Offline reinforcement learning (RL) aims to learn policies from static datasets of previously collected trajectories. Existing methods for offline RL either constrain the learned policy to the support of offline data or utilize model-based virtual environments to generate simulated rollouts. However, these methods suffer from (i) poor generalization to unseen states; and (ii) trivial improvement from low-qualified rollout simulation. In this paper, we propose offline trajectory generalization through world transformers for offline reinforcement learning (OTTO). Specifically, we use casual Transformers, a.k.a. World Transformers, to predict state dynamics and the immediate reward. Then we propose four strategies to use World Transformers to generate high-rewarded trajectory simulation by perturbing the offline data. Finally, we jointly use offline data with simulated data to train an offline RL algorithm. OTTO serves as a plug-in module and can be integrated with existing offline RL methods to enhance them with better generalization capability of transformers and high-rewarded data augmentation. Conducting extensive experiments on D4RL benchmark datasets, we verify that OTTO significantly outperforms state-of-the-art offline RL methods.
Radio Frequency Fingerprint Identification (RFFI), which exploits non-ideal hardware-induced unique distortion resident in the transmit signals to identify an emitter, is emerging as a means to enhance the security of communication systems. Recently, machine learning has achieved great success in developing state-of-the-art RFFI models. However, few works consider cross-receiver RFFI problems, where the RFFI model is trained and deployed on different receivers. Due to altered receiver characteristics, direct deployment of RFFI model on a new receiver leads to significant performance degradation. To address this issue, we formulate the cross-receiver RFFI as a model adaptation problem, which adapts the trained model to unlabeled signals from a new receiver. We first develop a theoretical generalization error bound for the adaptation model. Motivated by the bound, we propose a novel method to solve the cross-receiver RFFI problem, which includes domain alignment and adaptive pseudo-labeling. The former aims at finding a feature space where both domains exhibit similar distributions, effectively reducing the domain discrepancy. Meanwhile, the latter employs a dynamic pseudo-labeling scheme to implicitly transfer the label information from the labeled receiver to the new receiver. Experimental results indicate that the proposed method can effectively mitigate the receiver impact and improve the cross-receiver RFFI performance.
In recent years, the rapid advancement of Artificial Intelligence Generated Content (AIGC) has attracted widespread attention. Among the AIGC, AI generated omnidirectional images hold significant potential for Virtual Reality (VR) and Augmented Reality (AR) applications, hence omnidirectional AIGC techniques have also been widely studied. AI-generated omnidirectional images exhibit unique distortions compared to natural omnidirectional images, however, there is no dedicated Image Quality Assessment (IQA) criteria for assessing them. This study addresses this gap by establishing a large-scale AI generated omnidirectional image IQA database named AIGCOIQA2024 and constructing a comprehensive benchmark. We first generate 300 omnidirectional images based on 5 AIGC models utilizing 25 text prompts. A subjective IQA experiment is conducted subsequently to assess human visual preferences from three perspectives including quality, comfortability, and correspondence. Finally, we conduct a benchmark experiment to evaluate the performance of state-of-the-art IQA models on our database. The database will be released to facilitate future research.
Landslides are one of the most destructive natural disasters in the world, posing a serious threat to human life and safety. The development of foundation models has provided a new research paradigm for large-scale landslide detection. The Segment Anything Model (SAM) has garnered widespread attention in the field of image segmentation. However, our experiment found that SAM performed poorly in the task of landslide segmentation. We propose TransLandSeg, which is a transfer learning approach for landslide semantic segmentation based on a vision foundation model (VFM). TransLandSeg outperforms traditional semantic segmentation models on both the Landslide4Sense dataset and the Bijie landslide dataset. Our proposed adaptive transfer learning (ATL) architecture enables the powerful segmentation capability of SAM to be transferred to landslide detection by training only 1.3% of the number of the parameters of SAM, which greatly improves the training efficiency of the model. Finally we also conducted ablation experiments on models with different ATL structures, concluded that the deployment location and residual connection of ATL play an important role in TransLandSeg accuracy improvement.
