JCJaehoon Cha

Jaehoon Cha

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Jaehoon Cha has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Meta‐Storyboard Using Feature‐Action Patterns: Concept, Techniques, and Case Studies

    by S. Khan, S. Jones, B. Bach, J. Cha, M. Chen, J. Meikle, J. C. Roberts, J. Thiyagalingam, J. Wood, P. D. Ritsos

    This paper introduces the concepts of meta‐storyboard (MSB) and meta‐storyboard authoring for data‐driven storytelling. Traditionally, in data‐driven storytelling, a storyboard is predefined for a given data set and the information an author wants to convey. However, such a static process does not account for dynamic data and does not scale up to the need to produce visual stories for different individual viewers and contexts. In a meta‐storyboard, an author anticipates potential features that may appear in dynamically arriving or selected data and consequently defines visualization actions to convey explanations about these features. An MSB thus includes feature‐action patterns as reusable design patterns for anticipated data features, viewers' requests, and application contexts. Through two case studies, we first developed the concept of MSB as a standalone software for communicating time series data in the context of COVID‐19; then, we broadened the concept as an API to tell stories about machine learning workflows. MSB is complementary to traditional methods for authoring storytelling visualizations, while providing an efficient means to construct data‐dependent storyboards for different data streams of similar contexts.

  • Open Access

    Disentangling autoencoders and spherical harmonics for efficient shape classification in crystal growth simulations

    by Jaehoon Cha, Steven Tendyra, Alvin J. Walisinghe, Adam R. Hill, Susmita Basak, Peter R. Spackman, Michael W. Anderson, Jeyan Thiyagalingam

    Controlling crystal growth is a challenge across numerous industries, as the functional properties of crystalline materials are determined during formation and often depend on particle shape. Current approaches rely on expensive, time-consuming experimental studies complemented by exhaustive parameter space simulations, creating significant computational and analytical burdens. Despite machine learning advances in crystal growth for structure-property relationships, applications targeting morphological control remain underdeveloped. Here, we demonstrate how disentangling autoencoders combined with particle aspect ratio and spherical harmonics descriptors can enhance simulation workflows for crystal growth. This approach reveals continuous transformation pathways between different crystal morphologies whilst preserving underlying crystallographic principles. Our method significantly reduces data analytics burdens, shortens design study timelines, and deepens understanding of crystal shape control. This framework enables more efficient exploration of possible crystal morphologies, facilitating the targeted design of crystalline materials with specific functional properties.


  • Open Access

    Discovering fully semantic representations via centroid- and orientation-aware feature learning

    by Jaehoon Cha, Jinhae Park, Samuel Pinilla, Kyle L. Morris, Christopher S. Allen, Mark I. Wilkinson, Jeyan Thiyagalingam

    Learning meaningful representations of images in scientific domains that are robust to variations in centroids and orientations remains an important challenge. Here we introduce centroid- and orientation-aware disentangling autoencoder (CODAE), an encoder–decoder-based neural network that learns meaningful content of objects in a latent space. Specifically, a combination of a translation- and rotation-equivariant encoder, Euler encoding and an image moment loss enables CODAE to extract features invariant to positions and orientations of objects of interest from randomly translated and rotated images. We evaluate this approach on several publicly available scientific datasets, including protein images from life sciences, four-dimensional scanning transmission electron microscopy data from material science and galaxy images from astronomy. The evaluation shows that CODAE learns centroids, orientations and their invariant features and outputs, as well as aligned reconstructions and the exact view reconstructions of the input images with high quality.


  • Open Access

    Deep Learning Evidence for Global Optimality of Gerver’s Sofa

    by Kuangdai Leng, Jia Bi, Jaehoon Cha, Samuel Pinilla, Jeyan Thiyagalingam

    The moving sofa problem, introduced by Leo Moser in 1966, seeks to determine the maximal area of a 2D shape that can navigate an L-shaped corridor of unit width. Joseph Gerver’s 1992 solution, providing a lower bound of approximately 2.2195, is the best known, though its global optimality remains unproven. This paper leverages neural networks’ approximation power and recent advances in invexity optimization to explore global optimality. We propose two approaches supporting Gerver’s conjecture that his sofa is the unique global maximum. The first approach uses continuous function learning, discarding assumptions about the monotonicity, symmetry, and differentiability of sofa movements. The sofa area is computed as a differentiable function using our “waterfall” algorithm, with the loss function incorporating both differential terms and initial conditions based on physics-informed machine learning. Extensive training with diverse network initialization consistently converges to Gerver’s solution. The second approach applies discrete optimization to the Kallus–Romik upper bound, improving it from 2.37 to 2.3337 for five rotation angles. As the number of angles increases, our model asymptotically converges to Gerver’s area from above, indicating that no larger sofa exists.


  • Open Access

    Deep Learning Evidence for Global Optimality of Gerver's Sofa

    by Kuangdai Leng, Jia Bi, Jaehoon Cha, Samuel Pinilla, Jeyan Thiyagalingam

    The Moving Sofa Problem, formally proposed by Leo Moser in 1966, seeks to determine the largest area of a two-dimensional shape that can navigate through an $L$-shaped corridor with unit width. The current best lower bound is about 2.2195, achieved by Joseph Gerver in 1992, though its global optimality remains unproven. In this paper, we investigate this problem by leveraging the universal approximation strength and computational efficiency of neural networks. We report two approaches, both supporting Gerver's conjecture that his shape is the unique global maximum. Our first approach is continuous function learning. We drop Gerver's assumptions that i) the rotation of the corridor is monotonic and symmetric and, ii) the trajectory of its corner as a function of rotation is continuously differentiable. We parameterize rotation and trajectory by independent piecewise linear neural networks (with input being some pseudo time), allowing for rich movements such as backward rotation and pure translation. We then compute the sofa area as a differentiable function of rotation and trajectory using our "waterfall" algorithm. Our final loss function includes differential terms and initial conditions, leveraging the principles of physics-informed machine learning. Under such settings, extensive training starting from diverse function initialization and hyperparameters is conducted, unexceptionally showing rapid convergence to Gerver's solution. Our second approach is via discrete optimization of the Kallus-Romik upper bound, which converges to the maximum sofa area from above as the number of rotation angles increases. We uplift this number to 10000 to reveal its asymptotic behavior. It turns out that the upper bound yielded by our models does converge to Gerver's area (within an error of 0.01% when the number of angles reaches 2100). We also improve their five-angle upper bound from 2.37 to 2.3337.

