Jaesik Choi

AAAI Conference 2025 Conference Paper

Diverse Rare Sample Generation with Pretrained GANs

• Subeen Lee
• Jiyeon Han
• Soyeon Kim
• Jaesik Choi

Deep generative models are proficient in generating realistic data but struggle with producing rare samples in low density regions due to their scarcity of training datasets and the mode collapse problem. While recent methods aim to improve the fidelity of generated samples, they often reduce diversity and coverage by ignoring rare and novel samples. This study proposes a novel approach for generating diverse rare samples from high-resolution image datasets with pretrained GANs. Our method employs gradient-based optimization of latent vectors within a multi-objective framework and utilizes normalizing flows for density estimation on the feature space. This enables the generation of diverse rare images, with controllable parameters for rarity, diversity, and similarity to a reference image. We demonstrate the effectiveness of our approach both qualitatively and quantitatively across various datasets and GANs without retraining or fine-tuning the pretrained GANs.

ICML Conference 2025 Conference Paper

Local Manifold Approximation and Projection for Manifold-Aware Diffusion Planning

• Kyowoon Lee
• Jaesik Choi

Recent advances in diffusion-based generative modeling have demonstrated significant promise in tackling long-horizon, sparse-reward tasks by leveraging offline datasets. While these approaches have achieved promising results, their reliability remains inconsistent due to the inherent stochastic risk of producing infeasible trajectories, limiting their applicability in safety-critical applications. We identify that the primary cause of these failures is inaccurate guidance during the sampling procedure, and demonstrate the existence of manifold deviation by deriving a lower bound on the guidance gap. To address this challenge, we propose Local Manifold Approximation and Projection (LoMAP), a training-free method that projects the guided sample onto a low-rank subspace approximated from offline datasets, preventing infeasible trajectory generation. We validate our approach on standard offline reinforcement learning benchmarks that involve challenging long-horizon planning. Furthermore, we show that, as a standalone module, LoMAP can be incorporated into the hierarchical diffusion planner, providing further performance enhancements.

ICLR Conference 2025 Conference Paper

Neural ODE Transformers: Analyzing Internal Dynamics and Adaptive Fine-tuning

• Anh Tong
• Thanh Nguyen-Tang
• Dongeun Lee 0001
• Duc Nguyen
• Toan M. Tran
• David Leo Wright Hall
• Cheongwoong Kang
• Jaesik Choi

Recent advancements in large language models (LLMs) based on transformer architectures have sparked significant interest in understanding their inner workings. In this paper, we introduce a novel approach to modeling transformer architectures using highly flexible non-autonomous neural ordinary differential equations (ODEs). Our proposed model parameterizes all weights of attention and feed-forward blocks through neural networks, expressing these weights as functions of a continuous layer index. Through spectral analysis of the model's dynamics, we uncover an increase in eigenvalue magnitude that challenges the weight-sharing assumption prevalent in existing theoretical studies. We also leverage the Lyapunov exponent to examine token-level sensitivity, enhancing model interpretability. Our neural ODE transformer demonstrates performance comparable to or better than vanilla transformers across various configurations and datasets, while offering flexible fine-tuning capabilities that can adapt to different architectural constraints.

ICLR Conference 2025 Conference Paper

Rethinking Shapley Value for Negative Interactions in Non-convex Games

• Wonjoon Chang
• Myeongjin Lee
• Jaesik Choi

We study causal interactions for payoff allocation in cooperative game theory, including quantifying feature attribution for deep learning models. Most feature attribution methods mainly stem from the criteria of the Shapley value, which assigns fair payoffs to players based on their expected contribution in a cooperative game. However, interactions between players in the game do not explicitly appear in the original formulation of the Shapley value. In this work, we reformulate the Shapley value to clarify the role of interactions and discuss implicit assumptions from a game-theoretical perspective. Our theoretical analysis demonstrates that when negative interactions exist—common in deep learning models—the efficiency axiom can lead to the undervaluation of attributions or payoffs. We suggest a new allocation rule that decomposes contributions into interactions and aggregates positive parts for non-convex games. Furthermore, we propose an approximation algorithm to reduce the cost of interaction computation which can be applied to differentiable functions such as deep learning models. Our approach mitigates counterintuitive attribution outcomes observed in existing methods, ensuring that features critical to a model’s decision receive appropriate attribution.

NeurIPS Conference 2025 Conference Paper

State-Covering Trajectory Stitching for Diffusion Planners

• Kyowoon Lee
• Jaesik Choi

Diffusion-based generative models are emerging as powerful tools for long-horizon planning in reinforcement learning (RL), particularly with offline datasets. However, their performance is fundamentally limited by the quality and diversity of training data. This often restricts their generalization to tasks outside their training distribution or longer planning horizons. To overcome this challenge, we propose State-Covering Trajectory Stitching (SCoTS), a novel reward-free trajectory augmentation method that incrementally stitches together short trajectory segments, systematically generating diverse and extended trajectories. SCoTS first learns a temporal distance-preserving latent representation that captures the underlying temporal structure of the environment, then iteratively stitches trajectory segments guided by directional exploration and novelty to effectively cover and expand this latent space. We demonstrate that SCoTS significantly improves the performance and generalization capabilities of diffusion planners on offline goal-conditioned benchmarks requiring stitching and long-horizon reasoning. Furthermore, augmented trajectories generated by SCoTS significantly improve the performance of widely used offline goal-conditioned RL algorithms across diverse environments.

