arXiv subjects

Abdul Basit

Publications and source records attributed to Abdul Basit.

At least 37 records · Page 2Linked to original sources

arXiv2025

A Survey of Adversarial Defenses in Vision-based Systems: Categorization, Methods and Challenges

Adversarial attacks have emerged as a major challenge to the trustworthy deployment of machine learning models, particularly in computer vision applications. These attacks have a varied level of potency and can be implemented in both white box and black box approaches. Practical attacks include methods to manipulate the physical world and enforce adversarial behaviour by the corresponding target neural network models. Multiple different approaches to mitigate different kinds of such attacks are available in the literature, each with their own advantages and limitations. In this survey, we present a comprehensive systematization of knowledge on adversarial defenses, focusing on two key computer vision tasks: image classification and object detection. We review the state-of-the-art adversarial defense techniques and categorize them for easier comparison. In addition, we provide a schematic representation of these categories within the context of the overall machine learning pipeline, facilitating clearer understanding and benchmarking of defenses. Furthermore, we map these defenses to the types of adversarial attacks and datasets where they are most effective, offering practical insights for researchers and practitioners. This study is necessary for understanding the scope of how the available defenses are able to address the adversarial threats, and their shortcomings as well, which is necessary for driving the research in this area in the most appropriate direction, with the aim of building trustworthy AI systems for regular practical use-cases.

cs.CV↗

arXiv2025

DRL-Based Maximization of the Sum Cross-Layer Achievable Rate for Networks Under Jamming

In quasi-static wireless networks characterized by infrequent changes in the transmission schedules of user equipment (UE), malicious jammers can easily deteriorate network performance. Accordingly, a key challenge in these networks is managing channel access amidst jammers and under dynamic channel conditions. In this context, we propose a robust learning-based mechanism for channel access in multi-cell quasi-static networks under jamming. The network comprises multiple legitimate UEs, including predefined UEs (pUEs) with stochastic predefined schedules and an intelligent UE (iUE) with an undefined transmission schedule, all transmitting over a shared, time-varying uplink channel. Jammers transmit unwanted packets to disturb the pUEs' and the iUE's communication. The iUE's learning process is based on the deep reinforcement learning (DRL) framework, utilizing a residual network (ResNet)-based deep Q-Network (DQN). To coexist in the network and maximize the network's sum cross-layer achievable rate (SCLAR), the iUE must learn the unknown network dynamics while concurrently adapting to dynamic channel conditions. Our simulation results reveal that, with properly defined state space, action space, and rewards in DRL, the iUE can effectively coexist in the network, maximizing channel utilization and the network's SCLAR by judiciously selecting transmission time slots and thus avoiding collisions and jamming.

eess.SY↗

arXiv2024

Survey of different Large Language Model Architectures: Trends, Benchmarks, and Challenges

Large Language Models (LLMs) represent a class of deep learning models adept at understanding natural language and generating coherent responses to various prompts or queries. These models far exceed the complexity of conventional neural networks, often encompassing dozens of neural network layers and containing billions to trillions of parameters. They are typically trained on vast datasets, utilizing architectures based on transformer blocks. Present-day LLMs are multi-functional, capable of performing a range of tasks from text generation and language translation to question answering, as well as code generation and analysis. An advanced subset of these models, known as Multimodal Large Language Models (MLLMs), extends LLM capabilities to process and interpret multiple data modalities, including images, audio, and video. This enhancement empowers MLLMs with capabilities like video editing, image comprehension, and captioning for visual content. This survey provides a comprehensive overview of the recent advancements in LLMs. We begin by tracing the evolution of LLMs and subsequently delve into the advent and nuances of MLLMs. We analyze emerging state-of-the-art MLLMs, exploring their technical features, strengths, and limitations. Additionally, we present a comparative analysis of these models and discuss their challenges, potential limitations, and prospects for future development.

