论文检索

Guipeng Lan, Shuai Xiao, Jiachen Yang, Jiabao Wen

How to balance the diversity and quality of results from generative models through perception rectification poses a significant challenge. Abnormal perception in generative models is typically caused by two factors: inadequate model structure and imbalanced data distribution. In response to this issue, we propose the dynamic model perception rectification algorithm (DMPRA) for generalized generative models. The core idea is to gain a comprehensive perception of the data in the generative model by appropriately highlighting the low-density samples in the perception space, also known as the minor group samples. The entire process can be summarized as "search-evaluation-adjustment". To identify low-density regions in the data manifold within the perception space of generative models, we introduce a filtering method based on extended neighborhood sampling. Based on the informational value of samples from low-density regions, our proposed mechanism generates informative weights to assess the significance of these samples in correcting the models' perception. By using dynamic adjustment, DMPRA ensures simultaneous enhancement of diversity and quality in the presence of imbalanced data distribution. Experimental results indicate that the algorithm has effectively improved Generative Adversarial Nets (GANs), Normalizing Flows (Flows), Variational Auto-Encoders (VAEs), and Diffusion Models (Diffusion).

Ella Lan

Diabetes mellitus is a global concern, and early detection can prevent serious complications. 50% of those with diabetes live undiagnosed, disproportionately afflicting low-income groups. Non-invasive methods have emerged for timely detection; however, their limited accuracy constrains clinical usage. In this research, we present a novel Higher Dimensional Transformer (HDformer), the first Transformer-based architecture which utilizes long-range photoplethysmography (PPG) to detect diabetes. The long-range PPG maximizes signal contextual information when compared to the less-than 30 second signals commonly used in existing research. To increase the computational efficiency of HDformer’s long-range processing, a new attention module, Time Square Attention (TSA), is invented to achieve linear computational complexity with respect to the token volume while retaining the local/global dependencies. TSA converts the 1D inputs into 2D representations, grouping the adjacent points into a single 2D token. It then generates dynamic patches and feeds them into a gated mixture-of-experts (MoE) network, optimizing the learning on different attention areas. HDformer achieves state-of-the-art results (sensitivity 98.4, accuracy 97.3, specificity 92.8, AUC 0.929) on the standard MIMIC-III dataset, surpassing existing research. Furthermore, we develop an end-to-end solution where a low-cost wearable is prototyped to connect with the HDformer in the Cloud via a mobile app. This scalable, convenient, and affordable approach provides instantaneous detection and continuous monitoring for individuals. It aids doctors in easily screening for diabetes and safeguards underprivileged communities. The enhanced versatility of HDformer allows for efficient processing and learning of long-range signals in general one-dimensional time-series sequences, particularly for all biomedical waveforms.

Joonwoo Kwon, Sooyoung Kim, Yuewei Lin, Shinjae Yoo, Jiook Cha

Neural style transfer (NST) has evolved significantly in recent years. Yet, despite its rapid progress and advancement, existing NST methods either struggle to transfer aesthetic information from a style effectively or suffer from high computational costs and inefficiencies in feature disentanglement due to using pre-trained models. This work proposes a lightweight but effective model, AesFA---Aesthetic Feature-Aware NST. The primary idea is to decompose the image via its frequencies to better disentangle aesthetic styles from the reference image while training the entire model in an end-to-end manner to exclude pre-trained models at inference completely. To improve the network's ability to extract more distinct representations and further enhance the stylization quality, this work introduces a new aesthetic feature: contrastive loss. Extensive experiments and ablations show the approach not only outperforms recent NST methods in terms of stylization quality, but it also achieves faster inference. Codes are available at https://github.com/Sooyyoungg/AesFA.

Muhammad Rifki Kurniawan, Xiang Song, Zhiheng Ma, Yuhang He, Yihong Gong, Yang Qi, Xing Wei