AAAI Conference 2025 Conference Paper

xPatch: Dual-Stream Time Series Forecasting with Exponential Seasonal-Trend Decomposition

• Artyom Stitsyuk
• Jaesik Choi

In recent years, the application of transformer-based models in time-series forecasting has received significant attention. While often demonstrating promising results, the transformer architecture encounters challenges in fully exploiting the temporal relations within time series data due to its attention mechanism. In this work, we design eXponential Patch (xPatch for short), a novel dual-stream architecture that utilizes exponential decomposition. Inspired by the classical exponential smoothing approaches, xPatch introduces the innovative seasonal-trend exponential decomposition module. Additionally, we propose a dual-flow architecture that consists of an MLP-based linear stream and a CNN-based non-linear stream. This model investigates the benefits of employing patching and channel-independence techniques within a non-transformer model. Finally, we develop a robust arctangent loss function and a sigmoid learning rate adjustment scheme, which prevent overfitting and boost forecasting performance.

IJCAI Conference 2024 Conference Paper

Memorizing Documents with Guidance in Large Language Models

• Bumjin Park
• Jaesik Choi

Training data plays a pivotal role in AI models. Large language models (LLMs) are trained with massive amounts of documents, and their parameters hold document-related contents. Recently, several studies identified content-specific locations in LLMs by examining the parameters. Instead of the post hoc interpretation, we propose another approach. We propose document-wise memory architecture to track document memories in training. The proposed architecture maps document representations to memory entries, which softly mask memories in the forward process of LLMs. Additionally, we propose document guidance loss, which increases the likelihood of text with document memories and reduces the likelihood of the text with the memories of other documents. Experimental results on Wikitext-103-v1 with Pythia-1B show that the proposed methods provide different memory entries for documents and high recall of document-related content in generation with trained document-wise memories.

AAAI Conference 2024 Conference Paper

Towards Diverse Perspective Learning with Selection over Multiple Temporal Poolings

• Jihyeon Seong
• Jungmin Kim
• Jaesik Choi

In Time Series Classification (TSC), temporal pooling methods that consider sequential information have been proposed. However, we found that each temporal pooling has a distinct mechanism, and can perform better or worse depending on time series data. We term this fixed pooling mechanism a single perspective of temporal poolings. In this paper, we propose a novel temporal pooling method with diverse perspective learning: Selection over Multiple Temporal Poolings (SoM-TP). SoM-TP dynamically selects the optimal temporal pooling among multiple methods for each data by attention. The dynamic pooling selection is motivated by the ensemble concept of Multiple Choice Learning (MCL), which selects the best among multiple outputs. The pooling selection by SoM-TP's attention enables a non-iterative pooling ensemble within a single classifier. Additionally, we define a perspective loss and Diverse Perspective Learning Network (DPLN). The loss works as a regularizer to reflect all the pooling perspectives from DPLN. Our perspective analysis using Layer-wise Relevance Propagation (LRP) reveals the limitation of a single perspective and ultimately demonstrates diverse perspective learning of SoM-TP. We also show that SoM-TP outperforms CNN models based on other temporal poolings and state-of-the-art models in TSC with extensive UCR/UEA repositories.

IJCAI Conference 2024 Conference Paper

Towards Dynamic Trend Filtering through Trend Point Detection with Reinforcement Learning

• Jihyeon Seong
• Sekwang Oh
• Jaesik Choi

Trend filtering simplifies complex time series data by applying smoothness to filter out noise while emphasizing proximity to the original data. However, existing trend filtering methods fail to reflect abrupt changes in the trend due to `approximateness, ' resulting in constant smoothness. This approximateness uniformly filters out the tail distribution of time series data, characterized by extreme values, including both abrupt changes and noise. In this paper, we propose Trend Point Detection formulated as a Markov Decision Process (MDP), a novel approach to identifying essential points that should be reflected in the trend, departing from approximations. We term these essential points as Dynamic Trend Points (DTPs) and extract trends by interpolating them. To identify DTPs, we utilize Reinforcement Learning (RL) within a discrete action space and a forecasting sum-of-squares loss function as a reward, referred to as the Dynamic Trend Filtering network (DTF-net). DTF-net integrates flexible noise filtering, preserving critical original subsequences while removing noise as required for other subsequences. We demonstrate that DTF-net excels at capturing abrupt changes compared to other trend filtering algorithms and enhances forecasting performance, as abrupt changes are predicted rather than smoothed out.

