Author name cluster

Minji Lee

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

10 papers

2 author rows

JBHI Journal 2026 Journal Article

EEG-Based Cross-Subject Prediction for Consciousness State Transitions Under Sedation Using a Deep Learning Framework

Ji-Hoon Jeong
Minji Lee
Chaewon Lee
Seong-Hyun Yu
Hyeong-Yeong Park
Se-Na Jang
Robert D. Sanders
Seong-Whan Lee

Intraoperative awareness due to inappropriate depth of anesthesia remains a critical concern in clinical practice. Traditional binary classifications of consciousness and unconsciousness may fail to capture the gradual and variable transitions that occur during the induction and emergence phases of anesthesia. These transitions can differ significantly across individuals and anesthetic agents. This study aims to classify three conscious states, such as consciousness, transitions, and unresponsiveness under sedation with propofol and midazolam, based on the electroencephalogram (EEG) signals. Using patient-controlled sedation paradigms, we identified transitions through behavioral responsiveness and propose a novel deep learning framework, Deep-ConTrans, incorporating common spatial pattern-based spatial filtering, multi-domain feature extraction, attention-based fusion, and domain-adversarial training for robust classification. The average classification accuracies were improved, achieving 93. 93 ( $\pm$ 3. 32)% for propofol and 97. 42 ( $\pm$ 1. 68)% for midazolam, respectively, demonstrating superior performance over conventional methods. The model also demonstrated strong cross-anesthetic generalizability, maintaining high performance when evaluated across propofol and midazolam in external validation. Specifically, transitions toward unresponsiveness were marked by increased delta power in frontal regions and increased alpha power in parietal regions. These spectral changes are consistent with cortical bistability and disruption in the posterior hot zone, both of which have been linked to alterations in consciousness. By identifying robust and drug-independent EEG signatures of transitions, this study highlights the potential for more granular and reliable intraoperative monitoring beyond binary assessment. The enhanced generalizability and sensitivity of Deep-ConTrans enable anesthesiologists to precisely identify critical transitions, thus improving anesthetic management, minimizing the risk of intraoperative awareness, and facilitating personalized sedation protocols based on real-time EEG dynamics.

Details DOI

ICRA Conference 2025 Conference Paper

Efficient Gradient-Based Inference for Manipulation Planning in Contact Factor Graphs

Jeongmin Lee
Sunkyung Park
Minji Lee
Dongjun Lee

This paper presents a framework designed to tackle a range of planning problems arise in manipulation, which typically involve complex geometric-physical reasoning related to contact and dynamic constraints. We introduce the Contact Factor Graph (CFG) to graphically model these diverse factors, enabling us to perform inference on the graphs to approximate the distribution and sample appropriate solutions. We propose a novel approach that can incorporate various phenomena of contact manipulation as differentiable factors, and develop an efficient inference algorithm for CFG that leverages this differentiability along with the conditional probabilities arising from the structured nature of contact. Our results demonstrate the capability of our framework in generating viable samples and approximating posterior distributions for various manipulation scenarios.

Details

ICML Conference 2025 Conference Paper

From Mechanistic Interpretability to Mechanistic Biology: Training, Evaluating, and Interpreting Sparse Autoencoders on Protein Language Models

Etowah Adams
Liam Bai
Minji Lee
Yiyang Yu
Mohammed AlQuraishi

Protein language models (pLMs) are powerful predictors of protein structure and function, learning through unsupervised training on millions of protein sequences. pLMs are thought to capture common motifs in protein sequences, but the specifics of pLM features are not well understood. Identifying these features would not only shed light on how pLMs work, but potentially uncover novel protein biologystudying the model to study the biology. Motivated by this, we train sparse autoencoders (SAEs) on the residual stream of a pLM, ESM-2. By characterizing SAE features, we determine that pLMs use a combination of generic features and family-specific features to represent a protein. In addition, we demonstrate how known sequence determinants of properties such as thermostability and subcellular localization can be identified by linear probing of SAE features. For predictive features without known functional associations, we hypothesize their role in unknown mechanisms and provide visualization tools to aid their interpretation. Our study gives a better understanding of the limitations of pLMs, and demonstrates how SAE features can be used to help generate hypotheses for biological mechanisms. We release our code, model weights, and feature visualizer.

Details

ICRA Conference 2025 Conference Paper

Narrow Passage Path Planning Using Collision Constraint Interpolation

Minji Lee
Jeongmin Lee
Dongjun Lee

Narrow passage path planning is a prevalent problem from industrial to household sites, often facing difficulties in finding feasible paths or requiring excessive computational resources. Given that deep penetration into the environment can cause optimization failure, we propose a framework to ensure feasibility throughout the process using a series of subproblems tailored for narrow passage problem. We begin by decomposing the environment into convex objects and initializing collision constraints with a subset of these objects. By continuously interpolating the collision constraints through the process of sequentially introducing remaining objects, our proposed framework generates subproblems that guide the optimization toward solving the narrow passage problem. Several examples are presented to demonstrate how the proposed framework addresses narrow passage path planning problems.

Details

ICML Conference 2024 Conference Paper

Robust Optimization in Protein Fitness Landscapes Using Reinforcement Learning in Latent Space

Minji Lee
Luiz Felipe Vecchietti
Hyunkyu Jung
Hyun Joo Ro
Meeyoung Cha
Ho Min Kim

Proteins are complex molecules responsible for different functions in nature. Enhancing the functionality of proteins and cellular fitness can significantly impact various industries. However, protein optimization using computational methods remains challenging, especially when starting from low-fitness sequences. We propose LatProtRL, an optimization method to efficiently traverse a latent space learned by an encoder-decoder leveraging a large protein language model. To escape local optima, our optimization is modeled as a Markov decision process using reinforcement learning acting directly in latent space. We evaluate our approach on two important fitness optimization tasks, demonstrating its ability to achieve comparable or superior fitness over baseline methods. Our findings and in vitro evaluation show that the generated sequences can reach high-fitness regions, suggesting a substantial potential of LatProtRL in lab-in-the-loop scenarios.