Recent studies have demonstrated the potency of leveraging prompts in Transformers for continual learning (CL). Nevertheless, employing a discrete key-prompt bottleneck can lead to selection mismatches and inappropriate prompt associations during testing. Furthermore, this approach hinders adaptive prompting due to the lack of shareability among nearly identical instances at more granular level. To address these challenges, we introduce the Evolving Parameterized Prompt Memory (EvoPrompt), a novel method involving adaptive and continuous prompting attached to pre-trained Vision Transformer (ViT), conditioned on specific instance. We formulate a continuous prompt function as a neural bottleneck and encode the collection of prompts on network weights. We establish a paired prompt memory system consisting of a stable reference and a flexible working prompt memory. Inspired by linear mode connectivity, we progressively fuse the working prompt memory and reference prompt memory during inter-task periods, resulting in continually evolved prompt memory. This fusion involves aligning functionally equivalent prompts using optimal transport and aggregating them in parameter space with an adjustable bias based on prompt node attribution. Additionally, to enhance backward compatibility, we propose compositional classifier initialization, which leverages prior prototypes from pre-trained models to guide the initialization of new classifiers in a subspace-aware manner. Comprehensive experiments validate that our approach achieves state-of-the-art performance in both class and domain incremental learning scenarios.

Anastasiia Kurmukova, Deniz Gunduz

This paper introduces a novel approach called "friendly attack" aimed at enhancing the performance of error correction channel codes. Inspired by the concept of adversarial attacks, our method leverages the idea of introducing slight perturbations to the neural network input, resulting in a substantial impact on the network's performance. By introducing small perturbations to fixed-point modulated codewords before transmission, we effectively improve the decoder's performance without violating the input power constraint. The perturbation design is accomplished by a modified iterative fast gradient method. This study investigates various decoder architectures suitable for computing gradients to obtain the desired perturbations. Specifically, we consider belief propagation (BP) for LDPC codes; the error correcting code transformer, BP and neural BP (NBP) for polar codes, and neural BCJR for convolutional codes. We demonstrate that the proposed friendly attack method can improve the reliability across different channels, modulations, codes, and decoders. This method allows us to increase the reliability of communication with a legacy receiver by simply modifying the transmitted codeword appropriately.

Atsutoshi Kumagai, Tomoharu Iwata, Yasuhiro Fujiwara

We propose a method to learn prediction models such as classifiers for unseen target tasks where labeled and unlabeled data are absent but a few relevant input features for solving the tasks are given. Although machine learning requires data for training, data are often difficult to collect in practice. On the other hand, for many applications, a few relevant features would be more easily obtained. Although zero-shot learning or zero-shot domain adaptation use external knowledge to adapt to unseen classes or tasks without data, relevant features have not been used in existing studies. The proposed method improves the generalization performance on the target tasks, where there are no data but a few relevant features are given, by meta-learning from labeled data in related tasks. In the meta-learning phase, it is essential to simulate test phases on target tasks where prediction model learning is required without data. To this end, our neural network-based prediction model is meta-learned such that it correctly responds to perturbations of the relevant features on randomly generated synthetic data. By this modeling, the prediction model can explicitly learn the discriminability of the relevant features without real target data. When unlabeled training data are available in the target tasks, the proposed method can incorporate such data to boost the performance in a unified framework. Our experiments demonstrate that the proposed method outperforms various existing methods with four real-world datasets.

Atli Kosson, Dongyang Fan, Martin Jaggi

Batch Normalization (BN) is widely used to stabilize the optimization process and improve the test performance of deep neural networks. The regularization effect of BN depends on the batch size and explicitly using smaller batch sizes with Batch Normalization, a method known as Ghost Batch Normalization (GBN), has been found to improve generalization in many settings. We investigate the effectiveness of GBN by disentangling the induced ``Ghost Noise'' from normalization and quantitatively analyzing the distribution of noise as well as its impact on model performance. Inspired by our analysis, we propose a new regularization technique called Ghost Noise Injection (GNI) that imitates the noise in GBN without incurring the detrimental train-test discrepancy effects of small batch training. We experimentally show that GNI can provide a greater generalization benefit than GBN. Ghost Noise Injection can also be beneficial in otherwise non-noisy settings such as layer-normalized networks, providing additional evidence of the usefulness of Ghost Noise in Batch Normalization as a regularizer.

Heejo Kong, Suneung Kim, Ho-Joong Kim, Seong-Whan Lee

Recent advances in semi-supervised learning (SSL) have relied on the optimistic assumption that labeled and unlabeled data share the same class distribution. However, this assumption is often violated in real-world scenarios, where unlabeled data may contain out-of-class samples. SSL with such uncurated unlabeled data leads training models to be corrupted. In this paper, we propose a robust SSL method for learning from uncurated real-world data within the context of open-set semi-supervised learning (OSSL). Unlike previous works that rely on feature similarity distance, our method exploits uncertainty in logits. By leveraging task-dependent predictions of logits, our method is capable of robust learning even in the presence of highly correlated outliers. Our key contribution is to present an unknown-aware graph regularization (UAG), a novel technique that enhances the performance of uncertainty-based OSSL frameworks. The technique addresses not only the conflict between training objectives for inliers and outliers but also the limitation of applying the same training rule for all outlier classes, which are existed on previous uncertainty-based approaches. Extensive experiments demonstrate that UAG surpasses state-of-the-art OSSL methods by a large margin across various protocols. Codes are available at https://github.com/heejokong/UAGreg.

