Qi Meng

Zekai Yu, Qi Meng, Qizhi Chu et al.

Tool calling extends large language models (LLMs) by enabling grounded interaction with external executable interfaces, thereby supporting environment-coupled problem solving. However, mainstream in-context learning (ICL) approaches typically incorporate detailed tool documentation and usage examples directly into the context. This results in substantial inference overhead and heightened risks of hallucination as the context length grows. Conversely, while tuning-based methods improve general tool-calling capabilities, they often fail to effectively internalize the specific details of previously seen tools, thereby retaining a dependency on in-context documentation. To address these limitations, we propose ParaTool, a framework that projects each tool into a dedicated, loadable set of parameters. By equipping a dynamic integration of these parameterized tools, the LLM can perform tool calling without relying on in-context documents or examples. Specifically, our approach consists of three stages: (1) parametric tool pre-training encapsulates the knowledge of different tools into independent parameter modules; (2) soft tool selection employs a gating network to dynamically weigh and aggregate relevant tool parameters; and (3) parametric tool fine-tuning jointly updates tool parameters to align the training and inference processes. Experiments on Stable ToolBench and BFCL demonstrate that ParaTool significantly outperforms strong ICL-based baselines, achieving superior performance while reducing computational complexity.

Rui Zhang, Qi Meng, Rongchan Zhu et al.

In scenarios with limited available data, training the function-to-function neural PDE solver in an unsupervised manner is essential. However, the efficiency and accuracy of existing methods are constrained by the properties of numerical algorithms, such as finite difference and pseudo-spectral methods, integrated during the training stage. These methods necessitate careful spatiotemporal discretization to achieve reasonable accuracy, leading to significant computational challenges and inaccurate simulations, particularly in cases with substantial spatiotemporal variations. To address these limitations, we propose the Monte Carlo Neural PDE Solver (MCNP Solver) for training unsupervised neural solvers via the PDEs' probabilistic representation, which regards macroscopic phenomena as ensembles of random particles. Compared to other unsupervised methods, MCNP Solver naturally inherits the advantages of the Monte Carlo method, which is robust against spatiotemporal variations and can tolerate coarse step size. In simulating the trajectories of particles, we employ Heun's method for the convection process and calculate the expectation via the probability density function of neighbouring grid points during the diffusion process. These techniques enhance accuracy and circumvent the computational issues associated with Monte Carlo sampling. Our numerical experiments on convection-diffusion, Allen-Cahn, and Navier-Stokes equations demonstrate significant improvements in accuracy and efficiency compared to other unsupervised baselines. The source code will be publicly available at: https://github.com/optray/MCNP.

Rui Zhang, Qi Meng, Zhi-Ming Ma

Neural operators have been explored as surrogate models for simulating physical systems to overcome the limitations of traditional partial differential equation (PDE) solvers. However, most existing operator learning methods assume that the data originate from a single physical mechanism, limiting their applicability and performance in more realistic scenarios. To this end, we propose Physical Invariant Attention Neural Operator (PIANO) to decipher and integrate the physical invariants (PI) for operator learning from the PDE series with various physical mechanisms. PIANO employs self-supervised learning to extract physical knowledge and attention mechanisms to integrate them into dynamic convolutional layers. Compared to existing techniques, PIANO can reduce the relative error by 13.6\%-82.2\% on PDE forecasting tasks across varying coefficients, forces, or boundary conditions. Additionally, varied downstream tasks reveal that the PI embeddings deciphered by PIANO align well with the underlying invariants in the PDE systems, verifying the physical significance of PIANO. The source code will be publicly available at: https://github.com/optray/PIANO.

Chenglong Wang, Yifu Huo, Yang Gan et al.

