Search Results for author: Jaesik Choi

Found 47 papers, 13 papers with code

Capsule Neural Networks as Noise Stabilizer for Time Series Data

no code implementations20 Mar 2024 Soyeon Kim, Jihyeon Seong, Hyunkyung Han, Jaesik Choi

In this paper, we investigate the effectiveness of CapsNets in analyzing highly sensitive and noisy time series sensor data.

Adversarial Attack Time Series +1

Towards Diverse Perspective Learning with Selection over Multiple Temporal Poolings

no code implementations14 Mar 2024 Jihyeon Seong, Jungmin Kim, Jaesik Choi

In this paper, we propose a novel temporal pooling method with diverse perspective learning: Selection over Multiple Temporal Poolings (SoM-TP).

Multiple-choice Time Series +1

CardioCaps: Attention-based Capsule Network for Class-Imbalanced Echocardiogram Classification

1 code implementation14 Mar 2024 Hyunkyung Han, Jihyeon Seong, Jaesik Choi

In this paper, we explore the potential of DR-CapsNets and propose CardioCaps, a novel attention-based DR-CapsNet architecture for class-imbalanced echocardiogram classification.

Image Classification regression

Understanding Distributed Representations of Concepts in Deep Neural Networks without Supervision

2 code implementations28 Dec 2023 Wonjoon Chang, Dahee Kwon, Jaesik Choi

Understanding intermediate representations of the concepts learned by deep learning classifiers is indispensable for interpreting general model behaviors.

Refining Diffusion Planner for Reliable Behavior Synthesis by Automatic Detection of Infeasible Plans

1 code implementation NeurIPS 2023 Kyowoon Lee, Seongun Kim, Jaesik Choi

We also illustrate that our approach presents explainability by presenting the attribution maps of the gap predictor and highlighting error-prone transitions, allowing for a deeper understanding of the generated plans.

Variational Curriculum Reinforcement Learning for Unsupervised Discovery of Skills

1 code implementation30 Oct 2023 Seongun Kim, Kyowoon Lee, Jaesik Choi

We validate the effectiveness of our approach on complex navigation and robotic manipulation tasks in terms of sample efficiency and state coverage speed.

reinforcement-learning Reinforcement Learning (RL) +1

Explaining the Decisions of Deep Policy Networks for Robotic Manipulations

no code implementations30 Oct 2023 Seongun Kim, Jaesik Choi

In this paper, we present an explicit analysis of deep policy models through input attribution methods to explain how and to what extent each input feature affects the decisions of the robot policy models.

SigFormer: Signature Transformers for Deep Hedging

1 code implementation20 Oct 2023 Anh Tong, Thanh Nguyen-Tang, Dongeun Lee, Toan Tran, Jaesik Choi

To mitigate such difficulties, we introduce SigFormer, a novel deep learning model that combines the power of path signatures and transformers to handle sequential data, particularly in cases with irregularities.

Impact of Co-occurrence on Factual Knowledge of Large Language Models

1 code implementation12 Oct 2023 Cheongwoong Kang, Jaesik Choi

Consequently, LLMs struggle to recall facts whose subject and object rarely co-occur in the pre-training dataset although they are seen during finetuning.

The Disharmony between BN and ReLU Causes Gradient Explosion, but is Offset by the Correlation between Activations

no code implementations23 Apr 2023 Inyoung Paik, Jaesik Choi

In this study, we analyze the occurrence and mitigation of gradient explosion both theoretically and empirically, and discover that the correlation between activations plays a key role in preventing the gradient explosion from persisting throughout the training.

Stress and Adaptation: Applying Anna Karenina Principle in Deep Learning for Image Classification

no code implementations22 Feb 2023 Nesma Mahmoud, Hanna Antson, Jaesik Choi, Osamu Shimmi, Kallol Roy

In our paper, we have generated artificial perturbations to our model by hot-swapping the activation and loss functions during the training.

Image Classification

Beyond Single Path Integrated Gradients for Reliable Input Attribution via Randomized Path Sampling

no code implementations ICCV 2023 Giyoung Jeon, Haedong Jeong, Jaesik Choi

We show that such noisy attribution can be reduced by aggregating attributions from the multiple paths instead of using a single path.

Explanation on Pretraining Bias of Finetuned Vision Transformer

no code implementations18 Nov 2022 Bumjin Park, Jaesik Choi

As the number of fine tuning of pretrained models increased, understanding the bias of pretrained model is essential.

Why Do Neural Language Models Still Need Commonsense Knowledge to Handle Semantic Variations in Question Answering?