cs.LG↗

arXiv2024

MindArm: Mechanized Intelligent Non-Invasive Neuro-Driven Prosthetic Arm System

Currently, individuals with arm mobility impairments (referred to as "patients") face limited technological solutions due to two key challenges: (1) non-invasive prosthetic devices are often prohibitively expensive and costly to maintain, and (2) invasive solutions require high-risk, costly brain surgery, which can pose a health risk. Therefore, current technological solutions are not accessible for all patients with different financial backgrounds. Toward this, we propose a low-cost technological solution called MindArm, an affordable, non-invasive neuro-driven prosthetic arm system. MindArm employs a deep neural network (DNN) to translate brain signals, captured by low-cost surface electroencephalogram (EEG) electrodes, into prosthetic arm movements. Utilizing an Open Brain Computer Interface and UDP networking for signal processing, the system seamlessly controls arm motion. In the compute module, we run a trained DNN model to interpret filtered micro-voltage brain signals, and then translate them into a prosthetic arm action via serial communication seamlessly. Experimental results from a fully functional prototype show high accuracy across three actions, with 91% for idle/stationary, 85% for handshake, and 84% for cup pickup. The system costs approximately $500-550, including $400 for the EEG headset and $100-150 for motors, 3D printing, and assembly, offering an affordable alternative for mind-controlled prosthetic devices.

cs.AI↗

arXiv2024

Generalized Tuza's conjecture for random hypergraphs

A celebrated conjecture of Tuza states that in any finite graph the minimum size of a cover of triangles by edges is at most twice the maximum size of a set of edge-disjoint triangles. For an $r$-uniform hypergraph ($r$-graph) $G$, let $τ(G)$ be the minimum size of a cover of edges by $(r-1)$-sets of vertices, and let $ν(G)$ be the maximum size of a set of edges pairwise intersecting in fewer than $r-1$ vertices. Aharoni and Zerbib proposed the following generalization of Tuza's conjecture: $$ \text{For any $r$-graph $G$, $τ(G)/ν(G) \leq \lceil(r+1)/2\rceil$.} $$ Let $H_r(n,p)$ be the uniformly random $r$-graph on $n$ vertices. We show that, for $r \in \{3, 4, 5\}$ and any $p = p(n)$, $H_r(n,p)$ satisfies the Aharoni-Zerbib conjecture with high probability (i.e., with probability approaching 1 as $n \rightarrow \infty$). We also show that there is a $C < 1$ such that, for any $r \geq 6$ and any $p = p(n)$, $τ(H_r(n, p))/ν(H_r(n, p)) \leq C r$ with high probability. Furthermore, we may take $C < 1/2 + \varepsilon$, for any $\varepsilon > 0$, by restricting to sufficiently large $r$ (depending on $\varepsilon$).

math.CO↗

arXiv2024

Borrowing Treasures from Neighbors: In-Context Learning for Multimodal Learning with Missing Modalities and Data Scarcity

Multimodal machine learning with missing modalities is an increasingly relevant challenge arising in various applications such as healthcare. This paper extends the current research into missing modalities to the low-data regime, i.e., a downstream task has both missing modalities and limited sample size issues. This problem setting is particularly challenging and also practical as it is often expensive to get full-modality data and sufficient annotated training samples. We propose to use retrieval-augmented in-context learning to address these two crucial issues by unleashing the potential of a transformer's in-context learning ability. Diverging from existing methods, which primarily belong to the parametric paradigm and often require sufficient training samples, our work exploits the value of the available full-modality data, offering a novel perspective on resolving the challenge. The proposed data-dependent framework exhibits a higher degree of sample efficiency and is empirically demonstrated to enhance the classification model's performance on both full- and missing-modality data in the low-data regime across various multimodal learning tasks. When only 1% of the training data are available, our proposed method demonstrates an average improvement of 6.1% over a recent strong baseline across various datasets and missing states. Notably, our method also reduces the performance gap between full-modality and missing-modality data compared with the baseline.

cs.LG↗

arXiv2024

MedAide: Leveraging Large Language Models for On-Premise Medical Assistance on Edge Devices

Large language models (LLMs) are revolutionizing various domains with their remarkable natural language processing (NLP) abilities. However, deploying LLMs in resource-constrained edge computing and embedded systems presents significant challenges. Another challenge lies in delivering medical assistance in remote areas with limited healthcare facilities and infrastructure. To address this, we introduce MedAide, an on-premise healthcare chatbot. It leverages tiny-LLMs integrated with LangChain, providing efficient edge-based preliminary medical diagnostics and support. MedAide employs model optimizations for minimal memory footprint and latency on embedded edge devices without server infrastructure. The training process is optimized using low-rank adaptation (LoRA). Additionally, the model is trained on diverse medical datasets, employing reinforcement learning from human feedback (RLHF) to enhance its domain-specific capabilities. The system is implemented on various consumer GPUs and Nvidia Jetson development board. MedAide achieves 77\% accuracy in medical consultations and scores 56 in USMLE benchmark, enabling an energy-efficient healthcare assistance platform that alleviates privacy concerns due to edge-based deployment, thereby empowering the community.