Recent advances in multimodal reward modeling have been largely driven by a paradigm shift from discriminative to generative approaches. Building on this progress, recent studies have further employed reinforcement learning from verifiable rewards (RLVR) to enhance multimodal reward models (MRMs). Despite their success, RLVR-based training typically relies on labeled multimodal preference data, which are costly and labor-intensive to obtain, making it difficult to scale MRM training. To overcome this limitation, we propose a Multi-Stage Reinforcement Learning (MSRL) approach, which can achieve scalable RL for MRMs with limited multimodal data. MSRL replaces the conventional RLVR-based training paradigm by first learning a generalizable reward reasoning capability from large-scale textual preference data, and then progressively transferring this capability to multimodal tasks through caption-based and fully multimodal reinforcement-learning stages. Furthermore, we introduce a cross-modal knowledge distillation approach to improve preference generalization within MSRL. Extensive experiments demonstrate that MSRL effectively scales the RLVR-based training of generative MRMs and substantially improves their performance across both visual understanding and visual generation tasks (e.g., from 66.6% to 75.9% on VL-RewardBench and from 70.2% to 75.7% on GenAI-Bench), without requiring additional multimodal preference annotations. Our code is available at: https://github.com/wangclnlp/MSRL.

Bohan Wang, Yushun Zhang, Huishuai Zhang et al.

Adam is widely adopted in practical applications due to its fast convergence. However, its theoretical analysis is still far from satisfactory. Existing convergence analyses for Adam rely on the bounded smoothness assumption, referred to as the \emph{L-smooth condition}. Unfortunately, this assumption does not hold for many deep learning tasks. Moreover, we believe that this assumption obscures the true benefit of Adam, as the algorithm can adapt its update magnitude according to local smoothness. This important feature of Adam becomes irrelevant when assuming globally bounded smoothness. This paper studies the convergence of randomly reshuffled Adam (RR Adam) with diminishing learning rate, which is the major version of Adam adopted in deep learning tasks. We present the first convergence analysis of RR Adam without the bounded smoothness assumption. We demonstrate that RR Adam can maintain its convergence properties when smoothness is linearly bounded by the gradient norm, referred to as the \emph{$(L_0, L_1)$-smooth condition. We further compare Adam to SGD when both methods use diminishing learning rate. We refine the existing lower bound of SGD and show that SGD can be slower than Adam. To our knowledge, this is the first time that Adam and SGD are rigorously compared in the same setting and the advantage of Adam is revealed.

Rui Zhang, Peiyan Hu, Qi Meng et al.

Navier-Stokes equations are significant partial differential equations that describe the motion of fluids such as liquids and air. Due to the importance of Navier-Stokes equations, the development on efficient numerical schemes is important for both science and engineer. Recently, with the development of AI techniques, several approaches have been designed to integrate deep neural networks in simulating and inferring the fluid dynamics governed by incompressible Navier-Stokes equations, which can accelerate the simulation or inferring process in a mesh-free and differentiable way. In this paper, we point out that the capability of existing deep Navier-Stokes informed methods is limited to handle non-smooth or fractional equations, which are two critical situations in reality. To this end, we propose the \emph{Deep Random Vortex Method} (DRVM), which combines the neural network with a random vortex dynamics system equivalent to the Navier-Stokes equation. Specifically, the random vortex dynamics motivates a Monte Carlo based loss function for training the neural network, which avoids the calculation of derivatives through auto-differentiation. Therefore, DRVM not only can efficiently solve Navier-Stokes equations involving rough path, non-differentiable initial conditions and fractional operators, but also inherits the mesh-free and differentiable benefits of the deep-learning-based solver. We conduct experiments on the Cauchy problem, parametric solver learning, and the inverse problem of both 2-d and 3-d incompressible Navier-Stokes equations. The proposed method achieves accurate results for simulation and inference of Navier-Stokes equations. Especially for the cases that include singular initial conditions, DRVM significantly outperforms existing PINN method.

Xinquan Huang, Wenlei Shi, Qi Meng et al.