Fang Kong, Shuai Li

Two-sided matching markets have been widely studied in the literature due to their rich applications. Since participants are usually uncertain about their preferences, online algorithms have recently been adopted to learn them through iterative interactions. An existing work initiates the study of this problem in a many-to-one setting with responsiveness. However, their results are far from optimal and lack guarantees of incentive compatibility. We first extend an existing algorithm for the one-to-one setting to this more general setting and show it achieves a near-optimal bound for player-optimal regret. Nevertheless, due to the substantial requirement for collaboration, a single player's deviation could lead to a huge increase in its own cumulative rewards and a linear regret for others. In this paper, we aim to enhance the regret bound in many-to-one markets while ensuring incentive compatibility. We first propose the adaptively explore-then-deferred-acceptance (AETDA) algorithm for responsiveness setting and derive an upper bound for player-optimal stable regret while demonstrating its guarantee of incentive compatibility. This result is a significant improvement over existing works. And to the best of our knowledge, it constitutes the first player-optimal guarantee in matching markets that offers such robust assurances. We also consider broader substitutable preferences, one of the most general conditions to ensure the existence of a stable matching and cover responsiveness. We devise an online DA (ODA) algorithm and establish an upper bound for the player-pessimal stable regret for this setting.

Patrick Kolpaczki, Viktor Bengs, Maximilian Muschalik, Eyke Hüllermeier

The Shapley value, which is arguably the most popular approach for assigning a meaningful contribution value to players in a cooperative game, has recently been used intensively in explainable artificial intelligence. Its meaningfulness is due to axiomatic properties that only the Shapley value satisfies, which, however, comes at the expense of an exact computation growing exponentially with the number of agents. Accordingly, a number of works are devoted to the efficient approximation of the Shapley value, most of them revolve around the notion of an agent's marginal contribution. In this paper, we propose with SVARM and Stratified SVARM two parameter-free and domain-independent approximation algorithms based on a representation of the Shapley value detached from the notion of marginal contribution. We prove unmatched theoretical guarantees regarding their approximation quality and provide empirical results including synthetic games as well as common explainability use cases comparing ourselves with state-of-the-art methods.

Masahiro Kohjima

Shuffled regression is the problem of learning regression models from shuffled data that consists of a set of input features and a set of target outputs where the correspondence between the input and output is unknown. This study proposes a new deep learning method for shuffled regression called Shuffled Deep Regression (SDR). We derive the sparse and stochastic variant of the Expectation-Maximization algorithm for SDR that iteratively updates discrete latent variables and the parameters of neural networks. The effectiveness of the proposal is confirmed by benchmark data experiments.

Rune Kjærsgaard, Ahcène Boubekki, Line Clemmensen

Prototypical self-explainable classifiers have emerged to meet the growing demand for interpretable AI systems. These classifiers are designed to incorporate high transparency in their decisions by basing inference on similarity with learned prototypical objects. While these models are designed with diversity in mind, the learned prototypes often do not sufficiently represent all aspects of the input distribution, particularly those in low density regions. Such lack of sufficient data representation, known as representation bias, has been associated with various detrimental properties related to machine learning diversity and fairness. In light of this, we introduce pantypes, a new family of prototypical objects designed to capture the full diversity of the input distribution through a sparse set of objects. We show that pantypes can empower prototypical self-explainable models by occupying divergent regions of the latent space and thus fostering high diversity, interpretability and fairness.