AAAI Conference 2024 Conference Paper

Understanding Distributed Representations of Concepts in Deep Neural Networks without Supervision

• Wonjoon Chang
• Dahee Kwon
• Jaesik Choi

Understanding intermediate representations of the concepts learned by deep learning classifiers is indispensable for interpreting general model behaviors. Existing approaches to reveal learned concepts often rely on human supervision, such as pre-defined concept sets or segmentation processes. In this paper, we propose a novel unsupervised method for discovering distributed representations of concepts by selecting a principal subset of neurons. Our empirical findings demonstrate that instances with similar neuron activation states tend to share coherent concepts. Based on the observations, the proposed method selects principal neurons that construct an interpretable region, namely a Relaxed Decision Region (RDR), encompassing instances with coherent concepts in the feature space. It can be utilized to identify unlabeled subclasses within data and to detect the causes of misclassifications. Furthermore, the applicability of our method across various layers discloses distinct distributed representations over the layers, which provides deeper insights into the internal mechanisms of the deep learning model.

ICRA Conference 2023 Conference Paper

Adaptive and Explainable Deployment of Navigation Skills via Hierarchical Deep Reinforcement Learning

• Kyowoon Lee
• Seongun Kim
• Jaesik Choi

For robotic vehicles to navigate robustly and safely in unseen environments, it is crucial to decide the most suitable navigation policy. However, most existing deep reinforcement learning based navigation policies are trained with a hand-engineered curriculum and reward function which are difficult to be deployed in a wide range of real-world scenarios. In this paper, we propose a framework to learn a family of low-level navigation policies and a high-level policy for deploying them. The main idea is that, instead of learning a single navigation policy with a fixed reward function, we simultaneously learn a family of policies that exhibit different behaviors with a wide range of reward functions. We then train the high-level policy which adaptively deploys the most suitable navigation skill. We evaluate our approach in simulation and the real world and demonstrate that our method can learn diverse navigation skills and adaptively deploy them. We also illustrate that our proposed hierarchical learning framework presents explainability by providing semantics for the behavior of an autonomous agent.

ICLR Conference 2023 Conference Paper

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

• Jiyeon Han 0001
• Hwanil Choi
• Yunjey Choi
• Junho Kim
• Jung-Woo Ha 0001
• Jaesik Choi

Evaluation metrics in image synthesis play a key role to measure performances of generative models. However, most metrics mainly focus on image fidelity. Existing diversity metrics are derived by comparing distributions, and thus they cannot quantify the diversity or rarity degree of each generated image. In this work, we propose a new evaluation metric, called `rarity score', to measure both image-wise uncommonness and model-wise diversified generation performance. We first show empirical observation that typical samples are close to each other and distinctive samples are far from each other in nearest-neighbor distances on latent spaces represented by feature extractor networks such as VGG16. We then show that one can effectively filter typical or distinctive samples with the proposed metric. We also use our metric to demonstrate that the extent to which different generative models produce rare images can be effectively compared. Further, our metric can be used to compare rarities between datasets that share the same concept such as CelebA-HQ and FFHQ. Finally, we analyze the use of metrics in different designs of feature extractors to better understand the relationship between feature spaces and resulting high-rarity images. Code will be publicly available for the research community.

NeurIPS Conference 2023 Conference Paper

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

• Kyowoon Lee
• Seongun Kim
• Jaesik Choi

Diffusion-based planning has shown promising results in long-horizon, sparse-reward tasks by training trajectory diffusion models and conditioning the sampled trajectories using auxiliary guidance functions. However, due to their nature as generative models, diffusion models are not guaranteed to generate feasible plans, resulting in failed execution and precluding planners from being useful in safety-critical applications. In this work, we propose a novel approach to refine unreliable plans generated by diffusion models by providing refining guidance to error-prone plans. To this end, we suggest a new metric named restoration gap for evaluating the quality of individual plans generated by the diffusion model. A restoration gap is estimated by a gap predictor which produces restoration gap guidance to refine a diffusion planner. We additionally present an attribution map regularizer to prevent adversarial refining guidance that could be generated from the sub-optimal gap predictor, which enables further refinement of infeasible plans. We demonstrate the effectiveness of our approach on three different benchmarks in offline control settings that require long-horizon planning. 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.

ICML Conference 2023 Conference Paper

Variational Curriculum Reinforcement Learning for Unsupervised Discovery of Skills

• Seongun Kim
• Kyowoon Lee
• Jaesik Choi

Mutual information-based reinforcement learning (RL) has been proposed as a promising framework for retrieving complex skills autonomously without a task-oriented reward function through mutual information (MI) maximization or variational empowerment. However, learning complex skills is still challenging, due to the fact that the order of training skills can largely affect sample efficiency. Inspired by this, we recast variational empowerment as curriculum learning in goal-conditioned RL with an intrinsic reward function, which we name Variational Curriculum RL (VCRL). From this perspective, we propose a novel approach to unsupervised skill discovery based on information theory, called Value Uncertainty Variational Curriculum (VUVC). We prove that, under regularity conditions, VUVC accelerates the increase of entropy in the visited states compared to the uniform curriculum. We validate the effectiveness of our approach on complex navigation and robotic manipulation tasks in terms of sample efficiency and state coverage speed. We also demonstrate that the skills discovered by our method successfully complete a real-world robot navigation task in a zero-shot setup and that incorporating these skills with a global planner further increases the performance.

