arXiv subjects

Yang Yu

Publications and source records attributed to Yang Yu.

At least 217 records · Page 12Linked to original sources

arXiv2024

Unambiguous fluctuation decomposition of the self-energy: pseudogap physics beyond spin fluctuations

Correlated electron systems may give rise to multiple effective interactions whose combined impact on quasiparticle properties can be difficult to disentangle. We introduce an unambiguous decomposition of the electronic self-energy which allows us to quantify the contributions of various effective interactions simultaneously. We use this tool to revisit the hole-doped Hubbard model within the dynamical cluster approximation, where commonly spin fluctuations are considered to be the origin of the pseudogap. While our fluctuation decomposition confirms that spin fluctuations indeed suppress antinodal electronic spectral weight, we show that they alone can not capture the pseudogap self-energy quantitatively. Nonlocal multi-boson Feynman diagrams yield substantial contributions and are needed for a quantitative description of the pseudogap.

cond-mat.str-el↗

arXiv2024

Efficient Recurrent Off-Policy RL Requires a Context-Encoder-Specific Learning Rate

Real-world decision-making tasks are usually partially observable Markov decision processes (POMDPs), where the state is not fully observable. Recent progress has demonstrated that recurrent reinforcement learning (RL), which consists of a context encoder based on recurrent neural networks (RNNs) for unobservable state prediction and a multilayer perceptron (MLP) policy for decision making, can mitigate partial observability and serve as a robust baseline for POMDP tasks. However, previous recurrent RL methods face training stability issues due to the gradient instability of RNNs. In this paper, we propose Recurrent Off-policy RL with Context-Encoder-Specific Learning Rate (RESeL) to tackle this issue. Specifically, RESeL uses a lower learning rate for context encoder than other MLP layers to ensure the stability of the former while maintaining the training efficiency of the latter. We integrate this technique into existing off-policy RL methods, resulting in the RESeL algorithm. We evaluated RESeL in 18 POMDP tasks, including classic, meta-RL, and credit assignment scenarios, as well as five MDP locomotion tasks. The experiments demonstrate significant improvements in training stability with RESeL. Comparative results show that RESeL achieves notable performance improvements over previous recurrent RL baselines in POMDP tasks, and is competitive with or even surpasses state-of-the-art methods in MDP tasks. Further ablation studies highlight the necessity of applying a distinct learning rate for the context encoder.

cs.LG↗

arXiv2024

ReMax: A Simple, Effective, and Efficient Reinforcement Learning Method for Aligning Large Language Models

Reinforcement Learning from Human Feedback (RLHF) is key to aligning Large Language Models (LLMs), typically paired with the Proximal Policy Optimization (PPO) algorithm. While PPO is a powerful method designed for general reinforcement learning tasks, it is overly sophisticated for LLMs, leading to laborious hyper-parameter tuning and significant computation burdens. To make RLHF efficient, we present ReMax, which leverages 3 properties of RLHF: fast simulation, deterministic transitions, and trajectory-level rewards. These properties are not exploited in PPO, making it less suitable for RLHF. Building on the renowned REINFORCE algorithm, ReMax does not require training an additional value model as in PPO and is further enhanced with a new variance reduction technique. ReMax offers several benefits over PPO: it is simpler to implement, eliminates more than 4 hyper-parameters in PPO, reduces GPU memory usage, and shortens training time. ReMax can save about 46% GPU memory than PPO when training a 7B model and enables training on A800-80GB GPUs without the memory-saving offloading technique needed by PPO. Applying ReMax to a Mistral-7B model resulted in a 94.78% win rate on the AlpacaEval leaderboard and a 7.739 score on MT-bench, setting a new SOTA for open-source 7B models. These results show the effectiveness of ReMax while addressing the limitations of PPO in LLMs.

