arXiv subjects

Sina Heydari

Publications and source records attributed to Sina Heydari.

9 recordsLinked to original sources

arXiv2026

CDAE: Enhancing Perturbation Robustness in Pretrained Language Models with Contrastive Denoising

Pre-trained language models have significantly improved sentence representation learning, yet their embedding remain sensitive to semantic preserving textual perturbations such as synonym substitution, masking and word dropout. This work proposes a lightweight Contrastive Denoising Autoencoder (CDAE) that refines pre-trained BERT embedding by jointly optimizing contrastive and reconstruction objective to learn perturbation-invariant representation. We evaluate the proposed framework using multiple perturbation strategies with varying strengths and compare it against the original BERT embeddings and SimCSE. Experimental results show that CDAE consistently preserves higher embedding similarity under perturbations, with the improvements becoming more pronounced as framework effectively enhances representation stability while preserving semantic information, highlighting perturbation-invariant learning as a promising direction for improving sentence embeddings. The source code is publicly available at: https://github.com/ComputationIASBS/CDAE

cs.AI↗

arXiv2026

HyperPersona: A Multi-Level Hypergraph Framework for Text-Based Automatic Personality Prediction

As a modern commodity, language has become a vast repository of socially and psychologically significant traits and concepts, reflecting the ways people encode pattern of thoughts, behaviors, and emotions into words. Text-based Automatic Personality Prediction (APP), seeks to infer personality from linguistic behavior, offering a scalable alternative to traditional psychometric assessments. Although text is inherently hierarchical, with the document-level capturing global features, the sentence-level encoding local semantics, and the word-level providing fine-grained lexical information, most existing approaches rely on shallow, sequential, or single-level representations that ignore the multi-level structure of written language. To address this, we propose HyperPersona, a framework that explicitly models the hierarchical organization of text (document, sentence, and word) through hypergraph structure, where a document and its sentences are represented as hyperedges, and the words are represented as nodes, enabling joint modeling of global, local, and lexical dependencies of text. Followed by a transformer-based graph encoder that learns interactions within and across these linguistic layers, yielding context-sensitive and structurally grounded feature representations for personality prediction. Experiments on the Big Five personality dimensions show that, while relying solely on text, HyperPersona effectively integrates multi-level linguistic cues, achieving superior performance compared to state-of-the-art baselines. These findings underscore the critical role of textual hierarchy in advancing human-like personality inference from natural language.

cs.AI↗

arXiv2025

Navigation of a Three-Link Microswimmer via Deep Reinforcement Learning

Motile microorganisms develop effective swimming gaits to adapt to complex biological environments. Translating this adaptability to smart microrobots presents significant challenges in motion planning and stroke design. In this work, we explore the use of reinforcement learning (RL) to develop stroke patterns for targeted navigation in a three-link swimmer model at low Reynolds numbers. Specifically, we design two RL-based strategies: one focusing on maximizing velocity (Velocity-Focused Strategy) and another balancing velocity with energy consumption (Energy-Aware Strategy). Our results demonstrate how the use of different reward functions influences the resulting stroke patterns developed via RL, which are compared with those obtained from traditional optimization methods. Furthermore, we showcase the capability of the RL-powered swimmer in adapting its stroke patterns in performing different navigation tasks, including tracing complex trajectories and pursuing moving targets. Taken together, this work highlights the potential of reinforcement learning as a versatile tool for designing efficient and adaptive microswimmers capable of sophisticated maneuvers in complex environments.

cs.RO↗

arXiv2025

RSAttAE: An Information-Aware Attention-based Autoencoder Recommender System

Recommender systems play a crucial role in modern life, including information retrieval, the pharmaceutical industry, retail, and entertainment. The entertainment sector, in particular, attracts significant attention and generates substantial profits. This work proposes a new method for predicting unknown user-movie ratings to enhance customer satisfaction. To achieve this, we utilize the MovieLens 100K dataset. Our approach introduces an attention-based autoencoder to create meaningful representations and the XGBoost method for rating predictions. The results demonstrate that our proposal outperforms most of the existing state-of-the-art methods. Availability: github.com/ComputationIASBS/RecommSys

cs.LG↗

arXiv2024

Deep Dive into Model-free Reinforcement Learning for Biological and Robotic Systems: Theory and Practice

Animals and robots exist in a physical world and must coordinate their bodies to achieve behavioral objectives. With recent developments in deep reinforcement learning, it is now possible for scientists and engineers to obtain sensorimotor strategies (policies) for specific tasks using physically simulated bodies and environments. However, the utility of these methods goes beyond the constraints of a specific task; they offer an exciting framework for understanding the organization of an animal sensorimotor system in connection to its morphology and physical interaction with the environment, as well as for deriving general design rules for sensing and actuation in robotic systems. Algorithms and code implementing both learning agents and environments are increasingly available, but the basic assumptions and choices that go into the formulation of an embodied feedback control problem using deep reinforcement learning may not be immediately apparent. Here, we present a concise exposition of the mathematical and algorithmic aspects of model-free reinforcement learning, specifically through the use of \textit{actor-critic} methods, as a tool for investigating the feedback control underlying animal and robotic behavior.