AAAI Conference 2022 Conference Paper

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

• 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. As widely used metrics for GANs focus more on the overall performance of the model, evaluation on the quality of individual generations or detection of defective generations is challenging. While recent studies try to detect featuremap units that cause artifacts and evaluate individual samples, these approaches require additional resources such as external networks or a number of training data to approximate the real data manifold. In this work, we propose the concept of local activation, and devise a metric on the local activation to detect artifact generations without additional supervision. We empirically verify that our approach can detect and correct artifact generations from GANs with various datasets. Finally, we discuss a geometrical analysis to partially reveal the relation between the proposed concept and low visual fidelity.

IJCAI Conference 2022 Conference Paper

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

• 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. Occasionally, they generate images that have defective or unnatural objects, which are referred to as `artifacts'. Research to investigate why these artifacts emerge and how they can be detected and removed has yet to be sufficiently carried out. To analyze this, we first hypothesize that rarely activated neurons and frequently activated neurons have different purposes and responsibilities for the progress of generating images. In this study, by analyzing the statistics and the roles for those neurons, we empirically show that rarely activated neurons are related to the failure results of making diverse objects and inducing artifacts. In addition, we suggest a correction method, called `Sequential Ablation’, to repair the defective part of the generated images without high computational cost and manual efforts.

NeurIPS Conference 2022 Conference Paper

Distilled Gradient Aggregation: Purify Features for Input Attribution in the Deep Neural Network

• Giyoung Jeon
• Haedong Jeong
• Jaesik Choi

Measuring the attribution of input features toward the model output is one of the popular post-hoc explanations on the Deep Neural Networks (DNNs). Among various approaches to compute the attribution, the gradient-based methods are widely used to generate attributions, because of its ease of implementation and the model-agnostic characteristic. However, existing gradient integration methods such as Integrated Gradients (IG) suffer from (1) the noisy attributions which cause the unreliability of the explanation, and (2) the selection for the integration path which determines the quality of explanations. FullGrad (FG) is an another approach to construct the reliable attributions by focusing the locality of piece-wise linear network with the bias gradient. Although FG has shown reasonable performance for the given input, as the shortage of the global property, FG is vulnerable to the small perturbation, while IG which includes the exploration over the input space is robust. In this work, we design a new input attribution method which adopt the strengths of both local and global attributions. In particular, we propose a novel approach to distill input features using weak and extremely positive contributor masks. We aggregate the intermediate local attributions obtained from the distillation sequence to provide reliable attribution. We perform the quantitative evaluation compared to various attribution methods and show that our method outperforms others. We also provide the qualitative result that our method obtains object-aligned and sharp attribution heatmap.

NeurIPS Conference 2022 Conference Paper

Learning Fractional White Noises in Neural Stochastic Differential Equations

• Anh Tong
• Thanh Nguyen-Tang
• Toan Tran
• Jaesik Choi

Differential equations play important roles in modeling complex physical systems. Recent advances present interesting research directions by combining differential equations with neural networks. By including noise, stochastic differential equations (SDEs) allows us to model data with uncertainty and measure imprecision. There are many variants of noises known to exist in many real-world data. For example, previously white noises are idealized and induced by Brownian motions. Nevertheless, there is a lack of machine learning models that can handle such noises. In this paper, we introduce a generalized fractional white noise to existing models and propose an efficient approximation of noise sample paths based on classical integration methods and sparse Gaussian processes. Our experimental results demonstrate that the proposed model can capture noise characteristics such as continuity from various time series data, therefore improving model fittings over existing models. We examine how we can apply our approach to score-based generative models, showing that there exists a case of our generalized noise resulting in a better image generation measure.

AAAI Conference 2021 Conference Paper

Characterizing Deep Gaussian Processes via Nonlinear Recurrence Systems

• 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). However, there exists a pathology of deep Gaussian processes that their learning capacities reduce significantly when the number of layers increases. In this paper, we present a new analysis in DGPs by studying its corresponding nonlinear dynamic systems to explain the issue. Existing work reports the pathology for the squared exponential kernel function. We extend our investigation to four types of common stationary kernel functions. The recurrence relations between layers are analytically derived, providing a tighter bound and the rate of convergence of the dynamic systems. We demonstrate our finding with a number of experimental results.