1 code implementation1 Sep 2022 Sunjae Kwon, Cheongwoong Kang, Jiyeon Han, Jaesik Choi

We exemplify the possibility to overcome the limitations of the MNLM-based RC models by enriching text with the required knowledge from an external commonsense knowledge repository in controlled experiments.

Question Answering Reading Comprehension

On the Relationship Between Adversarial Robustness and Decision Region in Deep Neural Network

no code implementations7 Jul 2022 SeongJin Park, Haedong Jeong, Giyoung Jeon, Jaesik Choi

In general, Deep Neural Networks (DNNs) are evaluated by the generalization performance measured on unseen data excluded from the training phase.

Adversarial Attack Adversarial Robustness

Rarity Score : A New Metric to Evaluate the Uncommonness of Synthesized Images

1 code implementation17 Jun 2022 Jiyeon Han, Hwanil Choi, Yunjey Choi, Junho Kim, Jung-Woo Ha, Jaesik Choi

In this work, we propose a new evaluation metric, called `rarity score', to measure the individual rarity of each image synthesized by generative models.

Image Generation

Variational Neural Temporal Point Process

no code implementations17 Feb 2022 Deokjun Eom, Sehyun Lee, Jaesik Choi

The intensity functions are computed using the distribution of latent variable so that we can predict event types and the arrival times of the events more accurately.

Point Processes

Can We Find Neurons that Cause Unrealistic Images in Deep Generative Networks?

1 code implementation17 Jan 2022 Hwanil Choi, Wonjoon Chang, Jaesik Choi

Even though Generative Adversarial Networks (GANs) have shown a remarkable ability to generate high-quality images, GANs do not always guarantee the generation of photorealistic images.

GAN image forensics Image Generation +2

An Unsupervised Way to Understand Artifact Generating Internal Units in Generative Neural Networks

no code implementations16 Dec 2021 Haedong Jeong, Jiyeon Han, Jaesik Choi

Despite significant improvements on the image generation performance of Generative Adversarial Networks (GANs), generations with low visual fidelity still have been observed.

Image Generation

Decoupled Kernel Neural Processes: Neural Network-Parameterized Stochastic Processes using Explicit Data-driven Kernel

no code implementations29 Sep 2021 Daehoon Gwak, Gyubok Lee, Jaehoon Lee, Jaesik Choi, Jaegul Choo, Edward Choi

To address this, we introduce a new neural stochastic processes, Decoupled Kernel Neural Processes (DKNPs), which explicitly learn a separate mean and kernel function to directly model the covariance between output variables in a data-driven manner.

Gaussian Processes

Empirical Study of the Decision Region and Robustness in Deep Neural Networks

no code implementations29 Sep 2021 SeongJin Park, Haedong Jeong, Giyoung Jeon, Jaesik Choi

In general, the Deep Neural Networks (DNNs) is evaluated by the generalization performance measured on the unseen data excluded from the training phase.

Adversarial Attack Adversarial Robustness

Conditional Temporal Neural Processes with Covariance Loss

1 code implementation ICML 2021 Boseon Yoo, Jiwoo Lee, Janghoon Ju, Seijun Chung, Soyeon Kim, Jaesik Choi

We introduce a novel loss function, Covariance Loss, which is conceptually equivalent to conditional neural processes and has a form of regularization so that is applicable to many kinds of neural networks.

Time Series Forecasting Traffic Prediction

Automatic Correction of Internal Units in Generative Neural Networks

no code implementations CVPR 2021 Ali Tousi, Haedong Jeong, Jiyeon Han, Hwanil Choi, Jaesik Choi

Generative Adversarial Networks (GANs) have shown satisfactory performance in synthetic image generation by devising complex network structure and adversarial training scheme.

Image Generation

Learning Compositional Sparse Gaussian Processes with a Shrinkage Prior

no code implementations21 Dec 2020 Anh Tong, Toan Tran, Hung Bui, Jaesik Choi

Choosing a proper set of kernel functions is an important problem in learning Gaussian Process (GP) models since each kernel structure has different model complexity and data fitness.

Gaussian Processes Time Series +1

Interpreting Deep Neural Networks with Relative Sectional Propagation by Analyzing Comparative Gradients and Hostile Activations

no code implementations7 Dec 2020 Woo-Jeoung Nam, Jaesik Choi, Seong-Whan Lee

As a result, it is possible to assign the bi-polar relevance scores of the target (positive) and hostile (negative) attributions while maintaining each attribution aligned with the importance.

Characterizing Deep Gaussian Processes via Nonlinear Recurrence Systems

no code implementations19 Oct 2020 Anh Tong, Jaesik Choi

Recent advances in Deep Gaussian Processes (DGPs) show the potential to have more expressive representation than that of traditional Gaussian Processes (GPs).