  • Open Access

    EMinsight : a tool to capture cryoEM microscope configuration and experimental outcomes for analysis and deposition

    by Daniel Hatton, Jaehoon Cha, Stephen Riggs, Peter J. Harrison, Jeyan Thiyagalingam, Daniel K. Clare, Kyle L. Morris

    The widespread adoption of cryoEM technologies for structural biology has pushed the discipline to new frontiers. A significant worldwide effort has refined the single-particle analysis (SPA) workflow into a reasonably standardized procedure. Significant investments of development time have been made, particularly in sample preparation, microscope data-collection efficiency, pipeline analyses and data archiving. The widespread adoption of specific commercial microscopes, software for controlling them and best practices developed at facilities worldwide has also begun to establish a degree of standardization to data structures coming from the SPA workflow. There is opportunity to capitalize on this moment in the maturation of the field, to capture metadata from SPA experiments and correlate the metadata with experimental outcomes, which is presented here in a set of programs called EMinsight . This tool aims to prototype the framework and types of analyses that could lead to new insights into optimal microscope configurations as well as to define methods for metadata capture to assist with the archiving of cryoEM SPA data. It is also envisaged that this tool will be useful to microscope operators and facilities looking to rapidly generate reports on SPA data-collection and screening sessions.

  • Deep Learning-Based Rainfall Prediction Using Cloud Image Analysis

    by Jongyun Byun, Changhyun Jun, Jinwon Kim, Jaehoon Cha, Roya Narimani

    This study presents a new research direction for predicting rainfall amount using cloud image data. Herein, we employ a convolutional neural networks (CNNs) to develop an image-value model from cloud image data collected from 20 May 2020 to 24 October 2020 using IoT sensors installed at two research locations in Seoul, Republic of Korea. First, we refine the dataset using data preprocessing in three steps: 1. day/night discrimination, 2. ratio adjustment, and 3. image augmentation. Second, we construct a binary classification model using one-hot encoding for the existence of rainfall. This reduces no-rain data instances and increases model performance, thereby enabling the model to extract image features. Finally, we develop a CNN-based image-value model for rainfall prediction with a well-organized model configuration. Rainfall existence results derived from the binary classification model used for model input as preprocessed cloud image data. Proposed rainfall prediction model exhibited 85.59% accuracy on cloud images with an average mean squared error (MSE) of 3.05 for observation data under 3 mm/h. In particular, single application of the function that divides boolean input by the standard deviation of the dataset within each characteristic resulted in a 17% increase in predicted rainfall accuracy. To the best of our knowledge, this is the first study to train CNN model to predict value (rainfall) with matched image data (cloud), which could be denoted as CNN-based image-value model. Notably, the proposed model can be further extended into other image datasets, including rain streaks with various backgrounds under different climatic conditions.


  • Open Access

    Disentangling Autoencoders (DAE)

    by Jaehoon Cha, Jeyan Thiyagalingam

    Noting the importance of factorizing (or disentangling) the latent space, we propose a novel, non-probabilistic disentangling framework for autoencoders, based on the principles of symmetry transformations in group-theory. To the best of our knowledge, this is the first deterministic model that is aiming to achieve disentanglement based on autoencoders without regularizers. The proposed model is compared to seven state-of-the-art generative models based on autoencoders and evaluated based on five supervised disentanglement metrics. The experimental results show that the proposed model can have better disentanglement when variances of each features are different. We believe that this model leads to a new field for disentanglement learning based on autoencoders without regularizers.

  • Open Access

    Investigation of Applicability of Impact Factors to Estimate Solar Irradiance: Comparative Analysis Using Machine Learning Algorithms

    by Jaehoon Cha, Moon Keun Kim, Sanghyuk Lee, Kyeong Soo Kim

    This study explores investigation of applicability of impact factors to estimate solar irradiance by four machine learning algorithms using climatic elements as comparative analysis: linear regression, support vector machines (SVM), a multi-layer neural network (MLNN), and a long short-term memory (LSTM) neural network. The methods show how actual climate factors impact on solar irradiation, and the possibility of estimating one year local solar irradiance using machine learning methodologies with four different algorithms. This study conducted readily accessible local weather data including temperature, wind velocity and direction, air pressure, the amount of total cloud cover, the amount of middle and low-layer cloud cover, and humidity. The results show that the artificial neural network (ANN) models provided more close information on solar irradiance than the conventional techniques (linear regression and SVM). Between the two ANN models, the LSTM model achieved better performance, improving accuracy by 31.7% compared to the MLNN model. Impact factor analysis also revealed that temperature and the amount of total cloud cover are the dominant factors affecting solar irradiance, and the amount of middle and low-layer cloud cover is also an important factor. The results from this work demonstrate that ANN models, especially ones based on LSTM, can provide accurate information of local solar irradiance using weather data without installing and maintaining on-site solar irradiance sensors.