Neural networks have shown great potential in accelerating the solution of partial differential equations (PDEs). Recently, there has been a growing interest in introducing physics constraints into training neural PDE solvers to reduce the use of costly data and improve the generalization ability. However, these physics constraints, based on certain finite dimensional approximations over the function space, must resolve the smallest scaled physics to ensure the accuracy and stability of the simulation, resulting in high computational costs from large input, output, and neural networks. This paper proposes a general acceleration methodology called NeuralStagger by spatially and temporally decomposing the original learning tasks into several coarser-resolution subtasks. We define a coarse-resolution neural solver for each subtask, which requires fewer computational resources, and jointly train them with the vanilla physics-constrained loss by simply arranging their outputs to reconstruct the original solution. Due to the perfect parallelism between them, the solution is achieved as fast as a coarse-resolution neural solver. In addition, the trained solvers bring the flexibility of simulating with multiple levels of resolution. We demonstrate the successful application of NeuralStagger on 2D and 3D fluid dynamics simulations, which leads to an additional $10\sim100\times$ speed-up. Moreover, the experiment also shows that the learned model could be well used for optimal control.

Peiyan Hu, Qi Meng, Bingguang Chen et al.

Stochastic partial differential equations (SPDEs) are significant tools for modeling dynamics in many areas including atmospheric sciences and physics. Neural Operators, generations of neural networks with capability of learning maps between infinite-dimensional spaces, are strong tools for solving parametric PDEs. However, they lack the ability to modeling SPDEs which usually have poor regularity due to the driving noise. As the theory of regularity structure has achieved great successes in analyzing SPDEs and provides the concept model feature vectors that well-approximate SPDEs' solutions, we propose the Neural Operator with Regularity Structure (NORS) which incorporates the feature vectors for modeling dynamics driven by SPDEs. We conduct experiments on various of SPDEs including the dynamic Phi41 model and the 2d stochastic Navier-Stokes equation, and the results demonstrate that the NORS is resolution-invariant, efficient, and achieves one order of magnitude lower error with a modest amount of data.

Wei Chen, Weitao Du, Zhi-Ming Ma et al.

We study a kind of new SDE that was arisen from the research on optimization in machine learning, we call it power-law dynamic because its stationary distribution cannot have sub-Gaussian tail and obeys power-law. We prove that the power-law dynamic is ergodic with unique stationary distribution, provided the learning rate is small enough. We investigate its first exist time. In particular, we compare the exit times of the (continuous) power-law dynamic and its discretization. The comparison can help guide machine learning algorithm.

Zihan Wang, Qi Meng, HaiFeng Lan et al.

Speech emotion recognition (SER) classifies audio into emotion categories such as Happy, Angry, Fear, Disgust and Neutral. While Speech Emotion Recognition (SER) is a common application for popular languages, it continues to be a problem for low-resourced languages, i.e., languages with no pretrained speech-to-text recognition models. This paper firstly proposes a language-specific model that extract emotional information from multiple pre-trained speech models, and then designs a multi-domain model that simultaneously performs SER for various languages. Our multidomain model employs a multi-gating mechanism to generate unique weighted feature combination for each language, and also searches for specific neural network structure for each language through a neural architecture search module. In addition, we introduce a contrastive auxiliary loss to build more separable representations for audio data. Our experiments show that our model raises the state-of-the-art accuracy by 3% for German and 14.3% for French.

Nian Ran, Peng Xiao, Yue Wang et al.

The application of large deep learning models in weather forecasting has led to significant advancements in the field, including higher-resolution forecasting and extended prediction periods exemplified by models such as Pangu and Fuxi. Despite these successes, previous research has largely been characterized by the neglect of extreme weather events, and the availability of datasets specifically curated for such events remains limited. Given the critical importance of accurately forecasting extreme weather, this study introduces a comprehensive dataset that incorporates high-resolution extreme weather cases derived from the High-Resolution Rapid Refresh (HRRR) data, a 3-km real-time dataset provided by NOAA. We also evaluate the current state-of-the-art deep learning models and Numerical Weather Prediction (NWP) systems on HR-Extreme, and provide a improved baseline deep learning model called HR-Heim which has superior performance on both general loss and HR-Extreme compared to others. Our results reveal that the errors of extreme weather cases are significantly larger than overall forecast error, highlighting them as an crucial source of loss in weather prediction. These findings underscore the necessity for future research to focus on improving the accuracy of extreme weather forecasts to enhance their practical utility.

