arXiv subjects

Zahed Rahmati

Publications and source records attributed to Zahed Rahmati.

14 recordsLinked to original sources

arXiv2026

Institutional Equity Holdings Prediction Using Node Affinities of Dynamic Graphs

Institutional equity holdings disclosed in SEC Form 13F filings provide a rich temporal record of portfolio decisions by large investment managers. However, forecasting future allocations and modeling future demand remains challenging due to disclosure lags, reporting noise, and strong persistence in institutional behavior. We introduce the first benchmark for these tasks using temporal graph machine learning, framing holdings prediction as node affinity prediction -- i.e., forecasting portfolio weights -- on a discrete-time temporal bipartite graph of managers and securities extracted from preprocessed filings. On a sampled dataset comprising 99 managers and the S\&P 500 index (503 securities, 209,351 temporal edges across 48 quarters from 2013--2025), Node Affinity prediction model using Virtual State (NAVIS) achieves a state-of-the-art test Normalized Discounted Cumulative Gain (NDCG) of 0.9127 with features (0.9121 without), outperforming all dynamic graph representation learning competitors by a substantial margin, and outperforming all heuristic methods. Remarkably, a simple Exponential Moving Average baseline achieves 0.8882, surpassing all dynamic graph models except NAVIS and all heuristics except Persistent Forecast (0.8891), highlighting the strong smoothness and persistence of institutional portfolios. Domain-specific node features provide only marginal gains (<1.2\%), indicating that temporal and structural signals in the 13F ownership graph already capture most of the predictable information. By benchmarking a suite of Temporal Graph Benchmark (TGB) models under the node affinity prediction setting, both with and without features, on real-world 13F data, this work provides a reproducible foundation for temporal graph machine learning in holdings prediction and portfolio allocation.

cs.LG↗

arXiv2026

Enhancing Graph Neural Networks Using Proximity Graphs for Dust Source Emission Forecasting

Accurate prediction of dust source emissions is critical for mitigating the significant environmental and health hazards posed by dust storms. Traditional forecasting methods often struggle to capture the complex spatiotemporal dynamics of these phenomena. In this paper, we demonstrate that proximity graphs enable Graph Neural Networks (GNNs) to effectively model the intricate spatial and temporal relationships between data points. Specifically, we use proximity graphs--such as Delaunay triangulation, Gabriel graph, k-Nearest Neighbor graph, and Yao graph--as the input for GNNs (including GraphSAGE, Graph Convolutional Networks, and Graph Attention Networks) to perform message passing. Our approach highlights the effectiveness of integrating proximity graphs with GNNs for robust and accurate dust source forecasting. To emphasize the importance of proximity graph representations, we compare our method against GNNs using random graphs for message passing. The results show that GNNs with proximity graphs significantly outperform those with random graphs and are also far superior to Long Short-Term Memory (LSTM) model in dust source emission forecasting.

cs.LG↗

arXiv2026

Semantic Reasoning in Medicine: The Role of Knowledge Graphs Across Five Key Domains

Knowledge graphs (KGs) have emerged as a promising solution for integrating and reasoning over complex biomedical and clinical data in healthcare. By representing structured relationships among entities such as diseases, drugs, symptoms, and patient records, KGs provide a semantic backbone for decision-making, prediction, recommendation, and personalized care. Recent advances have demonstrated their utility across diverse medical applications--including clinical decision support systems, disease and treatment outcome prediction, health recommender systems, precision medicine, and medical question answering--where KGs often enhance interpretability, semantic coherence, and patient-specific reasoning. In parallel, a growing body of work focuses on medical KG generation itself, proposing frameworks that construct graphs from EHRs, clinical narratives, biomedical literature, and web resources using ontologies, semantic web technologies, deep-learning-based information extraction, and hybrid neuro-symbolic pipelines. Despite this progress, significant challenges remain, including limited and fragmented knowledge coverage, difficulties in aligning heterogeneous data sources, the fragility of current reasoning and representation-learning methods on dense multi-relational graphs, and unresolved issues related to privacy, bias, and accountability. This survey reviews and categorizes current research on KGs in medicine along both application-oriented and methodology-oriented dimensions, discusses their benefits and technical foundations, and outlines key limitations and open research directions. By analyzing trends, architectures, and evaluation practices, this work aims to guide future developments in KG-driven medical AI systems and support their safe and effective integration into healthcare environments.