ICML Conference 2021 Conference Paper

Conditional Temporal Neural Processes with Covariance Loss

• 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. With the proposed loss, mappings from input variables to target variables are highly affected by dependencies of target variables as well as mean activation and mean dependencies of input and target variables. This nature enables the resulting neural networks to become more robust to noisy observations and recapture missing dependencies from prior information. In order to show the validity of the proposed loss, we conduct extensive sets of experiments on real-world datasets with state-of-the-art models and discuss the benefits and drawbacks of the proposed Covariance Loss.

IROS Conference 2021 Conference Paper

Explaining the Decisions of Deep Policy Networks for Robotic Manipulations

• Seongun Kim
• Jaesik Choi

Deep policy networks enable robots to learn behaviors to solve various real-world complex tasks in an end-to-end fashion. However, they lack transparency to provide the reasons of actions. Thus, such a black-box model often results in low reliability and disruptive actions during the deployment of the robot in practice. To enhance its transparency, it is important to explain robot behaviors by considering the extent to which each input feature contributes to determining a given action. 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. To this end, we present two methods for applying input attribution methods to robot policy networks: (1) we measure the importance factor of each joint torque to re ect the influence of the motor torque on the end-effector movement, and (2) we modify a relevance propagation method to handle negative inputs and outputs in deep policy networks properly. To the best of our knowledge, this is the first report to identify the dynamic changes of input attributions of multi-modal sensor inputs in deep policy networks online for robotic manipulation.

AAAI Conference 2021 Conference Paper

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

• Woo-Jeoung Nam
• Jaesik Choi
• Seong-Whan Lee

The clear transparency of Deep Neural Networks (DNNs) is hampered by complex internal structures and nonlinear transformations along deep hierarchies. In this paper, we propose a new attribution method, Relative Sectional Propagation (RSP), for fully decomposing the output predictions with the characteristics of class-discriminative attributions and clear objectness. We carefully revisit some shortcomings of backpropagation-based attribution methods, which are trade-off relations in decomposing DNNs. We define hostile factor as an element that interferes with finding the attributions of the target and propagate it in a distinguishable way to overcome the non-suppressed nature of activated neurons. 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. We also present the purging techniques to prevent the decrement of the gap between the relevance scores of the target and hostile attributions during backward propagation by eliminating the conflicting units to channel attribution map. Therefore, our method makes it possible to decompose the predictions of DNNs with clearer class-discriminativeness and detailed elucidations of activation neurons compared to the conventional attribution methods. In a verified experimental environment, we report the results of the assessments: (i) Pointing Game, (ii) mIoU, and (iii) Model Sensitivity with PAS- CAL VOC 2007, MS COCO 2014, and ImageNet datasets. The results demonstrate that our method outperforms existing backward decomposition methods, including distinctive and intuitive visualizations.

AAAI Conference 2021 Conference Paper

Learning Compositional Sparse Gaussian Processes with a Shrinkage Prior

• Anh Tong
• Toan M 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. Recently, automatic kernel composition methods provide not only accurate prediction but also attractive interpretability through search-based methods. However, existing methods suffer from slow kernel composition learning. To tackle large-scaled data, we propose a new sparse approximate posterior for GPs, MultiSVGP, constructed from groups of inducing points associated with individual additive kernels in compositional kernels. We demonstrate that this approximation provides a better fit to learn compositional kernels given empirical observations. We also provide theoretically justification on error bound when compared to the traditional sparse GP. In contrast to the search-based approach, we present a novel probabilistic algorithm to learn a kernel composition by handling the sparsity in the kernel selection with Horseshoe prior. We demonstrate that our model can capture characteristics of time series with significant reductions in computational time and have competitive regression performance on real-world data sets.

ICRA Conference 2020 Conference Paper

A Single Multi-Task Deep Neural Network with Post-Processing for Object Detection with Reasoning and Robotic Grasp Detection

• Dongwon Park
• Yonghyeok Seo
• Dongju Shin
• Jaesik Choi
• Se Young Chun

Applications of deep neural network (DNN) based object and grasp detections could be expanded significantly when the network output is processed by a high-level reasoning over relationship of objects. Recently, robotic grasp detection and object detection with reasoning have been investigated using DNNs. There have been efforts to combine these multitasks using separate networks so that robots can deal with situations of grasping specific target objects in the cluttered, stacked, complex piles of novel objects from a single RGB-D camera. We propose a single multi-task DNN that yields accurate detections of objects, grasp position and relationship reasoning among objects. Our proposed methods yield state-of-the-art performance with the accuracy of 98. 6% and 74. 2% with the computation speed of 33 and 62 frame per second on VMRD and Cornell datasets, respectively. Our methods also yielded 95. 3% grasp success rate for novel object grasping tasks with a 4-axis robot arm and 86. 7% grasp success rate in cluttered novel objects with a humanoid robot.