cs.AI↗

arXiv2024

DRL-Based Dynamic Channel Access and SCLAR Maximization for Networks Under Jamming

This paper investigates a deep reinforcement learning (DRL)-based approach for managing channel access in wireless networks. Specifically, we consider a scenario in which an intelligent user device (iUD) shares a time-varying uplink wireless channel with several fixed transmission schedule user devices (fUDs) and an unknown-schedule malicious jammer. The iUD aims to harmoniously coexist with the fUDs, avoid the jammer, and adaptively learn an optimal channel access strategy in the face of dynamic channel conditions, to maximize the network's sum cross-layer achievable rate (SCLAR). Through extensive simulations, we demonstrate that when we appropriately define the state space, action space, and rewards within the DRL framework, the iUD can effectively coexist with other UDs and optimize the network's SCLAR. We show that the proposed algorithm outperforms the tabular Q-learning and a fully connected deep neural network approach.

eess.SP↗

arXiv2024

Effects of reservoir squeezing on the amplification of quantum correlation

The dynamics of quantum discord is studied in a system of two identical noninteracting qubits coupled to a common squeezed vacuum bath through non-demolition interactions. We concern on how reservoir squeezing influences the dynamical behaviors of quantum discord when both qubits are initially prepared in $X$-type states. We find that the critical time exhibits the sudden change of quantum discord, which is of great significance for the quantum discord amplification. Furthermore, depending on the initial parameters of the system, we numerically calculate the interval when the critical time is finite or infinite. For the finite critical time, we show that the squeezing phase of the bath can prolong the critical time while the squeezing strength exhibits the opposite effect. For infinite critical time, even if there is no sudden transition point, reservoir squeezing still has an effect on the amplification of quantum discord, and the time to reach steady-state quantum discord can be changed by adjusting the squeezing parameters. Fianlly, we investigate the quantum speed limit time for a two-qubit system under squeezed reservoir, and find that the quantum speed limit time can be reduced via the adjustment of the squeezing parameters and the initial parameters. Remarkably, in the short time limit, reservoir squeezing has an obvious influence on the degree of amplification of quantum discord. Our study presents a promising approach to controlling the amplification of quantum correlation.

quant-ph↗

arXiv2024

HgbNet: predicting hemoglobin level/anemia degree from EHR data

Anemia is a prevalent medical condition that typically requires invasive blood tests for diagnosis and monitoring. Electronic health records (EHRs) have emerged as valuable data sources for numerous medical studies. EHR-based hemoglobin level/anemia degree prediction is non-invasive and rapid but still faces some challenges due to the fact that EHR data is typically an irregular multivariate time series containing a significant number of missing values and irregular time intervals. To address these issues, we introduce HgbNet, a machine learning-based prediction model that emulates clinicians' decision-making processes for hemoglobin level/anemia degree prediction. The model incorporates a NanDense layer with a missing indicator to handle missing values and employs attention mechanisms to account for both local irregularity and global irregularity. We evaluate the proposed method using two real-world datasets across two use cases. In our first use case, we predict hemoglobin level/anemia degree at moment T+1 by utilizing records from moments prior to T+1. In our second use case, we integrate all historical records with additional selected test results at moment T+1 to predict hemoglobin level/anemia degree at the same moment, T+1. HgbNet outperforms the best baseline results across all datasets and use cases. These findings demonstrate the feasibility of estimating hemoglobin levels and anemia degree from EHR data, positioning HgbNet as an effective non-invasive anemia diagnosis solution that could potentially enhance the quality of life for millions of affected individuals worldwide. To our knowledge, HgbNet is the first machine learning model leveraging EHR data for hemoglobin level/anemia degree prediction.