Gaussian Processes

Improved Predictive Deep Temporal Neural Networks with Trend Filtering

no code implementations16 Oct 2020 YoungJin Park, Deokjun Eom, Byoungki Seo, Jaesik Choi

We reveal that the predictive performance of deep temporal neural networks improves when the training data is temporally processed by a trend filtering.

Time Series Time Series Analysis

Interpretation of Deep Temporal Representations by Selective Visualization of Internally Activated Nodes

no code implementations27 Apr 2020 Sohee Cho, Ginkyeng Lee, Wonjoon Chang, Jaesik Choi

Recently deep neural networks demonstrate competitive performances in classification and regression tasks for many temporal or sequential data.

Clustering General Classification

An Efficient Explorative Sampling Considering the Generative Boundaries of Deep Generative Neural Networks

no code implementations12 Dec 2019 Giyoung Jeon, Haedong Jeong, Jaesik Choi

Despite of recent advances in generative networks, identifying the image generation mechanism still remains challenging.

Image Generation

Why Do Masked Neural Language Models Still Need Common Sense Knowledge?

no code implementations8 Nov 2019 Sunjae Kwon, Cheongwoong Kang, Jiyeon Han, Jaesik Choi

From the test, we observed that MNLMs partially understand various types of common sense knowledge but do not accurately understand the semantic meaning of relations.

Common Sense Reasoning Question Answering

Relative Attributing Propagation: Interpreting the Comparative Contributions of Individual Units in Deep Neural Networks

1 code implementation1 Apr 2019 Woo-Jeoung Nam, Shir Gur, Jaesik Choi, Lior Wolf, Seong-Whan Lee

As Deep Neural Networks (DNNs) have demonstrated superhuman performance in a variety of fields, there is an increasing interest in understanding the complex internal mechanisms of DNNs.

Parametric Information Bottleneck to Optimize Stochastic Neural Networks

no code implementations ICLR 2018 Thanh T. Nguyen, Jaesik Choi

Here, we propose Parametric Information Bottleneck (PIB) for a neural network by utilizing (only) its model parameters explicitly to approximate the compression and the relevance.

Layer-wise Learning of Stochastic Neural Networks with Information Bottleneck

no code implementations4 Dec 2017 Thanh T. Nguyen, Jaesik Choi

Information Bottleneck (IB) is a generalization of rate-distortion theory that naturally incorporates compression and relevance trade-offs for learning.

Adversarial Robustness

Grouped Convolutional Neural Networks for Multivariate Time Series

no code implementations29 Mar 2017 Subin Yi, Janghoon Ju, Man-Ki Yoon, Jaesik Choi

In experiments with two real-world datasets, we demonstrate that our group CNNs outperform existing CNN based regression methods.

Anomaly Detection Clustering +2

Discovering Latent Covariance Structures for Multiple Time Series

no code implementations28 Mar 2017 Anh Tong, Jaesik Choi

In this paper, we present a new GP model which naturally handles multiple time series by placing an Indian Buffet Process (IBP) prior on the presence of shared kernels.

Time Series Time Series Analysis

Make Hawkes Processes Explainable by Decomposing Self-Triggering Kernels

no code implementations27 Mar 2017 Rafael Lima, Jaesik Choi

We demonstrate that the new automatic kernel decomposition procedure outperforms the existing methods on the prediction of discrete events in real-world data.

Automatic Generation of Probabilistic Programming from Time Series Data

no code implementations4 Jul 2016 Anh Tong, Jaesik Choi

In this paper, we provide a new perspective to build expressive probabilistic program from continue time series data when the structure of model is not given.

Descriptive Probabilistic Programming +3

Searching for Topological Symmetry in Data Haystack

no code implementations11 Mar 2016 Kallol Roy, Anh Tong, Jaesik Choi

To compute the symmetry in a grid structure, we introduce three legal grid moves (i) Commutation (ii) Cyclic Permutation (iii) Stabilization on sets of local grid squares, grid blocks.

Global Deconvolutional Networks for Semantic Segmentation

no code implementations12 Feb 2016 Vladimir Nekrasov, Janghoon Ju, Jaesik Choi

Semantic image segmentation is a principal problem in computer vision, where the aim is to correctly classify each individual pixel of an image into a semantic label.

Autonomous Driving Image Classification +4

The Automatic Statistician: A Relational Perspective

no code implementations26 Nov 2015 Yunseong Hwang, Anh Tong, Jaesik Choi

Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions.

Gaussian Processes Time Series +1

Cannot find the paper you are looking for? You can Submit a new open access paper.