cs.RO↗

arXiv2021

Active tail flexion in concert with passive hydrodynamic forces improves swimming speed and efficiency

Fish typically swim by periodic bending of their bodies. Bending seems to follow a universal rule; it occurs at about one-third from the posterior end of the fish body with a maximum bending angle of about $30^o$. However, the hydrodynamic mechanisms that shaped this convergent design and its potential benefit to fish in terms of swimming speed and efficiency are not well understood. It is also unclear to what extent this bending is active or follows passively from the interaction of a flexible posterior with the fluid environment. Here, we use a self-propelled two-link model, with fluid-structure interactions described in the context of the vortex sheet method, to analyze the effects of both active and passive body bending on the swimming performance. We find that passive bending is more efficient but could reduce swimming speed compared to rigid flapping, but the addition of active bending could enhance both speed and efficiency. Importantly, we find that the phase difference between the posterior and anterior sections of the body is an important kinematic factor that influences performance, and that active antiphase flexion, consistent with the passive flexion phase, can simultaneously enhance speed and efficiency in a region of the design space that overlaps with biological observations. Our results are consistent with the hypothesis that fish that actively bend their bodies in a fashion that exploits passive hydrodynamics can at once improve speed and efficiency.

physics.flu-dyn↗

arXiv2021

School cohesion, speed, and efficiency are modulated by the swimmers flapping motion

Fish schools are ubiquitous in marine life. Although flow interactions are thought to be beneficial for schooling, their exact effects on the speed, energetics, and stability of the group remain elusive. Recent experiments suggest that flow interactions stabilize in-tandem formations of heaving foils. Here, we propose a minimal approach based on the vortex sheet model that captures salient features of the flow interactions among flapping swimmers, and we study the free swimming of a pair of in-line swimmers driven with identical heaving or pitching motions. We find that, independent of the flapping mode, the follower passively stabilizes at discrete locations in the wake of the leader, consistent with the heaving foil experiments, but pitching swimmers exhibit tighter and more cohesive formations. Further, in comparison to swimming alone, pitching motions increase the energetic efficiency of the group while heaving motions result in a slight increase in the swimming speed. These results recapitulate that flow interactions provide a passive mechanism that promotes school cohesion, and provide novel insight into the role of the flapping mode in controlling the emergent properties of the school.

physics.flu-dyn↗

arXiv2020

Learning to swim in potential flow

Fish swim by undulating their bodies. These propulsive motions require coordinated shape changes of a body that interacts with its fluid environment, but the specific shape coordination that leads to robust turning and swimming motions remains unclear. To address the problem of underwater motion planning, we propose a simple model of a three-link fish swimming in a potential flow environment and we use model-free reinforcement learning for shape control. We arrive at optimal shape changes for two swimming tasks: swimming in a desired direction and swimming towards a known target. This fish model belongs to a class of problems in geometric mechanics, known as driftless dynamical systems, which allow us to analyze the swimming behavior in terms of geometric phases over the shape space of the fish. These geometric methods are less intuitive in the presence of drift. Here, we use the shape space analysis as a tool for assessing, visualizing, and interpreting the control policies obtained via reinforcement learning in the absence of drift. We then examine the robustness of these policies to drift-related perturbations. Although the fish has no direct control over the drift itself, it learns to take advantage of the presence of moderate drift to reach its target.

q-bio.QM↗

arXiv2019

Sea star inspired crawling and bouncing

The oral surface of sea stars is lined with arrays of tube feet that enable them to achieve highly controlled locomotion on various terrains. The activity of the tube feet is orchestrated by a nervous system that is distributed throughout the body without a central brain. How such a distributed nervous system produces a coordinated locomotion is yet to be understood. We develop mathematical models of the biomechanics of the tube feet and the sea star body. In the model, the feet are coupled mechanically through their structural connection to a rigid body. We formulate hierarchical control laws that capture salient features of the sea star nervous system. Namely, at the tube foot level, the power and recovery strokes follow a state-dependent feedback controller. At the system level, a directionality command is communicated through the nervous system to all tube feet. We study the locomotion gaits afforded by this hierarchical control model. We find that these minimally-coupled tube feet coordinate to generate robust forward locomotion, reminiscent of the crawling motion of sea stars, on various terrains and for heterogeneous tube feet parameters and initial conditions. Our model also predicts a transition from crawling to bouncing consistent with recent experiments. We conclude by commenting on the implications of these findings for understanding the neuromechanics of sea stars and their potential application to autonomous robotic systems.

physics.bio-ph↗