cs.LG↗

arXiv2023

Partitioning axis-parallel lines in 3D

Let $L$ be a set of $n$ axis-parallel lines in $\mathbb{R}^3$. We are are interested in partitions of $\mathbb{R}^3$ by a set $H$ of three planes such that each open cell in the arrangement $\mathcal{A}(H)$ is intersected by as few lines from $L$ as possible. We study such partitions in three settings, depending on the type of splitting planes that we allow. We obtain the following results. $\bullet$ There are sets $L$ of $n$ axis-parallel lines such that, for any set $H$ of three splitting planes, there is an open cell in $\mathcal{A}(H)$ that intersects at least~$\lfloor n/3 \rfloor-1 \approx \frac{1}{3}n$ lines. $\bullet$ If we require the splitting planes to be axis-parallel, then there are sets $L$ of $n$ axis-parallel lines such that, for any set $H$ of three splitting planes, there is an open cell in $\mathcal{A}(H)$ that intersects at least $\frac{3}{2}\lfloor n/4 \rfloor -1 \approx \left( \frac{1}{3}+\frac{1}{24}\right) n$ lines. Furthermore, for any set $L$ of $n$ axis-parallel lines, there exists a set $H$ of three axis-parallel splitting planes such that each open cell in $\mathcal{A}(H)$ intersects at most $\frac{7}{18} n = \left( \frac{1}{3}+\frac{1}{18}\right) n$ lines. $\bullet$ For any set $L$ of $n$ axis-parallel lines, there exists a set $H$ of three axis-parallel and mutually orthogonal splitting planes, such that each open cell in $\mathcal{A}(H)$ intersects at most $\lceil \frac{5}{12} n \rceil \approx \left( \frac{1}{3}+\frac{1}{12}\right) n$ lines.

cs.CG↗

arXiv2023

Detection and Localization of Firearm Carriers in Complex Scenes for Improved Safety Measures

Detecting firearms and accurately localizing individuals carrying them in images or videos is of paramount importance in security, surveillance, and content customization. However, this task presents significant challenges in complex environments due to clutter and the diverse shapes of firearms. To address this problem, we propose a novel approach that leverages human-firearm interaction information, which provides valuable clues for localizing firearm carriers. Our approach incorporates an attention mechanism that effectively distinguishes humans and firearms from the background by focusing on relevant areas. Additionally, we introduce a saliency-driven locality-preserving constraint to learn essential features while preserving foreground information in the input image. By combining these components, our approach achieves exceptional results on a newly proposed dataset. To handle inputs of varying sizes, we pass paired human-firearm instances with attention masks as channels through a deep network for feature computation, utilizing an adaptive average pooling layer. We extensively evaluate our approach against existing methods in human-object interaction detection and achieve significant results (AP=77.8\%) compared to the baseline approach (AP=63.1\%). This demonstrates the effectiveness of leveraging attention mechanisms and saliency-driven locality preservation for accurate human-firearm interaction detection. Our findings contribute to advancing the fields of security and surveillance, enabling more efficient firearm localization and identification in diverse scenarios.

cs.CV↗

arXiv2023

Covering triangular grids with multiplicity

Motivated by classical work of Alon and Füredi, we introduce and address the following problem: determine the minimum number of affine hyperplanes in $\mathbb{R}^d$ needed to cover every point of the triangular grid $T_d(n) := \{(x_1,\dots,x_d)\in\mathbb{Z}_{\ge 0}^d\mid x_1+\dots+x_d\le n-1\}$ at least $k$ times. For $d = 2$, we solve the problem exactly for $k \leq 4$, and obtain a partial solution for $k > 4$. We also obtain an asymptotic formula (in $n$) for all $d \geq k - 2$. The proofs rely on combinatorial arguments and linear programming.

math.CO↗

arXiv2022

Improved bounds on a generalization of Tuza's conjecture

For an $r$-uniform hypergraph $H$, let $ν^{(m)}(H)$ denote the maximum size of a set~$M$ of edges in $H$ such that every two edges in $M$ intersect in less than $m$ vertices, and let $τ^{(m)}(H)$ denote the minimum size of a collection $C$ of $m$-sets of vertices such that every edge in $H$ contains an element of $C$. The fractional analogues of these parameters are denoted by $ν^{*(m)}(H)$ and $τ^{*(m)}(H)$, respectively. Generalizing a famous conjecture of Tuza on covering triangles in a graph, Aharoni and Zerbib conjectured that for every $r$-uniform hypergraph $H$, $τ^{(r-1)}(H)/ν^{(r-1)}(H) \leq \lceil{\frac{r+1}{2}}\rceil$. In this paper we prove bounds on the ratio between the parameters $τ^{(m)}$ and $ν^{(m)}$, and their fractional analogues. Our main result is that, for every $r$-uniform hypergraph~$H$, \[ τ^{*(r-1)}(H)/ν^{(r-1)}(H) \le \begin{cases} \frac{3}{4}r - \frac{r}{4(r+1)} &\text{for }r\text{ even,}\\ \frac{3}{4}r - \frac{r}{4(r+2)} &\text{for }r\text{ odd.} \\ \end{cases} \] This improves the known bound of $r-1$. We also prove that, for every $r$-uniform hypergraph $H$, $τ^{(m)}(H)/ν^{*(m)}(H) \le \operatorname{ex}_m(r, m+1)$, where the Turán number $\operatorname{ex}_r(n, k)$ is the maximum number of edges in an $r$-uniform hypergraph on $n$ vertices that does not contain a copy of the complete $r$-uniform hypergraph on $k$ vertices. Finally, we prove further bounds in the special cases $(r,m)=(4,2)$ and $(r,m)=(4,3)$.

