Kyung-Min Kim

Gaon An, Junhyeok Lee, Xingdong Zuo et al.

Preference-based reinforcement learning (PbRL) is an approach that enables RL agents to learn from preference, which is particularly useful when formulating a reward function is challenging. Existing PbRL methods generally involve a two-step procedure: they first learn a reward model based on given preference data and then employ off-the-shelf reinforcement learning algorithms using the learned reward model. However, obtaining an accurate reward model solely from preference information, especially when the preference is from human teachers, can be difficult. Instead, we propose a PbRL algorithm that directly learns from preference without requiring any reward modeling. To achieve this, we adopt a contrastive learning framework to design a novel policy scoring metric that assigns a high score to policies that align with the given preferences. We apply our algorithm to offline RL tasks with actual human preference labels and show that our algorithm outperforms or is on par with the existing PbRL methods. Notably, on high-dimensional control tasks, our algorithm surpasses offline RL methods that learn with ground-truth reward information. Finally, we show that our algorithm can be successfully applied to fine-tune large language models.

Jinyoung Park, Seongjun Yun, Hyeonjin Park et al.

Transformer-based models have recently shown success in representation learning on graph-structured data beyond natural language processing and computer vision. However, the success is limited to small-scale graphs due to the drawbacks of full dot-product attention on graphs such as the quadratic complexity with respect to the number of nodes and message aggregation from enormous irrelevant nodes. To address these issues, we propose Deformable Graph Transformer (DGT) that performs sparse attention via dynamically sampled relevant nodes for efficiently handling large-scale graphs with a linear complexity in the number of nodes. Specifically, our framework first constructs multiple node sequences with various criteria to consider both structural and semantic proximity. Then, combining with our learnable Katz Positional Encodings, the sparse attention is applied to the node sequences for learning node representations with a significantly reduced computational cost. Extensive experiments demonstrate that our DGT achieves state-of-the-art performance on 7 graph benchmark datasets with 2.5 - 449 times less computational cost compared to transformer-based graph models with full attention.

Kyuyong Shin, Hanock Kwak, Wonjae Kim et al.

Recent studies have proposed unified user modeling frameworks that leverage user behavior data from various applications. Many of them benefit from utilizing users' behavior sequences as plain texts, representing rich information in any domain or system without losing generality. Hence, a question arises: Can language modeling for user history corpus help improve recommender systems? While its versatile usability has been widely investigated in many domains, its applications to recommender systems still remain underexplored. We show that language modeling applied directly to task-specific user histories achieves excellent results on diverse recommendation tasks. Also, leveraging additional task-agnostic user histories delivers significant performance benefits. We further demonstrate that our approach can provide promising transfer learning capabilities for a broad spectrum of real-world recommender systems, even on unseen domains and services.

Kashif Rasul, Young-Jin Park, Max Nihlén Ramström et al.

Time series models aim for accurate predictions of the future given the past, where the forecasts are used for important downstream tasks like business decision making. In practice, deep learning based time series models come in many forms, but at a high level learn some continuous representation of the past and use it to output point or probabilistic forecasts. In this paper, we introduce a novel autoregressive architecture, VQ-AR, which instead learns a \emph{discrete} set of representations that are used to predict the future. Extensive empirical comparison with other competitive deep learning models shows that surprisingly such a discrete set of representations gives state-of-the-art or equivalent results on a wide variety of time series datasets. We also highlight the shortcomings of this approach, explore its zero-shot generalization capabilities, and present an ablation study on the number of representations. The full source code of the method will be available at the time of publication with the hope that researchers can further investigate this important but overlooked inductive bias for the time series domain.

Jin-Duk Park, Kyung-Min Kim, Won-Yong Shin

Collaborative filtering (CF) remains essential in recommender systems, leveraging user--item interactions to provide personalized recommendations. Meanwhile, a number of CF techniques have evolved into sophisticated model architectures based on multi-layer perceptrons (MLPs). However, MLPs often suffer from catastrophic forgetting, and thus lose previously acquired knowledge when new information is learned, particularly in dynamic environments requiring continual learning. To tackle this problem, we propose CF-KAN, a new CF method utilizing Kolmogorov-Arnold networks (KANs). By learning nonlinear functions on the edge level, KANs are more robust to the catastrophic forgetting problem than MLPs. Built upon a KAN-based autoencoder, CF-KAN is designed in the sense of effectively capturing the intricacies of sparse user--item interactions and retaining information from previous data instances. Despite its simplicity, our extensive experiments demonstrate 1) CF-KAN's superiority over state-of-the-art methods in recommendation accuracy, 2) CF-KAN's resilience to catastrophic forgetting, underscoring its effectiveness in both static and dynamic recommendation scenarios, and 3) CF-KAN's edge-level interpretation facilitating the explainability of recommendations.