Yang Liu, Jinxuan Cai, Yishen Li et al.

Large language model-based (LLM-based) multi-agent systems (MAS) are increasingly used to extend agentic problem solving via role specialization and collaboration. MAS workflows can be naturally modeled as directed computation graphs, where nodes execute agents/sub-workflows and edges encode dependencies and message passing. However, implementing complex graph workflows in current frameworks still requires substantial manual effort, offers limited reuse, and makes it difficult to integrate heterogeneous external context sources. To overcome these limitations, we present MASFactory, a graph-centric framework for orchestrating LLM-based MAS. It introduces Vibe Graphing, a human-in-the-loop approach that compiles natural-language intent into an editable workflow specification and then into an executable graph. In addition, the framework provides reusable components and pluggable context integration, as well as a visualizer for topology preview, runtime tracing, and human-in-the-loop interaction. We evaluate MASFactory on seven public benchmarks, validating both reproduction consistency for representative MAS methods and the effectiveness of Vibe Graphing. Our code (https://github.com/BUPT-GAMMA/MASFactory) and video (https://youtu.be/ANynzVfY32k) are publicly available.

Tiantian Yang, Xuanle Ren, Qingdian Wan et al.

Adders are fundamental building blocks in modern digital systems, and their performance, power, and area (PPA) directly impact system efficiency. Contemporary adders typically use parallel-prefix architectures with established PPA trade-offs, but these often fail to deliver globally optimal PPA for specific design goals. Prior work lacks netlist-/cell-level awareness, and general synthesis heuristics are not adder-specific, resulting in suboptimal PPA. To address this, we propose AXON, an automated netlist optimization framework for adders. It performs design space exploration from architectural to netlist level, integrating prefix topology search with standard-cell-aware mapping via a hierarchical approach to quickly converge to near-optimal PPA solutions. We also introduce a hybrid ultra-high-speed adder combining parallel-prefix and Ling architectures to shorten the critical path. Experiments on TSMC 28nm library show AXON improves delay, area-delay product, and energy-delay product by up to 10.3%, 12.6%, and 32.1% respectively, compared to commercial synthesis tools.

Haosen Li, Qi Meng, Jiahao Li et al.

Partial differential equation (PDE) simulation holds extensive significance in scientific research. Currently, the integration of deep neural networks to learn solution operators of PDEs has introduced great potential. In this paper, we present UniFluids, a conditional flow-matching framework that harnesses the scalability of diffusion Transformer to unify learning of solution operators across diverse PDEs with varying dimensionality and physical variables. Unlike the autoregressive PDE foundation models, UniFluids adopts flow-matching to achieve parallel sequence generation, making it the first such approach for unified operator learning. Specifically, the introduction of a unified four-dimensional spatiotemporal representation for the heterogeneous PDE datasets enables joint training and conditional encoding. Furthermore, we find the effective dimension of the PDE dataset is much lower than its patch dimension. We thus employ $x$-prediction in the flow-matching operator learning, which is verified to significantly improve prediction accuracy. We conduct a large-scale evaluation of UniFluids on several PDE datasets covering spatial dimensions 1D, 2D and 3D. Experimental results show that UniFluids achieves strong prediction accuracy and demonstrates good scalability and cross-scenario generalization capability. The code will be released later.

Zheyan Li, Yuantu Zhu, Hao Ni et al.