cs.LG↗

arXiv2026

Where Black-box Drug-Target Interaction Prediction Models Look: Cross-Method Explainability

Drug-target interaction (DTI) and affinity (DTA) predictors increasingly achieve strong benchmark scores, yet their internal use of sequence, fingerprint, and graph features often remains opaque. We present an interpretability audit of BridgeDPI architecture on three different datasets including Gao, Human, and C.elegans. This study combines gradient-based attributions -- integrated gradients, saliency, layer-wise relevance propagation, SmoothGrad, and SmoothGrad-IG -- with feature-wise occlusion ablation and strict intersection consensus across methods to reduce single-explainer bias. We summarize sensitivity and signed effects at raw inputs, at the bridge similarity scaffold, and through the graph convolution, including edge-level sensitivities and targeted edge removals. The results show that explainability is most informative when treated as model criticism: it reveals modality dominance, padding and special-token artifacts, dataset-dependent cooperative versus suppressive effects across layers, and chemistry-consistent fragment and composition motifs where methods agree. These analyses do not substitute for structural or experimental ground truth, yet they can provide testable hypotheses for downstream validation in computational drug discovery pipelines. More broadly, applying modern XAI to contemporary DTI/DTA models is still an early pass over the rich structure implicit in trained weights and data -- yet even this first layer of scrutiny already helps researchers relate predictions to drug- and target-side representations and to prioritize external validation.

cs.LG↗

arXiv2025

Drug-Target Interaction/Affinity Prediction: Deep Learning Models and Advances Review

Drug discovery remains a slow and expensive process that involves many steps, from detecting the target structure to obtaining approval from the Food and Drug Administration (FDA), and is often riddled with safety concerns. Accurate prediction of how drugs interact with their targets and the development of new drugs by using better methods and technologies have immense potential to speed up this process, ultimately leading to faster delivery of life-saving medications. Traditional methods used for drug-target interaction prediction show limitations, particularly in capturing complex relationships between drugs and their targets. As an outcome, deep learning models have been presented to overcome the challenges of interaction prediction through their precise and efficient end results. By outlining promising research avenues and models, each with a different solution but similar to the problem, this paper aims to give researchers a better idea of methods for even more accurate and efficient prediction of drug-target interaction, ultimately accelerating the development of more effective drugs. A total of 180 prediction methods for drug-target interactions were analyzed throughout the period spanning 2016 to 2025 using different frameworks based on machine learning, mainly deep learning and graph neural networks. Additionally, this paper discusses the novelty, architecture, and input representation of these models.

q-bio.QM↗

arXiv2024

Relational Graph Convolutional Networks for Sentiment Analysis

With the growth of textual data across online platforms, sentiment analysis has become crucial for extracting insights from user-generated content. While traditional approaches and deep learning models have shown promise, they cannot often capture complex relationships between entities. In this paper, we propose leveraging Relational Graph Convolutional Networks (RGCNs) for sentiment analysis, which offer interpretability and flexibility by capturing dependencies between data points represented as nodes in a graph. We demonstrate the effectiveness of our approach by using pre-trained language models such as BERT and RoBERTa with RGCN architecture on product reviews from Amazon and Digikala datasets and evaluating the results. Our experiments highlight the effectiveness of RGCNs in capturing relational information for sentiment analysis tasks.

cs.CL↗

arXiv2023

Edge2Node: Reducing Edge Prediction to Node Classification

Despite the success of graph neural network models in node classification, edge prediction (the task of predicting missing or potential links between nodes in a graph) remains a challenging problem for these models. A common approach for edge prediction is to first obtain the embeddings of two nodes, and then a predefined scoring function is used to predict the existence of an edge between the two nodes. Here, we introduce a preliminary idea called Edge2Node which suggests to directly obtain an embedding for each edge, without the need for a scoring function. This idea wants to create a new graph H based on the graph G given for the edge prediction task, and then suggests reducing the edge prediction task on G to a node classification task on H. We anticipate that this introductory method could stimulate further investigations for edge prediction task.