cs.LG↗

arXiv2024

Preventive Audits for Data Applications Before Data Sharing in the Power IoT

With the increase in data volume, more types of data are being used and shared, especially in the power Internet of Things (IoT). However, the processes of data sharing may lead to unexpected information leakage because of the ubiquitous relevance among the different data, thus it is necessary for data owners to conduct preventive audits for data applications before data sharing to avoid the risk of key information leakage. Considering that the same data may play completely different roles in different application scenarios, data owners should know the expected data applications of the data buyers in advance and provide modified data that are less relevant to the private information of the data owners and more relevant to the nonprivate information that the data buyers need. In this paper, data sharing in the power IoT is regarded as the background, and the mutual information of the data and their implicit information is selected as the data feature parameter to indicate the relevance between the data and their implicit information or the ability to infer the implicit information from the data. Therefore, preventive audits should be conducted based on changes in the data feature parameters before and after data sharing. The probability exchange adjustment method is proposed as the theoretical basis of preventive audits under simplified consumption, and the corresponding optimization models are constructed and extended to more practical scenarios with multivariate characteristics. Finally, case studies are used to validate the effectiveness of the proposed preventive audits.

cs.CR↗

arXiv2024

Adaptive Negative Evidential Deep Learning for Open-set Semi-supervised Learning

Semi-supervised learning (SSL) methods assume that labeled data, unlabeled data and test data are from the same distribution. Open-set semi-supervised learning (Open-set SSL) considers a more practical scenario, where unlabeled data and test data contain new categories (outliers) not observed in labeled data (inliers). Most previous works focused on outlier detection via binary classifiers, which suffer from insufficient scalability and inability to distinguish different types of uncertainty. In this paper, we propose a novel framework, Adaptive Negative Evidential Deep Learning (ANEDL) to tackle these limitations. Concretely, we first introduce evidential deep learning (EDL) as an outlier detector to quantify different types of uncertainty, and design different uncertainty metrics for self-training and inference. Furthermore, we propose a novel adaptive negative optimization strategy, making EDL more tailored to the unlabeled dataset containing both inliers and outliers. As demonstrated empirically, our proposed method outperforms existing state-of-the-art methods across four datasets.

cs.LG↗

arXiv2024

Knowledgeable Agents by Offline Reinforcement Learning from Large Language Model Rollouts

Reinforcement learning (RL) trains agents to accomplish complex tasks through environmental interaction data, but its capacity is also limited by the scope of the available data. To obtain a knowledgeable agent, a promising approach is to leverage the knowledge from large language models (LLMs). Despite previous studies combining LLMs with RL, seamless integration of the two components remains challenging due to their semantic gap. This paper introduces a novel method, Knowledgeable Agents from Language Model Rollouts (KALM), which extracts knowledge from LLMs in the form of imaginary rollouts that can be easily learned by the agent through offline reinforcement learning methods. The primary challenge of KALM lies in LLM grounding, as LLMs are inherently limited to textual data, whereas environmental data often comprise numerical vectors unseen to LLMs. To address this, KALM fine-tunes the LLM to perform various tasks based on environmental data, including bidirectional translation between natural language descriptions of skills and their corresponding rollout data. This grounding process enhances the LLM's comprehension of environmental dynamics, enabling it to generate diverse and meaningful imaginary rollouts that reflect novel skills. Initial empirical evaluations on the CLEVR-Robot environment demonstrate that KALM enables agents to complete complex rephrasings of task goals and extend their capabilities to novel tasks requiring unprecedented optimal behaviors. KALM achieves a success rate of 46% in executing tasks with unseen goals, substantially surpassing the 26% success rate achieved by baseline methods. Furthermore, KALM effectively enables the LLM to comprehend environmental dynamics, resulting in the generation of meaningful imaginary rollouts that reflect novel skills and demonstrate the seamless integration of large language models and reinforcement learning.

cs.LG↗

arXiv2024

Principle of Information Increase: An Operational Perspective of Information Gain in the Foundations of Quantum Theory