Ben Kizaric, Daniel Pimentel-Alarcón

Low dimensional models like PCA are often used to simplify complex datasets by learning a single approximating subspace. This paradigm has expanded to union of subspaces models, like those learned by subspace clustering. In this paper, we present Principal Component Trees (PCTs), a graph structure that generalizes these ideas to identify mixtures of components that together describe the subspace structure of high-dimensional datasets. Each node in a PCT corresponds to a principal component of the data, and the edges between nodes indicate the components that must be mixed to produce a subspace that approximates a portion of the data. In order to construct PCTs, we propose two angle-distribution hypothesis tests to detect subspace clusters in the data. To analyze, compare, and select the best PCT model, we define two persistent homology measures that describe their shape. We show our construction yields two key properties of PCTs, namely ancestral orthogonality and non-decreasing singular values. Our main theoretical results show that learning PCTs reduces to PCA under multivariate normality, and that PCTs are efficient parameterizations of intersecting union of subspaces. Finally, we use PCTs to analyze neural network latent space, word embeddings, and reference image datasets.

Kathryn E. Kirchoff, Travis Maxfield, Alexander Tropsha, Shawn M. Gomez

In deep learning for drug discovery, molecular representations are often based on sequences, known as SMILES, which allow for straightforward implementation of natural language processing methodologies, one being the sequence-to-sequence autoencoder. However, we observe that training an autoencoder solely on SMILES is insufficient to learn molecular representations that are semantically meaningful, where semantics are specified by the structural (graph-to-graph) similarities between molecules. We demonstrate by example that SMILES-based autoencoders may map structurally similar molecules to distant codes, resulting in an incoherent latent space that does not necessarily respect the semantic similarities between molecules. To address this shortcoming we propose Semantically-Aware Latent Space Autoencoder (SALSA) for molecular representations: a SMILES-based transformer autoencoder modified with a contrastive task aimed at learning graph-to-graph similarities between molecules. To accomplish this, we develop a novel dataset comprised of sets of structurally similar molecules and opt for a supervised contrastive loss that is able to incorporate full sets of positive samples. We evaluate semantic awareness of SALSA representations by comparing to its ablated counterparts, and show empirically that SALSA learns representations that maintain 1) structural awareness, 2) physicochemical awareness, 3) biological awareness, and 4) semantic continuity.

Rolando Kindelan Nuñez, Mircea Petrache, Mauricio Cerda, Nancy Hitschfeld

Persistence diagrams (PD)s play a central role in topological data analysis, and are used in an ever increasing variety of applications. The comparison of PD data requires computing distances among large sets of PDs, with metrics which are accurate, theoretically sound, and fast to compute. Especially for denser multi-dimensional PDs, such comparison metrics are lacking. While on the one hand, Wasserstein-type distances have high accuracy and theoretical guarantees, they incur high computational cost. On the other hand, distances between vectorizations such as Persistence Statistics (PS)s have lower computational cost, but lack the accuracy guarantees and theoretical properties of a true distance over PD space. In this work we introduce a class of pseudodistances called Extended Topological Pseudodistances (ETD)s, which have tunable complexity, and can approximate Sliced and classical Wasserstein distances at the high-complexity extreme, while being computationally lighter and close to Persistence Statistics at the lower complexity extreme, and thus allow users to interpolate between the two metrics. We build theoretical comparisons to show how to fit our new distances at an intermediate level between persistence vectorizations and Wasserstein distances. We also experimentally verify that ETDs outperform PSs in terms of accuracy and outperform Wasserstein and Sliced Wasserstein distances in terms of computational complexity.

Young-Jin Kim, Min-Jun Kim, Kyunghwan An, Jinwoo Ahn, Jaeseok Kim, Yu-Jung Heo, Du-Seong Chang, Eun-Sol Kim

With the ability to collect vast amounts of image and natural language data from the web, there has been a remarkable advancement in Large-scale Language Models (LLMs). This progress has led to the emergence of chatbots and dialogue systems capable of fluent conversations with humans. As the variety of devices enabling interactions between humans and agents expands, and the performance of text-based dialogue systems improves, there has been recently proposed research on visual dialog. However, visual dialog requires understanding sequences of pairs consisting of images and sentences, making it challenging to gather sufficient data for training large-scale models from the web. In this paper, we propose a new multimodal learning method leveraging existing large-scale models designed for each modality, to enable model training for visual dialog with small visual dialog datasets. The key ideas of our approach are: 1) storing the history or context during the progression of visual dialog in the form of spatiotemporal graphs, and 2) introducing small modulation blocks between modality-specific models and the graphs to align the semantic spaces. For implementation, we introduce a novel structure-aware cross-attention method, which retrieves relevant image and text knowledge for utterance generation from the pretrained models. For experiments, we achieved a new state-of-the-art performance on three visual dialog datasets, including the most challenging one COMET.