cs.LG↗

arXiv2022

Emanation Graph: A Plane Geometric Spanner with Steiner Points

An emanation graph of grade $k$ on a set of points is a plane spanner made by shooting $2^{k+1}$ equally spaced rays from each point, where the shorter rays stop the longer ones upon collision. The collision points are the Steiner points of the spanner. Emanation graphs of grade one were studied by Mondal and Nachmanson in the context of network visualization. They proved that the spanning ratio of such a graph is bounded by $(2+\sqrt{2})\approx 3.414$. We improve this upper bound to $\sqrt{10} \approx 3.162$ and show this to be tight, i.e., there exist emanation graphs with spanning ratio $\sqrt{10}$. We show that for every fixed $k$, the emanation graphs of grade $k$ are constant spanners, where the constant factor depends on $k$. An emanation graph of grade two may have twice the number of edges compared to grade one graphs. Hence we introduce a heuristic method for simplifying them. In particular, we compare simplified emanation graphs against Shewchuk's constrained Delaunay triangulations on both synthetic and real-life datasets. Our experimental results reveal that the simplified emanation graphs outperform constrained Delaunay triangulations in common quality measures (e.g., edge count, angular resolution, average degree, total edge length) while maintaining a comparable spanning ratio and Steiner point count.

cs.CG↗

arXiv2021

BERT-DRE: BERT with Deep Recursive Encoder for Natural Language Sentence Matching

This paper presents a deep neural architecture, for Natural Language Sentence Matching (NLSM) by adding a deep recursive encoder to BERT so called BERT with Deep Recursive Encoder (BERT-DRE). Our analysis of model behavior shows that BERT still does not capture the full complexity of text, so a deep recursive encoder is applied on top of BERT. Three Bi-LSTM layers with residual connection are used to design a recursive encoder and an attention module is used on top of this encoder. To obtain the final vector, a pooling layer consisting of average and maximum pooling is used. We experiment our model on four benchmarks, SNLI, FarsTail, MultiNLI, SciTail, and a novel Persian religious questions dataset. This paper focuses on improving the BERT results in the NLSM task. In this regard, comparisons between BERT-DRE and BERT are conducted, and it is shown that in all cases, BERT-DRE outperforms BERT. The BERT algorithm on the religious dataset achieved an accuracy of 89.70%, and BERT-DRE architectures improved to 90.29% using the same dataset.

cs.CL↗

arXiv2020

RAC Drawings in Subcubic Area

In this paper, we study tradeoffs between curve complexity and area of Right Angle Crossing drawings (RAC drawings), which is a challenging theoretical problem in graph drawing. Given a graph with $n$ vertices and $m$ edges, we provide a RAC drawing algorithm with curve complexity $6$ and area $O(n^{2.75})$, which takes time $O(n+m)$. Our algorithm improves the previous upper bound $O(n^3)$, by Di Giacomo et al., on the area of RAC drawings.

cs.CG↗

arXiv2014

Kinetic $k$-Semi-Yao Graph and its Applications

This paper introduces a new proximity graph, called the $k$-Semi-Yao graph ($k$-SYG), on a set $P$ of points in $\mathbb{R}^d$, which is a supergraph of the $k$-nearest neighbor graph ($k$-NNG) of $P$. We provide a kinetic data structure (KDS) to maintain the $k$-SYG on moving points, where the trajectory of each point is a polynomial function whose degree is bounded by some constant. Our technique gives the first KDS for the theta graph (\ie, $1$-SYG) in $\mathbb{R}^d$. It generalizes and improves on previous work on maintaining the theta graph in $\mathbb{R}^2$. As an application, we use the kinetic $k$-SYG to provide the first KDS for maintenance of all the $k$-nearest neighbors in $\mathbb{R}^d$, for any $k\geq 1$. Previous works considered the $k=1$ case only. Our KDS for all the $1$-nearest neighbors is deterministic. The best previous KDS for all the $1$-nearest neighbors in $ \mathbb{R}^d$ is randomized. Our structure and analysis are simpler and improve on this work for the $k=1$ case. We also provide a KDS for all the $(1+ε)$-nearest neighbors, which in fact gives better performance than previous KDS's for maintenance of all the exact $1$-nearest neighbors. As another application, we present the first KDS for answering reverse $k$-nearest neighbor queries on moving points in $ \mathbb{R}^d$, for any $k\geq 1$.