Details

ICRA Conference 2023 Conference Paper

Differentiable Dynamics Simulation Using Invariant Contact Mapping and Damped Contact Force

Minji Lee
Jeongmin Lee 0002
Dongjun Lee

The gradient of typical differentiable simulation is uninformative for two reasons: 1) non-smoothness in contact dynamics not considered properly, and 2) excessive local minima generated from the smoothing procedure. To tackle this issue, we first propose differentiable contact dynamics with an invariant contact set and coordinate differentiation using a signed distance function (SDF). Also, to eliminate the undesirable jittering caused by the smoothing procedure, which induces extra local minima, and to achieve a smooth and informative gradient, we further endow our framework with a novel damped contact model. Various optimization problems are implemented to demonstrate the usefulness and efficacy of our differentiable framework.

Details

ICRA Conference 2023 Conference Paper

Modular and Parallelizable Multibody Physics Simulation via Subsystem-Based ADMM

Jeongmin Lee 0002
Minji Lee
Dongjun Lee

In this paper, we present a new multibody physics simulation framework that utilizes the subsystem-based struc-ture and the Alternating Direction Method of Multiplier (ADMM). The major challenge in simulating complex high degree of freedom systems is a large number of coupled con-straints and large-sized matrices. To address this challenge, we first split the multibody into several subsystems and reformulate the dynamics equation into a subsystem perspective based on the structure of their interconnection. Then we utilize ADMM with our novel subsystem-based variable splitting scheme to solve the equation, which allows parallelizable and modular architecture. The resulting algorithm is fast, scalable, versatile, and converges well while maintaining solution consistency. Sev-eral illustrative examples are implemented with performance evaluation results showing advantages over other state-of-the-art algorithms.

Details

ICRA Conference 2021 Conference Paper

A Parallelized Iterative Algorithm for Real-Time Simulation of Long Flexible Cable Manipulation

Jeongmin Lee 0002
Minji Lee
Jaemin Yoon
Dongjun Lee

We propose a novel real-time physically-accurate simulator for long flexible cable manipulation. We first discretize the cable into multiple rigid link segments, each with complementarity-based contact model and inter-segment compliant coupling; and partition the cable into a number of subsystems, each composed with a number of consecutive links. We then formulate the inter-subsystem consistency constraint as a certain analytical condition among the inter-subsystem coupling and the contact impulses; and solve each subsystem dynamics in parallel with the contact model together with this consistency condition in an iterative manner, achieving both the speed and the accuracy of the simulation. A novel post-regulation scheme is also proposed to further speed up the simulation. Experimental validation/demonstration are also performed to show the theory.

Details

IROS Conference 2021 Conference Paper

Real-Time Physically-Accurate Simulation of Robotic Snap Connection Process

Minji Lee
Jeongmin Lee 0002
Jaemin Yoon
Dongjun Lee

We propose a novel real-time physically-accurate simulation framework for the snap connection process. For this, we first notice the peculiarities of the process, namely, small/smooth deformation, stiff connector and segmented contact. We then design our simulation to fully exploit these peculiarities by adopting the following strategies: 1) the technique of passive midpoint integration (PMI [1]), which allows for stable simulation of arbitrarily light/stiff system by enforcing discrete-time passivity; 2) linear finite element method (FEM [2]) modeling, which is adequate to deal with the small snap connector deformation while providing much faster speed as compared to nonlinear FEM; 3) segmentation of the snap connector FEM model and solving of each segment individually with their coupling analytically eliminated, thereby, further speeding up the simulation; 4) balanced model reduction (BMR [3]) to further reduce the dimension of each segment purely analytically without any prior experiment or simulation; and 5) parallelized data-driven collision detection, which turns out to further significantly speed up our simulation. Experimentally-verified simulations are also performed to show the efficacy of our proposed simulation framework.

Details

YNIMG Journal 2015 Journal Article

Which motor cortical region best predicts imagined movement?

Chang-hyun Park
Won Hyuk Chang
Minji Lee
Gyu Hyun Kwon
Laehyun Kim
Sung Tae Kim
Yun-Hee Kim

In brain-computer interfacing (BCI), motor imagery is used to provide a gateway to an effector action or behavior. However, in contrast to the main functional role of the primary motor cortex (M1) in motor execution, the M1's involvement in motor imagery has been debated, while the roles of secondary motor areas such as the premotor cortex (PMC) and supplementary motor area (SMA) in motor imagery have been proposed. We examined which motor cortical region had the greatest predictive ability for imagined movement among the primary and secondary motor areas. For two modes of motor performance, executed movement and imagined movement, in 12 healthy subjects who performed two types of motor task, hand grasping and hand rotation, we used the multivariate Bayes method to compare predictive ability between the primary and secondary motor areas (M1, PMC, and SMA) contralateral to the moved hand. With the distributed representation of activation, executed movement was best predicted from the M1 while imagined movement from the SMA, among the three motor cortical regions, in both types of motor task. In addition, the most predictive information about the distinction between executed movement and imagined movement was contained in the M1. The greater predictive ability of the SMA for imagined movement suggests its functional role that could be applied to motor imagery-based BCI.

Details DOI