A measurement performed on a quantum system is an act of gaining information about its state, a view that is widespread in practical and foundational work in quantum theory. However, the concept of information in quantum theory reconstructions is multiply-defined, and its conceptual foundations remain surprisingly under-explored. In this paper, we investigate the gain of information in quantum measurements from an operational viewpoint. We show that the continuous extension of the Shannon entropy naturally admits two distinct measures of information gain, differential information gain and relative information gain, and that these have radically different characteristics. In particular, while differential information gain can increase or decrease as additional data is acquired, relative information gain consistently grows, and moreover exhibits asymptotic indifference to the data or choice of Bayesian prior. In order to make a principled choice between these measures, we articulate a Principle of Information Increase, which incorporates Summhammer's proposal that more data from measurements leads to more knowledge about the system, and also takes into consideration black swan events. This principle favors differential information gain as the more relevant metric in two-outcome quantum systems, and guides the selection of priors for these information measures. Finally, we show that, of the beta distribution priors, the Jeffreys' binomial prior is the prior ensures maximal robustness of information gain to the particular data sequence obtained in a run of experiments.

quant-ph↗

arXiv2024

Disentangling Policy from Offline Task Representation Learning via Adversarial Data Augmentation

Offline meta-reinforcement learning (OMRL) proficiently allows an agent to tackle novel tasks while solely relying on a static dataset. For precise and efficient task identification, existing OMRL research suggests learning separate task representations that be incorporated with policy input, thus forming a context-based meta-policy. A major approach to train task representations is to adopt contrastive learning using multi-task offline data. The dataset typically encompasses interactions from various policies (i.e., the behavior policies), thus providing a plethora of contextual information regarding different tasks. Nonetheless, amassing data from a substantial number of policies is not only impractical but also often unattainable in realistic settings. Instead, we resort to a more constrained yet practical scenario, where multi-task data collection occurs with a limited number of policies. We observed that learned task representations from previous OMRL methods tend to correlate spuriously with the behavior policy instead of reflecting the essential characteristics of the task, resulting in unfavorable out-of-distribution generalization. To alleviate this issue, we introduce a novel algorithm to disentangle the impact of behavior policy from task representation learning through a process called adversarial data augmentation. Specifically, the objective of adversarial data augmentation is not merely to generate data analogous to offline data distribution; instead, it aims to create adversarial examples designed to confound learned task representations and lead to incorrect task identification. Our experiments show that learning from such adversarial samples significantly enhances the robustness and effectiveness of the task identification process and realizes satisfactory out-of-distribution generalization.

cs.LG↗

arXiv2024

The Structure of Quantum Questions

In classical physics, a single measurement can in principle reveal the state of a system. However, quantum theory permits numerous non-equivalent measurements on a physical system, each providing only limited information about the state. This set of various measurements on a quantum system indicates a rich internal structure. We illuminate this structure for both individual and composite systems by conceptualizing measurements as questions with a finite number of outcomes. We create a mathematical question structure to explore the underlying properties, employing the concept of information as a key tool representing our knowledge gained from asking these questions. We subsequently propose informational assumptions based on properties observed from measurements on qubits, generalizing these to higher dimensional systems. Our informational assumptions shape the correlations between subsystems, which are symbolized as classical logical gates. Interestingly, systems with prime number dimensions exhibit unique property: the logical gate can be expressed simply as a linear equation under modular arithmetic. We also identify structures in quantum theory that correspond to those in the structure of quantum questions. For instance, the questions determining the system correspond to generalized Pauli matrices, and the logical gate connecting questions in subsystems is directly related to the tensor product combining operators. Based on these correspondences, we present two equivalent scenarios regarding the evolution of systems and the change of information within both quantum questions and quantum mechanics.