AAAI Conference 2020 Conference Paper

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

• Giyoung Jeon
• Haedong Jeong
• Jaesik Choi

Deep generative neural networks (DGNNs) have achieved realistic and high-quality data generation. In particular, the adversarial training scheme has been applied to many DGNNs and has exhibited powerful performance. Despite of recent advances in generative networks, identifying the image generation mechanism still remains challenging. In this paper, we present an explorative sampling algorithm to analyze generation mechanism of DGNNs. Our method efficiently obtains samples with identical attributes from a query image in a perspective of the trained model. We define generative boundaries which determine the activation of nodes in the internal layer and probe inside the model with this information. To handle a large number of boundaries, we obtain the essential set of boundaries using optimization. By gathering samples within the region surrounded by generative boundaries, we can empirically reveal the characteristics of the internal layers of DGNNs. We also demonstrate that our algorithm can find more homogeneous, the model specific samples compared to the variations of -based sampling method.

AAAI Conference 2020 Conference Paper

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

• 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. In this paper, we propose Relative Attributing Propagation (RAP), which decomposes the output predictions of DNNs with a new perspective of separating the relevant (positive) and irrelevant (negative) attributions according to the relative influence between the layers. The relevance of each neuron is identified with respect to its degree of contribution, separated into positive and negative, while preserving the conservation rule. Considering the relevance assigned to neurons in terms of relative priority, RAP allows each neuron to be assigned with a bi-polar importance score concerning the output: from highly relevant to highly irrelevant. Therefore, our method makes it possible to interpret DNNs with much clearer and attentive visualizations of the separated attributions than the conventional explaining methods. To verify that the attributions propagated by RAP correctly account for each meaning, we utilize the evaluation metrics: (i) Outside-inside relevance ratio, (ii) Segmentation mIOU and (iii) Region perturbation. In all experiments and metrics, we present a sizable gap in comparison to the existing literature.

IJCAI Conference 2019 Conference Paper

Confirmatory Bayesian Online Change Point Detection in the Covariance Structure of Gaussian Processes

• Jiyeon Han
• Kyowoon Lee
• Anh Tong
• Jaesik Choi

In the analysis of sequential data, the detection of abrupt changes is important in predicting future events. In this paper, we propose statistical hypothesis tests for detecting covariance structure changes in locally smooth time series modeled by Gaussian Processes (GPs). We provide theoretically justified thresholds for the tests, and use them to improve Bayesian Online Change Point Detection (BOCPD) by confirming statistically significant changes and non-changes. Our Confirmatory BOCPD (CBOCPD) algorithm finds multiple structural breaks in GPs even when hyperparameters are not tuned precisely. We also provide conditions under which CBOCPD provides the lower prediction error compared to BOCPD. Experimental results on synthetic and real-world datasets show that our proposed algorithm outperforms existing methods for the prediction of nonstationarity in terms of both regression error and log-likelihood.

ICML Conference 2019 Conference Paper

Discovering Latent Covariance Structures for Multiple Time Series

• Anh Tong
• Jaesik Choi

Analyzing multivariate time series data is important to predict future events and changes of complex systems in finance, manufacturing, and administrative decisions. The expressiveness power of Gaussian Process (GP) regression methods has been significantly improved by compositional covariance structures. 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. Our selective covariance structure decomposition allows exploiting shared parameters over a set of multiple, selected time series. We also investigate the well-definedness of the models when infinite latent components are introduced. We present a pragmatic search algorithm which explores a larger structure space efficiently. Experiments conducted on five real-world data sets demonstrate that our new model outperforms existing methods in term of structure discoveries and predictive performances.

ICML Conference 2018 Conference Paper

Deep Reinforcement Learning in Continuous Action Spaces: a Case Study in the Game of Simulated Curling

• Kyowoon Lee
• Sol-A. Kim
• Jaesik Choi
• Seong-Whan Lee

Many real-world applications of reinforcement learning require an agent to select optimal actions from continuous spaces. Recently, deep neural networks have successfully been applied to games with discrete actions spaces. However, deep neural networks for discrete actions are not suitable for devising strategies for games where a very small change in an action can dramatically affect the outcome. In this paper, we present a new self-play reinforcement learning framework which equips a continuous search algorithm which enables to search in continuous action spaces with a kernel regression method. Without any hand-crafted features, our network is trained by supervised learning followed by self-play reinforcement learning with a high-fidelity simulator for the Olympic sport of curling. The program trained under our framework outperforms existing programs equipped with several hand-crafted features and won an international digital curling competition.

ICML Conference 2016 Conference Paper

Automatic Construction of Nonparametric Relational Regression Models for Multiple Time Series

• Yunseong Hwang
• Anh Tong
• Jaesik Choi

Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions. The Automatic Bayesian Covariance Discovery (ABCD) system constructs natural-language description of time-series data by treating unknown time-series data nonparametrically using GP with a composite covariance kernel function. Unfortunately, learning a composite covariance kernel with a single time-series data set often results in less informative kernel that may not give qualitative, distinctive descriptions of data. We address this challenge by proposing two relational kernel learning methods which can model multiple time-series data sets by finding common, shared causes of changes. We show that the relational kernel learning methods find more accurate models for regression problems on several real-world data sets; US stock data, US house price index data and currency exchange rate data.