Woosung Kim, Donghyeon Ki, Byung-Jun Lee

One of the major challenges of offline reinforcement learning (RL) is dealing with distribution shifts that stem from the mismatch between the trained policy and the data collection policy. Stationary distribution correction estimation algorithms (DICE) have addressed this issue by regularizing the policy optimization with f-divergence between the state-action visitation distributions of the data collection policy and the optimized policy. While such regularization naturally integrates to derive an objective to get optimal state-action visitation, such an implicit policy optimization framework has shown limited performance in practice. We observe that the reduced performance is attributed to the biased estimate and the properties of conjugate functions of f-divergence regularization. In this paper, we improve the regularized implicit policy optimization framework by relieving the bias and reshaping the conjugate function by relaxing the constraints. We show that the relaxation adjusts the degree of involvement of the sub-optimal samples in optimization, and we derive a new offline RL algorithm that benefits from the relaxed framework, improving from a previous implicit policy optimization algorithm by a large margin.

Woo Kyung Kim, Minjong Yoo, Honguk Woo

Skill-based reinforcement learning (RL) approaches have shown considerable promise, especially in solving long-horizon tasks via hierarchical structures. These skills, learned task-agnostically from offline datasets, can accelerate the policy learning process for new tasks. Yet, the application of these skills in different domains remains restricted due to their inherent dependency on the datasets, which poses a challenge when attempting to learn a skill-based policy via RL for a target domain different from the datasets' domains. In this paper, we present a novel offline skill learning framework DuSkill which employs a guided Diffusion model to generate versatile skills extended from the limited skills in datasets, thereby enhancing the robustness of policy learning for tasks in different domains. Specifically, we devise a guided diffusion-based skill decoder in conjunction with the hierarchical encoding to disentangle the skill embedding space into two distinct representations, one for encapsulating domain-invariant behaviors and the other for delineating the factors that induce domain variations in the behaviors. Our DuSkill framework enhances the diversity of skills learned offline, thus enabling to accelerate the learning procedure of high-level policies for different domains. Through experiments, we show that DuSkill outperforms other skill-based imitation learning and RL algorithms for several long-horizon tasks, demonstrating its benefits in few-shot imitation and online RL.

Taehoon Kim, Jaeyoo Park, Bohyung Han

We propose a novel class incremental learning approach, which incorporates a feature augmentation technique motivated by adversarial attacks. We employ a classifier learned in the past to complement training examples of previous tasks. The proposed approach has an unique perspective to utilize the previous knowledge in class incremental learning since it augments features of arbitrary target classes using examples in other classes via adversarial attacks on a previously learned classifier. By allowing the Cross-Class Feature Augmentations (CCFA), each class in the old tasks conveniently populates samples in the feature space, which alleviates the collapse of the decision boundaries caused by sample deficiency for the previous tasks, especially when the number of stored exemplars is small. This idea can be easily incorporated into existing class incremental learning algorithms without any architecture modification. Extensive experiments on the standard benchmarks show that our method consistently outperforms existing class incremental learning methods by significant margins in various scenarios, especially under an environment with an extremely limited memory budget.

Sungyoon Kim, Yunseon Choi, Daiki E. Matsunaga, Kee-Eung Kim

Offline Goal-Conditioned Reinforcement Learning (Offline GCRL) is an important problem in RL that focuses on acquiring diverse goal-oriented skills solely from pre-collected behavior datasets. In this setting, the reward feedback is typically absent except when the goal is achieved, which makes it difficult to learn policies especially from a finite dataset of suboptimal behaviors. In addition, realistic scenarios involve long-horizon planning, which necessitates the extraction of useful skills within sub-trajectories. Recently, the conditional diffusion model has been shown to be a promising approach to generate high-quality long-horizon plans for RL. However, their practicality for the goal-conditioned setting is still limited due to a number of technical assumptions made by the methods. In this paper, we propose SSD (Sub-trajectory Stitching with Diffusion), a model-based offline GCRL method that leverages the conditional diffusion model to address these limitations. In summary, we use the diffusion model that generates future plans conditioned on the target goal and value, with the target value estimated from the goal-relabeled offline dataset. We report state-of-the-art performance in the standard benchmark set of GCRL tasks, and demonstrate the capability to successfully stitch the segments of suboptimal trajectories in the offline data to generate high-quality plans.