quant-ph↗

arXiv2024

SplatFlow: Learning Multi-frame Optical Flow via Splatting

The occlusion problem remains a crucial challenge in optical flow estimation (OFE). Despite the recent significant progress brought about by deep learning, most existing deep learning OFE methods still struggle to handle occlusions; in particular, those based on two frames cannot correctly handle occlusions because occluded regions have no visual correspondences. However, there is still hope in multi-frame settings, which can potentially mitigate the occlusion issue in OFE. Unfortunately, multi-frame OFE (MOFE) remains underexplored, and the limited studies on it are mainly specially designed for pyramid backbones or else obtain the aligned previous frame's features, such as correlation volume and optical flow, through time-consuming backward flow calculation or non-differentiable forward warping transformation. This study proposes an efficient MOFE framework named SplatFlow to address these shortcomings. SplatFlow introduces the differentiable splatting transformation to align the previous frame's motion feature and designs a Final-to-All embedding method to input the aligned motion feature into the current frame's estimation, thus remodeling the existing two-frame backbones. The proposed SplatFlow is efficient yet more accurate, as it can handle occlusions properly. Extensive experimental evaluations show that SplatFlow substantially outperforms all published methods on the KITTI2015 and Sintel benchmarks. Especially on the Sintel benchmark, SplatFlow achieves errors of 1.12 (clean pass) and 2.07 (final pass), with surprisingly significant 19.4% and 16.2% error reductions, respectively, from the previous best results submitted. The code for SplatFlow is available at https://github.com/wwsource/SplatFlow.

cs.CV↗

arXiv2024

Policy Optimization in RLHF: The Impact of Out-of-preference Data

Aligning intelligent agents with human preferences and values is important. This paper examines two popular alignment methods: Direct Preference Optimization (DPO) and Reward-Model-Based Policy Optimization (RMB-PO). A variant of RMB-PO, referred to as RMB-PO+ is also considered. These methods, either explicitly or implicitly, learn a reward model from preference data and differ in the data used for policy optimization to unlock the generalization ability of the reward model. In particular, compared with DPO, RMB-PO additionally uses policy-generated data, and RMB-PO+ further leverages new, preference-free data. We examine the impact of such out-of-preference data. Our study, conducted through controlled and synthetic experiments, demonstrates that DPO performs poorly, whereas RMB-PO+ performs the best. In particular, even when providing the policy model with a good feature representation, we find that policy optimization with adequate out-of-preference data significantly improves performance by harnessing the reward model's generalization capabilities.

cs.LG↗

arXiv2024

Empowering Language Models with Active Inquiry for Deeper Understanding

The rise of large language models (LLMs) has revolutionized the way that we interact with artificial intelligence systems through natural language. However, LLMs often misinterpret user queries because of their uncertain intention, leading to less helpful responses. In natural human interactions, clarification is sought through targeted questioning to uncover obscure information. Thus, in this paper, we introduce LaMAI (Language Model with Active Inquiry), designed to endow LLMs with this same level of interactive engagement. LaMAI leverages active learning techniques to raise the most informative questions, fostering a dynamic bidirectional dialogue. This approach not only narrows the contextual gap but also refines the output of the LLMs, aligning it more closely with user expectations. Our empirical studies, across a variety of complex datasets where LLMs have limited conversational context, demonstrate the effectiveness of LaMAI. The method improves answer accuracy from 31.9% to 50.9%, outperforming other leading question-answering frameworks. Moreover, in scenarios involving human participants, LaMAI consistently generates responses that are superior or comparable to baseline methods in more than 82% of the cases. The applicability of LaMAI is further evidenced by its successful integration with various LLMs, highlighting its potential for the future of interactive language models.

cs.CL↗

arXiv2024

PALoc: Advancing SLAM Benchmarking with Prior-Assisted 6-DoF Trajectory Generation and Uncertainty Estimation