UAI Conference 2016 Conference Paper

Improving Imprecise Compressive Sensing Models

• Dongeun Lee 0001
• Rafael Lima
• Jaesik Choi

Random sampling in compressive sensing (CS) enables the compression of large amounts of input signals in an efficient manner, which is useful for many applications. CS reconstructs the compressed signals exactly with overwhelming probability when incoming data can be sparsely represented with a few components. However, the theory of CS framework including random sampling has been focused on exact recovery of signal; impreciseness in signal recovery has been neglected. This can be problematic when there is uncertainty in the number of sparse components such as signal sparsity in dynamic systems that can change over time. We present a new theoretical framework that handles uncertainty in signal recovery from the perspective of recovery success and quality. We show that the signal recovery success in our model is more accurate than the success probability analysis in the CS framework. Our model is then extended to the case where the success or failure of signal recovery can be relaxed. We represent the number of components included in signal recovery with a righttailed distribution and focus on recovery quality. Experimental results confirm the accuracy of our model in dynamic systems.

IJCAI Conference 2015 Conference Paper

A Deterministic Partition Function Approximation for Exponential Random Graph Models

• Wen Pu
• Jaesik Choi
• Yunseong Hwang
• Eyal Amir

Exponential Random Graphs Models (ERGM) are common, simple statistical models for social network and other network structures. Unfortunately, inference and learning with them is hard even for small networks because their partition functions are intractable for precise computation. In this paper, we introduce a new quadratic time deterministic approximation to these partition functions. Our main insight enabling this advance is that subgraph statistics is sufficient to derive a lower bound for partition functions given that the model is not dominated by a few graphs. The proposed method differs from existing methods in its ways of exploiting asymptotic properties of subgraph statistics. Compared to the current Monte Carlo simulation based methods, the new method is scalable, stable, and precise enough for inference tasks.

AAAI Conference 2015 Conference Paper

Learning Relational Kalman Filtering

• Jaesik Choi
• Eyal Amir
• Tianfang Xu
• Albert Valocchi

The Kalman Filter (KF) is pervasively used to control a vast array of consumer, health and defense products. By grouping sets of symmetric state variables, the Relational Kalman Filter (RKF) enables us to scale the exact KF for large-scale dynamic systems. In this paper, we provide a parameter learning algorithm for RKF, and a regrouping algorithm that prevents the degeneration of the relational structure for efficient filtering. The proposed algorithms significantly expand the applicability of the RKFs by solving the following questions: (1) how to learn parameters for RKF from partial observations; and (2) how to regroup the degenerated state variables by noisy real-world observations. To our knowledge, this is the first paper on learning parameters in relational continuous probabilistic models. We show that our new algorithms significantly improve the accuracy and the efficiency of filtering large-scale dynamic systems.

UAI Conference 2012 Conference Paper

Lifted Relational Variational Inference

• Jaesik Choi
• Eyal Amir

Hybrid continuous-discrete models naturally represent many real-world applications in robotics, finance, and environmental engineering. Inference with large-scale models is challenging because relational structures deteriorate rapidly during inference with observations. The main contribution of this paper is an efficient relational variational inference algorithm that factors largescale probability models into simpler variational models, composed of mixtures of iid (Bernoulli) random variables. The algorithm takes probability relational models of largescale hybrid systems and converts them to a close-to-optimal variational models. Then, it efficiently calculates marginal probabilities on the variational models by using a latent (or lifted) variable elimination or a lifted stochastic sampling. This inference is unique because it maintains the relational structure upon individual observations and during inference steps.

AAAI Conference 2011 Conference Paper

Efficient Methods for Lifted Inference with Aggregate Factors

• Jaesik Choi
• Rodrigo de Salvo Braz
• Hung Bui

Aggregate factors (that is, those based on aggregate functions such as SUM, AVERAGE, AND etc) in probabilistic relational models can compactly represent dependencies among a large number of relational random variables. However, propositional inference on a factor aggregating n k-valued random variables into an r-valued result random variable is O(rk2n ). Lifted methods can ameliorate this to O(rnk ) in general and O(rk log n) for commutative associative aggregators. In this paper, we propose (a) an exact solution constant in n when k=2 for certain aggregate operations such as AND, OR and SUM, and (b) a close approximation for inference with aggregate factors with time complexity constant in n. This approximate inference involves an analytical solution for some operations when k>2. The approximation is based on the fact that the typically used aggregate functions can be represented by linear constraints in the standard (k−1)-simplex in Rk where k is the number of possible values for random variables. This includes even aggregate functions that are commutative but not associative (e. g. , the MODE operator that chooses the most frequent value). Our algorithm takes polynomial time in k (which is only 2 for binary variables) regardless of r and n, and the error decreases as n increases. Therefore, for most applications (in which a close approximation suffices) our algorithm is a much more efficient solution than existing algorithms. We present experimental results supporting these claims. We also present a (c) third contribution which further optimizes aggregations over multiple groups of random variables with distinct distributions.

