound is singly exponential in $k$. In particular, since $\ell$-equivalence implies that the homotopy groups up to dimension $\ell$ of $|Δ|$ are isomorphic to those of $S$, we obtain a reduction (having singly exponential complexity) of the problem of computing the first $\ell$ homotopy groups of $S$ to the combinatorial problem of computing the first $\ell$ homotopy groups of a finite simplicial complex of size bounded by $(sd)^{k^{O(\ell)}}$.

math.AT↗

arXiv2022

Sequents, barcodes, and homology

We consider the problem of generating hypothesis from data based on ideas from logic. We introduce a notion of barcodes, which we call sequent barcodes, that mirrors the barcodes in persistent homology theory in topological data analysis. We prove a theoretical result on the stability of these barcodes in analogy with similar results in persistent homology theory. Additionally we show that our new notion of barcodes can be interpreted in terms of a persistent homology of a particular filtration of topological spaces induced by the data. Finally, we discuss a concrete application of the sequent barcodes in a discovery problem arising from the area of cancer genomics.

math.AT↗

arXiv2022

Computing the homology functor on semi-algebraic maps and diagrams

Developing an algorithm for computing the Betti numbers of semi-algebraic sets with singly exponential complexity has been a holy grail in algorithmic semi-algebraic geometry and only partial results are known. In this paper we consider the more general problem of computing the image under the homology functor of a semi-algebraic map $f:X \rightarrow Y$ between closed and bounded semi-algebraic sets. For every fixed $\ell \geq 0$ we give an algorithm with singly exponential complexity that computes bases of the homology groups $\mathrm{H}_i(X), \mathrm{H}_i(Y)$ (with rational coefficients) and a matrix with respect to these bases of the induced linear maps $\mathrm{H}_i(f):\mathrm{H}_i(X) \rightarrow \mathrm{H}_i(Y), 0 \leq i \leq \ell$. We generalize this algorithm to more general (zigzag) diagrams of maps between closed and bounded semi-algebraic sets and give a singly exponential algorithm for computing the homology functors on such diagrams. This allows us to give an algorithm with singly exponential complexity for computing barcodes of semi-algebraic zigzag persistent homology in small dimensions.

math.AT↗

arXiv2022

Persistent homology of semi-algebraic sets

We give an algorithm with singly exponential complexity for computing the barcodes up to dimension $\ell$ (for any fixed $\ell \geq 0$) of the filtration of a given semi-algebraic set by the sub-level sets of a given polynomial. Our algorithm is the first algorithm for this problem with singly exponential complexity, and generalizes the corresponding results for computing the Betti numbers up to dimension $\ell$ of semi-algebraic sets with no filtration present.

math.AT↗

arXiv2022

Inferring COVID-19 Biological Pathways from Clinical Phenotypes via Topological Analysis

COVID-19 has caused thousands of deaths around the world and also resulted in a large international economic disruption. Identifying the pathways associated with this illness can help medical researchers to better understand the properties of the condition. This process can be carried out by analyzing the medical records. It is crucial to develop tools and models that can aid researchers with this process in a timely manner. However, medical records are often unstructured clinical notes, and this poses significant challenges to developing the automated systems. In this article, we propose a pipeline to aid practitioners in analyzing clinical notes and revealing the pathways associated with this disease. Our pipeline relies on topological properties and consists of three steps: 1) pre-processing the clinical notes to extract the salient concepts, 2) constructing a feature space of the patients to characterize the extracted concepts, and finally, 3) leveraging the topological properties to distill the available knowledge and visualize the result. Our experiments on a publicly available dataset of COVID-19 clinical notes testify that our pipeline can indeed extract meaningful pathways.

cs.CL↗

arXiv2021

Semi-Supervised Text Classification via Self-Pretraining

We present a neural semi-supervised learning model termed Self-Pretraining. Our model is inspired by the classic self-training algorithm. However, as opposed to self-training, Self-Pretraining is threshold-free, it can potentially update its belief about previously labeled documents, and can cope with the semantic drift problem. Self-Pretraining is iterative and consists of two classifiers. In each iteration, one classifier draws a random set of unlabeled documents and labels them. This set is used to initialize the second classifier, to be further trained by the set of labeled documents. The algorithm proceeds to the next iteration and the classifiers' roles are reversed. To improve the flow of information across the iterations and also to cope with the semantic drift problem, Self-Pretraining employs an iterative distillation process, transfers hypotheses across the iterations, utilizes a two-stage training model, uses an efficient learning rate schedule, and employs a pseudo-label transformation heuristic. We have evaluated our model in three publicly available social media datasets. Our experiments show that Self-Pretraining outperforms the existing state-of-the-art semi-supervised classifiers across multiple settings. Our code is available at https://github.com/p-karisani/self_pretraining.

cs.CL↗

arXiv2020

Mining Coronavirus (COVID-19) Posts in Social Media

World Health Organization (WHO) characterized the novel coronavirus (COVID-19) as a global pandemic on March 11th, 2020. Before this and in late January, more specifically on January 27th, while the majority of the infection cases were still reported in China and a few cruise ships, we began crawling social media user postings using the Twitter search API. Our goal was to leverage machine learning and linguistic tools to better understand the impact of the outbreak in China. Unlike our initial expectation to monitor a local outbreak, COVID-19 rapidly spread across the globe. In this short article we report the preliminary results of our study on automatically detecting the positive reports of COVID-19 from social media user postings using state-of-the-art machine learning models.

cs.CL↗

arXiv2014

On the star arboricity of hypercubes

A Hypercube $Q_n$ is a graph in which the vertices are all binary vectors of length n, and two vertices are adjacent if and only if their components differ in exactly one place. A galaxy or a star forest is a union of vertex disjoint stars. The star arboricity of a graph $G$, ${\rm sa}(G)$, is the minimum number of galaxies which partition the edge set of $G$. In this paper among other results, we determine the exact values of ${\rm sa}(Q_n)$ for $n \in \{2^k-3, 2^k+1, 2^k+2, 2^i+2^j-4\}$, $i \geq j \geq 2$. We also improve the last known upper bound of ${\rm sa}(Q_n)$ and show the relation between ${\rm sa}(G)$ and square coloring.

math.CO↗

arXiv2013

On the construction of tree decompositions of hypercubes

There are different concepts regarding to tree decomposition of a graph $G$. For the Hypercube $Q_n$, these concepts have been shown to have many applications. But some diverse papers on this subject make it difficult to follow what is precisely known. In this note first we will mention some known results on the tree decomposition of hypercubes and then introduce new explicit constructions for the previously known and unknown cases.

math.CO↗