Accurately generating ground truth (GT) trajectories is essential for Simultaneous Localization and Mapping (SLAM) evaluation, particularly under varying environmental conditions. This study introduces a systematic approach employing a prior map-assisted framework for generating dense six-degree-of-freedom (6-DoF) GT poses for the first time, enhancing the fidelity of both indoor and outdoor SLAM datasets. Our method excels in handling degenerate and stationary conditions frequently encountered in SLAM datasets, thereby increasing robustness and precision. A significant aspect of our approach is the detailed derivation of covariances within the factor graph, enabling an in-depth analysis of pose uncertainty propagation. This analysis crucially contributes to demonstrating specific pose uncertainties and enhancing trajectory reliability from both theoretical and empirical perspectives. Additionally, we provide an open-source toolbox (https://github.com/JokerJohn/Cloud_Map_Evaluation) for map evaluation criteria, facilitating the indirect assessment of overall trajectory precision. Experimental results show at least a 30\% improvement in map accuracy and a 20\% increase in direct trajectory accuracy compared to the Iterative Closest Point (ICP) \cite{sharp2002icp} algorithm across diverse campus environments, with substantially enhanced robustness. Our open-source solution (https://github.com/JokerJohn/PALoc), extensively applied in the FusionPortable\cite{Jiao2022Mar} dataset, is geared towards SLAM benchmark dataset augmentation and represents a significant advancement in SLAM evaluations.

cs.RO↗

arXiv2024

ACT: Empowering Decision Transformer with Dynamic Programming via Advantage Conditioning

Decision Transformer (DT), which employs expressive sequence modeling techniques to perform action generation, has emerged as a promising approach to offline policy optimization. However, DT generates actions conditioned on a desired future return, which is known to bear some weaknesses such as the susceptibility to environmental stochasticity. To overcome DT's weaknesses, we propose to empower DT with dynamic programming. Our method comprises three steps. First, we employ in-sample value iteration to obtain approximated value functions, which involves dynamic programming over the MDP structure. Second, we evaluate action quality in context with estimated advantages. We introduce two types of advantage estimators, IAE and GAE, which are suitable for different tasks. Third, we train an Advantage-Conditioned Transformer (ACT) to generate actions conditioned on the estimated advantages. Finally, during testing, ACT generates actions conditioned on a desired advantage. Our evaluation results validate that, by leveraging the power of dynamic programming, ACT demonstrates effective trajectory stitching and robust action generation in spite of the environmental stochasticity, outperforming baseline methods across various benchmarks. Additionally, we conduct an in-depth analysis of ACT's various design choices through ablation studies. Our code is available at https://github.com/LAMDA-RL/ACT.

cs.LG↗

arXiv2024

The effective field theory approach to the strong coupling issue in $f(T)$ gravity with a non-minimally coupled scalar field

The Hamiltonian analysis for $f(T)$ gravity implies the existence of at least one scalar-type degree of freedom (DoF). However, this scalar DoF of $f(T)$ gravity does not manifest in linear perturbations around a cosmological background, which indicates an underlying strong coupling problem. In this work we expand the scope by introducing an extra scalar field non-minimally coupled to $f(T)$ gravity, aiming to address or alleviate the aforementioned strong coupling problem. Employing the effective field theory (EFT) approach, we provide a class of torsional EFT forms up to second order operators, avoiding the Ostrogradsky ghost. To illustrate this phenomenon, we study a simple model and perform a detailed analysis of its linear scalar perturbations. The results demonstrate that the coupling terms in this toy model are necessary to avoid the initial degenerate situation. The complete avoidance of new constraints requires more coupling terms. Once this vanishing scalar DoF starts propagating in cosmological background at linear level, this phenomenon will demand a revisit of the strong coupling issue that arises in $f(T)$ gravity, particularly in the presence of matter coupling.