cs.CG↗

arXiv2014

Kinetic Reverse $k$-Nearest Neighbor Problem

This paper provides the first solution to the kinetic reverse $k$-nearest neighbor (\rknn) problem in $\mathbb{R}^d$, which is defined as follows: Given a set $P$ of $n$ moving points in arbitrary but fixed dimension $d$, an integer $k$, and a query point $q\notin P$ at any time $t$, report all the points $p\in P$ for which $q$ is one of the $k$-nearest neighbors of $p$.

cs.CG↗

arXiv2014

A Simple, Faster Method for Kinetic Proximity Problems

For a set of $n$ points in the plane, this paper presents simple kinetic data structures (KDS's) for solutions to some fundamental proximity problems, namely, the all nearest neighbors problem, the closest pair problem, and the Euclidean minimum spanning tree (EMST) problem. Also, the paper introduces KDS's for maintenance of two well-studied sparse proximity graphs, the Yao graph and the Semi-Yao graph. We use sparse graph representations, the Pie Delaunay graph and the Equilateral Delaunay graph, to provide new solutions for the proximity problems. Then we design KDS's that efficiently maintain these sparse graphs on a set of $n$ moving points, where the trajectory of each point is assumed to be an algebraic function of constant maximum degree $s$. We use the kinetic Pie Delaunay graph and the kinetic Equilateral Delaunay graph to create KDS's for maintenance of the Yao graph, the Semi-Yao graph, all the nearest neighbors, the closest pair, and the EMST. Our KDS's use $O(n)$ space and $O(n\log n)$ preprocessing time. We provide the first KDS's for maintenance of the Semi-Yao graph and the Yao graph. Our KDS processes $O(n^2β_{2s+2}(n))$ (resp. $O(n^3β_{2s+2}^2(n)\log n)$) events to maintain the Semi-Yao graph (resp. the Yao graph); each event can be processed in time $O(\log n)$ in an amortized sense. Here, $β_s(n)$ is an extremely slow-growing function. Our KDS for maintenance of all the nearest neighbors and the closest pair processes $O(n^2β^2_{2s+2}(n)\log n)$ events. For maintenance of the EMST, our KDS processes $O(n^3β_{2s+2}^2(n)\log n)$ events. For all three of these problems, each event can be handled in time $O(\log n)$ in an amortized sense. We improve the previous randomized kinetic algorithm for maintenance of all the nearest neighbors by Agarwal, Kaplan, and Sharir, and the previous EMST KDS by Rahmati and Zarei.

cs.CG↗

arXiv2013

Kinetic Data Structures for the Semi-Yao Graph and All Nearest Neighbors in R^d

This paper presents a simple kinetic data structure for maintaining all the nearest neighbors of a set of $n$ moving points in $\mathbb{R}^d$, where the trajectory of each point is an algebraic function of at most constant degree $s$. The approach is based on maintaining the edges of the Semi-Yao graph, a sparse graph whose edge set includes the pairs of nearest neighbors as a subset. Our kinetic data structure (KDS) for maintaining all the nearest neighbors is deterministic. It processes $O(n^2β_{2s+2}^2(n)\log n)$ events with a total cost of $O(n^2β_{2s+2}(n)\log^{d+1} n)$. Here, $β_s(n)$ is an extremely slow-growing function. The best previous KDS for all the nearest neighbors in $ \mathbb{R}^d$ is by Agarwal, Kaplan, and Sharir (TALG 2008). It is a randomized result. Our structure and analysis are simpler than theirs. Also, we improve their result by a factor of $\log^d n$ in the number of events and by a $\log n$ factor in the total cost. This paper generalizes and improves the 2013 work of Rahmati, King and Whitesides (SoCG 2013) on maintaining the Semi-Yao graph in $\mathbb{R}^2$; its new technique provides the first KDS for the Semi-Yao graph in $\mathbb{R}^d$. Our KDS is local in the worst case, meaning that only a constant number of events is associated with any one point at any time. For maintaining all the nearest neighbors, neither our KDS nor the KDS by Agarwal~\etal~is local, and furthermore, each event in our KDS and in their KDS is handled in polylogarithmic time in an amortized sense. Finally, in this paper, we also give a KDS for maintenance of all the $(1+ε)$-nearest neighbors which is local and each event can be handled in a polylogarithmic worst-case time.

cs.CG↗