Transformer-based models have demonstrated remarkable in-context learning capabilities, prompting extensive research into its underlying mechanisms. Recent studies have suggested that Transformers can implement first-order optimization algorithms for in-context learning and even second order ones for the case of linear regression. In this work, we study whether Transformers can perform higher order optimization methods, beyond the case of linear regression. We establish that linear attention Transformers with ReLU layers can approximate second order optimization algorithms for the task of logistic regression and achieve $\epsilon$ error with only a logarithmic to the error more layers. As a by-product we demonstrate the ability of even linear attention-only Transformers in implementing a single step of Newton's iteration for matrix inversion with merely two layers. These results suggest the ability of the Transformer architecture to implement complex algorithms, beyond gradient descent.
Reinforcement learning from human feedback (RLHF) has been extensively employed to align large language models with user intent. However, proximal policy optimization (PPO) based RLHF is occasionally unstable requiring significant hyperparameter finetuning, and computationally expensive to maximize the estimated reward during alignment. Recently, direct preference optimization (DPO) is proposed to address those challenges. However, DPO relies on contrastive responses generated from human annotator and alternative LLM, instead of the policy model, limiting the effectiveness of the RLHF. In this paper, we addresses both challenges by systematically combining rejection sampling (RS) and DPO. Our proposed method, RS-DPO, initiates with the development of a supervised fine-tuned policy model (SFT). A varied set of k responses per prompt are sampled directly from the SFT model. RS-DPO identifies pairs of contrastive samples based on their reward distribution. Finally, we apply DPO with the contrastive samples to align the model to human preference. Our experiments indicate that our proposed method effectively fine-tunes LLMs with limited resource environments, leading to improved alignment with user intent. Furthermore, it outperforms existing methods, including RS, PPO, and DPO.
Learning a good history representation is one of the core challenges of reinforcement learning (RL) in partially observable environments. Recent works have shown the advantages of various auxiliary tasks for facilitating representation learning. However, the effectiveness of such auxiliary tasks has not been fully convincing, especially in partially observable environments that require long-term memorization and inference. In this empirical study, we investigate the effectiveness of future prediction for learning the representations of histories, possibly of extensive length, in partially observable environments. We first introduce an approach that decouples the task of learning history representations from policy optimization via future prediction. Then, our main contributions are two-fold: (a) we demonstrate that the performance of reinforcement learning is strongly correlated with the prediction accuracy of future observations in partially observable environments, and (b) our approach can significantly improve the overall end-to-end approach by preventing high-variance noisy signals from reinforcement learning objectives to influence the representation learning. We illustrate our claims on three types of benchmarks that necessitate the ability to process long histories for high returns.
The inference of multicellular self-assembly is the central quest of understanding morphogenesis, including embryos, organoids, tumors, and many others. However, it has been tremendously difficult to identify structural features that can indicate multicellular dynamics. Here we propose to harness the predictive power of graph-based deep neural networks (GNN) to discover important graph features that can predict dynamics. To demonstrate, we apply a physically informed GNN (piGNN) to predict the motility of multicellular collectives from a snapshot of their positions both in experiments and simulations. We demonstrate that piGNN is capable of navigating through complex graph features of multicellular living systems, which otherwise can not be achieved by classical mechanistic models. With increasing amounts of multicellular data, we propose that collaborative efforts can be made to create a multicellular data bank (MDB) from which it is possible to construct a large multicellular graph model (LMGM) for general-purposed predictions of multicellular organization.
Can we build a single large model for a wide range of PDE-related scientific learning tasks? Can this model generalize to new PDEs, even of new forms, without any fine-tuning? In-context operator learning and the corresponding model In-Context Operator Networks (ICON) represent an initial exploration of these questions. The capability of ICON regarding the first question has been demonstrated previously. In this paper, we present a detailed methodology for solving PDE problems with ICON, and show how a single ICON model can make forward and reverse predictions for different equations with different strides, provided with appropriately designed data prompts. We show the positive evidence to the second question, i.e., ICON can generalize well to some PDEs with new forms without any fine-tuning. This is exemplified through a study on 1D scalar nonlinear conservation laws, a family of PDEs with temporal evolution. We also show how to broaden the range of problems that an ICON model can address, by transforming functions and equations to ICON's capability scope. We believe that the progress in this paper is a significant step towards the goal of training a foundation model for PDE-related tasks under the in-context operator learning framework.