gr-qc↗

arXiv2024

FedGT: Federated Node Classification with Scalable Graph Transformer

Graphs are widely used to model relational data. As graphs are getting larger and larger in real-world scenarios, there is a trend to store and compute subgraphs in multiple local systems. For example, recently proposed \emph{subgraph federated learning} methods train Graph Neural Networks (GNNs) distributively on local subgraphs and aggregate GNN parameters with a central server. However, existing methods have the following limitations: (1) The links between local subgraphs are missing in subgraph federated learning. This could severely damage the performance of GNNs that follow message-passing paradigms to update node/edge features. (2) Most existing methods overlook the subgraph heterogeneity issue, brought by subgraphs being from different parts of the whole graph. To address the aforementioned challenges, we propose a scalable \textbf{Fed}erated \textbf{G}raph \textbf{T}ransformer (\textbf{FedGT}) in the paper. Firstly, we design a hybrid attention scheme to reduce the complexity of the Graph Transformer to linear while ensuring a global receptive field with theoretical bounds. Specifically, each node attends to the sampled local neighbors and a set of curated global nodes to learn both local and global information and be robust to missing links. The global nodes are dynamically updated during training with an online clustering algorithm to capture the data distribution of the corresponding local subgraph. Secondly, FedGT computes clients' similarity based on the aligned global nodes with optimal transport. The similarity is then used to perform weighted averaging for personalized aggregation, which well addresses the data heterogeneity problem. Moreover, local differential privacy is applied to further protect the privacy of clients. Finally, extensive experimental results on 6 datasets and 2 subgraph settings demonstrate the superiority of FedGT.

cs.LG↗

arXiv2024

Beimingwu: A Learnware Dock System

The learnware paradigm proposed by Zhou [2016] aims to enable users to reuse numerous existing well-trained models instead of building machine learning models from scratch, with the hope of solving new user tasks even beyond models' original purposes. In this paradigm, developers worldwide can submit their high-performing models spontaneously to the learnware dock system (formerly known as learnware market) without revealing their training data. Once the dock system accepts the model, it assigns a specification and accommodates the model. This specification allows the model to be adequately identified and assembled to reuse according to future users' needs, even if they have no prior knowledge of the model. This paradigm greatly differs from the current big model direction and it is expected that a learnware dock system housing millions or more high-performing models could offer excellent capabilities for both planned tasks where big models are applicable; and unplanned, specialized, data-sensitive scenarios where big models are not present or applicable. This paper describes Beimingwu, the first open-source learnware dock system providing foundational support for future research of learnware paradigm.The system significantly streamlines the model development for new user tasks, thanks to its integrated architecture and engine design, extensive engineering implementations and optimizations, and the integration of various algorithms for learnware identification and reuse. Notably, this is possible even for users with limited data and minimal expertise in machine learning, without compromising the raw data's security. Beimingwu supports the entire process of learnware paradigm. The system lays the foundation for future research in learnware-related algorithms and systems, and prepares the ground for hosting a vast array of learnwares and establishing a learnware ecosystem.

cs.SE↗

arXiv2023

Generalizable Task Representation Learning for Offline Meta-Reinforcement Learning with Data Limitations

Generalization and sample efficiency have been long-standing issues concerning reinforcement learning, and thus the field of Offline Meta-Reinforcement Learning~(OMRL) has gained increasing attention due to its potential of solving a wide range of problems with static and limited offline data. Existing OMRL methods often assume sufficient training tasks and data coverage to apply contrastive learning to extract task representations. However, such assumptions are not applicable in several real-world applications and thus undermine the generalization ability of the representations. In this paper, we consider OMRL with two types of data limitations: limited training tasks and limited behavior diversity and propose a novel algorithm called GENTLE for learning generalizable task representations in the face of data limitations. GENTLE employs Task Auto-Encoder~(TAE), which is an encoder-decoder architecture to extract the characteristics of the tasks. Unlike existing methods, TAE is optimized solely by reconstruction of the state transition and reward, which captures the generative structure of the task models and produces generalizable representations when training tasks are limited. To alleviate the effect of limited behavior diversity, we consistently construct pseudo-transitions to align the data distribution used to train TAE with the data distribution encountered during testing. Empirically, GENTLE significantly outperforms existing OMRL methods on both in-distribution tasks and out-of-distribution tasks across both the given-context protocol and the one-shot protocol.

cs.LG↗