Seungjae Jung, Kyung-Min Kim

Survival analysis appears in various fields such as medicine, economics, engineering, and business. Recent studies showed that the Ordinary Differential Equation (ODE) modeling framework unifies many existing survival models while the framework is flexible and widely applicable. However, naively applying the ODE framework to survival analysis problems may model fiercely changing density function which may worsen the model's performance. Though we can apply L1 or L2 regularizers to the ODE model, their effect on the ODE modeling framework is barely known. In this paper, we propose hazard gradient penalty (HGP) to enhance the performance of a survival analysis model. Our method imposes constraints on local data points by regularizing the gradient of hazard function with respect to the data point. Our method applies to any survival analysis model including the ODE modeling framework and is easy to implement. We theoretically show that our method is related to minimizing the KL divergence between the density function at a data point and that of the neighborhood points. Experimental results on three public benchmarks show that our approach outperforms other regularization methods.

Ayush Jain, Norio Kosaka, Xinhu Li et al.

In reinforcement learning, off-policy actor-critic methods like DDPG and TD3 use deterministic policy gradients: the Q-function is learned from environment data, while the actor maximizes it via gradient ascent. We observe that in complex tasks such as dexterous manipulation and restricted locomotion with mobility constraints, the Q-function exhibits many local optima, making gradient ascent prone to getting stuck. To address this, we introduce SAVO, an actor architecture that (i) generates multiple action proposals and selects the one with the highest Q-value, and (ii) approximates the Q-function repeatedly by truncating poor local optima to guide gradient ascent more effectively. We evaluate tasks such as restricted locomotion, dexterous manipulation, and large discrete-action space recommender systems and show that our actor finds optimal actions more frequently and outperforms alternate actor architectures.

Dasol Hwang, Jinyoung Park, Sunyoung Kwon et al.

In recent years, graph neural networks (GNNs) have been widely adopted in the representation learning of graph-structured data and provided state-of-the-art performance in various applications such as link prediction, node classification, and recommendation. Motivated by recent advances of self-supervision for representation learning in natural language processing and computer vision, self-supervised learning has been recently studied to leverage unlabeled graph-structured data. However, employing self-supervision tasks as auxiliary tasks to assist a primary task has been less explored in the literature on graphs. In this paper, we propose a novel self-supervised auxiliary learning framework to effectively learn graph neural networks. Moreover, this work is the first study showing that a meta-path prediction is beneficial as a self-supervised auxiliary task for heterogeneous graphs. Our method is learning to learn a primary task with various auxiliary tasks to improve generalization performance. The proposed method identifies an effective combination of auxiliary tasks and automatically balances them to improve the primary task. Our methods can be applied to any graph neural network in a plug-in manner without manual labeling or additional data. Also, it can be extended to any other auxiliary tasks. Our experiments demonstrate that the proposed method consistently improves the performance of node classification and link prediction.

Seungjae Jung, Kyung-Min Kim, Hanock Kwak et al.

Probabilistic time-series models become popular in the forecasting field as they help to make optimal decisions under uncertainty. Despite the growing interest, a lack of thorough analysis hinders choosing what is worth applying for the desired task. In this paper, we analyze the performance of three prominent probabilistic time-series models for sales forecasting. To remove the role of random chance in architecture's performance, we make two experimental principles; 1) Large-scale dataset with various cross-validation sets. 2) A standardized training and hyperparameter selection. The experimental results show that a simple Multi-layer Perceptron and Linear Regression outperform the probabilistic models on RMSE without any feature engineering. Overall, the probabilistic models fail to achieve better performance on point estimation, such as RMSE and MAPE, than comparably simple baselines. We analyze and discuss the performances of probabilistic time-series models.

Young-Jin Park, Kyuyong Shin, Kyung-Min Kim

Graph representation learning is gaining popularity in a wide range of applications, such as social networks analysis, computational biology, and recommender systems. However, different with positive results from many academic studies, applying graph neural networks (GNNs) in a real-world application is still challenging due to non-stationary environments. The underlying distribution of streaming data changes unexpectedly, resulting in different graph structures (a.k.a., concept drift). Therefore, it is essential to devise a robust graph learning technique so that the model does not overfit to the training graphs. In this work, we present Hop Sampling, a straightforward regularization method that can effectively prevent GNNs from overfishing. The hop sampling randomly selects the number of propagation steps rather than fixing it, and by doing so, it encourages the model to learn meaningful node representation for all intermediate propagation layers and to experience a variety of plausible graphs that are not in the training set. Particularly, we describe the use case of our method in recommender systems, a representative example of the real-world non-stationary case. We evaluated hop sampling on a large-scale real-world LINE dataset and conducted an online A/B/n test in LINE Coupon recommender systems of LINE Wallet Tab. Experimental results demonstrate that the proposed scheme improves the prediction accuracy of GNNs. We observed hop sampling provides 7.97% and 16.93% improvements for NDCG and MAP compared to non-regularized GNN models in our online service. Furthermore, models using hop sampling alleviate the oversmoothing issue in GNNs enabling a deeper model as well as more diversified representation.