math.CO↗

arXiv2021

On the number of ordinary lines determined by sets in complex space

Kelly's theorem states that a set of $n$ points affinely spanning $\mathbb{C}^3$ must determine at least one ordinary complex line (a line passing through exactly two of the points). Our main theorem shows that such sets determine at least $3n/2$ ordinary lines, unless the configuration has $n-1$ points in a plane and one point outside the plane (in which case there are at least $n-1$ ordinary lines). In addition, when at most $2n/3$ points are contained in any plane, we prove a theorem giving stronger bounds that take advantage of the existence of lines with 4 and more points (in the spirit of Melchior's and Hirzebruch's inequalities). Furthermore, when the points span 4 or more dimensions, with at most $2n/3$ points contained in any three dimensional affine subspace, we show that there must be a quadratic number of ordinary lines.

math.CO↗

arXiv2021

An improved sum-product bound for quaternions

We show that there exists an absolute constant $c > 0$, such that, for any finite set $A$ of quaternions, \[ \max\{|A+A, |AA| \} \gtrsim |A|^{4/3 + c}. \] This generalizes a sum-product bound for real numbers proved by Konyagin and Shkredov.

math.CO↗

arXiv2021

Zarankiewicz's problem for semilinear hypergraphs

A bipartite graph $H = \left(V_1, V_2; E \right)$ with $|V_1| + |V_2| = n$ is semilinear if $V_i \subseteq \mathbb{R}^{d_i}$ for some $d_i$ and the edge relation $E$ consists of the pairs of points $(x_1, x_2) \in V_1 \times V_2$ satisfying a fixed Boolean combination of $s$ linear equalities and inequalities in $d_1 + d_2$ variables for some $s$. We show that for a fixed $k$, the number of edges in a $K_{k,k}$-free semilinear $H$ is almost linear in $n$, namely $|E| = O_{s,k,\varepsilon}(n^{1+\varepsilon})$ for any $\varepsilon > 0$; and more generally, $|E| = O_{s,k,r,\varepsilon}(n^{r-1 + \varepsilon})$ for a $K_{k, \ldots,k}$-free semilinear $r$-partite $r$-uniform hypergraph. As an application, we obtain the following incidence bound: given $n_1$ points and $n_2$ open boxes with axis parallel sides in $\mathbb{R}^d$ such that their incidence graph is $K_{k,k}$-free, there can be at most $O_{k,\varepsilon}(n^{1+\varepsilon})$ incidences. The same bound holds if instead of boxes one takes polytopes cut out by the translates of an arbitrary fixed finite set of halfspaces. We also obtain matching upper and (superlinear) lower bounds in the case of dyadic boxes on the plane, and point out some connections to the model-theoretic trichotomy in $o$-minimal structures (showing that the failure of an almost linear bound for some definable graph allows one to recover the field operations from that graph in a definable manner).

math.CO↗

arXiv2021

On the independent set sequence of a tree

Alavi, Malde, Schwenk and Erdős asked whether the independent set sequence of every tree is unimodal. Here we make some observations about this question. We show that for the uniformly random (labelled) tree, asymptotically almost surely (a.a.s.) the initial approximately 49.5\% of the sequence is increasing while the terminal approximately 38.8\% is decreasing. Our approach uses the Matrix Tree Theorem, combined with computation. We also present a generalization of a result of Levit and Mandrescu, concerning the final one-third of the independent set sequence of a König-Egerváry graph.

math.CO↗