IJCAI Conference 2011 Conference Paper

Lifted Relational Kalman Filtering

• Jaesik Choi
• Abner Guzman-Rivera
• Eyal Amir

Kalman Filtering is a computational tool with widespread applications in robotics, financial and weather forecasting, environmental engineering and defense. Given observation and state transition models, the Kalman Filter (KF) recursively estimates the state variables of a dynamic system. However, the KF requires a cubic time matrix inversion operation at every timestep which prevents its application in domains with large numbers of state variables. We propose Relational Gaussian Models to represent and model dynamic systems with large numbers of variables efficiently. Furthermore, we devise an exact lifted Kalman Filtering algorithm which takes only linear time in the number of random variables at every timestep. We prove that our algorithm takes linear time in the number of state variables even when individual observations apply to each variable. To our knowledge, this is the first lifted (linear time) algorithm for filtering with continuous dynamic relational models.

UAI Conference 2010 Conference Paper

Lifted Inference for Relational Continuous Models

• Jaesik Choi
• Eyal Amir
• David J. Hill 0002

Relational Continuous Models (RCMs) represent joint probability densities over attributes of objects, when the attributes have continuous domains. With relational representations, they can model joint probability distributions over large numbers of variables compactly in a natural way. This paper presents a new exact lifted inference algorithm for RCMs, thus it scales up to large models of real world applications. The algorithm applies to Relational Pairwise Models which are (relational) products of potentials of arity 2. Our algorithm is unique in two ways. First, it substantially improves the efficiency of lifted inference with variables of continuous domains. When a relational model has Gaussian potentials, it takes only linear-time compared to cubic time of previous methods. Second, it is the first exact inference algorithm which handles RCMs in a lifted way. The algorithm is illustrated over an example from econometrics. Experimental results show that our algorithm outperforms both a groundlevel inference algorithm and an algorithm built with previously-known lifted methods.

IJCAI Conference 2009 Conference Paper

• Hannaneh Hajishirzi
• Afsaneh Shirazi
• Jaesik Choi
• Eyal Amir

In many real-world situations we are charged with detecting change as soon as possible. Important examples include detecting medical conditions, detecting security breaches, and updating caches of distributed databases. In those situations, sensing can be expensive, but it is also important to detect change in a timely manner. In this paper we present tractable greedy algorithms and prove that they solve this decision problem either optimally or approximate the optimal solution in many cases. Our problem model is a POMDP that includes a cost for sensing, a cost for delayed detection, a reward for successful detection, and no-cost partial observations. Making optimal decisions is difficult in general. We show that our tractable greedy approach finds optimal policies for sensing both a single variable and multiple correlated variables. Further, we provide approximations for the optimal solution to multiple hidden or observed variables per step. Our algorithms outperform previous algorithms in experiments over simulated data and live Wikipedia WWW pages.

ICRA Conference 2009 Conference Paper

Combining planning and motion planning

• Jaesik Choi
• Eyal Amir

Robotic manipulation is important for real, physical world applications. General Purpose manipulation with a robot (eg. delivering dishes, opening doors with a key, etc.) is demanding. It is hard because (1) objects are constrained in position and orientation, (2) many non-spatial constraints interact (or interfere) with each other, and (3) robots may have multi-degree of freedoms (DOF). In this paper we solve the problem of general purpose robotic manipulation using a novel combination of planning and motion planning. Our approach integrates motions of a robot with other (non-physical or external-to-robot) actions to achieve a goal while manipulating objects. It differs from previous, hierarchical approaches in that (a) it considers kinematic constraints in configuration space (C-space) together with constraints over object manipulations; (b) it automatically generates high-level (logical) actions from a C-space based motion planning algorithm; and (c) it decomposes a planning problem into small segments, thus reducing the complexity of planning.

IROS Conference 2007 Conference Paper

Factor-guided motion planning for a robot arm

• Jaesik Choi
• Eyal Amir

Motion planning for robotic arms is important for real, physical world applications. The planning for arms with high-degree-of-freedom (DOF) is hard because its search space is large (exponential in the number of joints), and the links may collide with static obstacles or other joints (self-collision). In this paper we present a motion planning algorithm that finds plans of motion from one arm configuration to a goal arm configuration in 2D space assuming no self-collision. Our algorithm is unique in two ways: (a) it utilizes the topology of the arm and obstacles to factor the search space and reduce the complexity of the planning problem using dynamic programming; and (b) it takes only polynomial time in the number of joints under some conditions. We provide a sufficient condition for polytime motion planning for 2D-space arms: if there is a path between two homotopic configurations, an embedded local planner finds a path within a polynomial time. The experimental results show that the proposed algorithm improves the performance of path planning for 2D arms.