Stochastic Partial Differential Equations (SPDEs) driven by random noise play a central role in modelling physical processes whose spatio-temporal dynamics can be rough, such as turbulence flows, superconductors, and quantum dynamics. To efficiently model these processes and make predictions, machine learning (ML)-based surrogate models are proposed, with their network architectures incorporating the spatio-temporal roughness in their design. However, it lacks an extensive and unified datasets for SPDE learning; especially, existing datasets do not account for the computational error introduced by noise sampling and the necessary renormalization required for handling singular SPDEs. We thus introduce SPDEBench, which is designed to solve typical SPDEs of physical significance (e.g., the $Φ^4_d$, wave, incompressible Navier--Stokes, and KdV equations) on 1D or 2D tori driven by white noise via ML methods. New datasets for singular SPDEs based on the renormalization process have been constructed, and novel ML models achieving the best results to date have been proposed. In particular, we investigate the impact of computational error introduced by noise sampling and renormalization on the performance comparison of ML models and highlight the importance of selecting high-quality test data for accurate evaluation. Results are benchmarked with traditional numerical solvers and ML-based models, including FNO, NSPDE and DLR-Net, etc. It is shown that, for singular SPDEs, naively applying ML models on data without specifying the numerical schemes can lead to significant errors and misleading conclusions. Our SPDEBench provides an open-source codebase that ensures full reproducibility of benchmarking across a variety of SPDE datasets while offering the flexibility to incorporate new datasets and machine learning baselines, making it a valuable resource for the community.

Jitai Hao, WeiWei Sun, Xin Xin et al.

Parameter-Efficient Fine-tuning (PEFT) facilitates the fine-tuning of Large Language Models (LLMs) under limited resources. However, the fine-tuning performance with PEFT on complex, knowledge-intensive tasks is limited due to the constrained model capacity, which originates from the limited number of additional trainable parameters. To overcome this limitation, we introduce a novel mechanism that fine-tunes LLMs with adapters of larger size yet memory-efficient. This is achieved by leveraging the inherent activation sparsity in the Feed-Forward Networks (FFNs) of LLMs and utilizing the larger capacity of Central Processing Unit (CPU) memory compared to Graphics Processing Unit (GPU). We store and update the parameters of larger adapters on the CPU. Moreover, we employ a Mixture of Experts (MoE)-like architecture to mitigate unnecessary CPU computations and reduce the communication volume between the GPU and CPU. This is particularly beneficial over the limited bandwidth of PCI Express (PCIe). Our method can achieve fine-tuning results comparable to those obtained with larger memory capacities, even when operating under more limited resources such as a 24GB memory single GPU setup, with acceptable loss in training efficiency. Our codes are available at https://github.com/CURRENTF/MEFT.

Xiaobo Liang, Lijun Wu, Juntao Li et al.

Dropout is a powerful and widely used technique to regularize the training of deep neural networks. In this paper, we introduce a simple regularization strategy upon dropout in model training, namely R-Drop, which forces the output distributions of different sub models generated by dropout to be consistent with each other. Specifically, for each training sample, R-Drop minimizes the bidirectional KL-divergence between the output distributions of two sub models sampled by dropout. Theoretical analysis reveals that R-Drop reduces the freedom of the model parameters and complements dropout. Experiments on $\bf{5}$ widely used deep learning tasks ($\bf{18}$ datasets in total), including neural machine translation, abstractive summarization, language understanding, language modeling, and image classification, show that R-Drop is universally effective. In particular, it yields substantial improvements when applied to fine-tune large-scale pre-trained models, e.g., ViT, RoBERTa-large, and BART, and achieves state-of-the-art (SOTA) performances with the vanilla Transformer model on WMT14 English$\to$German translation ($\bf{30.91}$ BLEU) and WMT14 English$\to$French translation ($\bf{43.95}$ BLEU), even surpassing models trained with extra large-scale data and expert-designed advanced variants of Transformer models. Our code is available at GitHub{\url{https://github.com/dropreg/R-Drop}}.

Bohan Wang, Huishuai Zhang, Qi Meng et al.

This paper aims to clearly distinguish between Stochastic Gradient Descent with Momentum (SGDM) and Adam in terms of their convergence rates. We demonstrate that Adam achieves a faster convergence compared to SGDM under the condition of non-uniformly bounded smoothness. Our findings reveal that: (1) in deterministic environments, Adam can attain the known lower bound for the convergence rate of deterministic first-order optimizers, whereas the convergence rate of Gradient Descent with Momentum (GDM) has higher order dependence on the initial function value; (2) in stochastic setting, Adam's convergence rate upper bound matches the lower bounds of stochastic first-order optimizers, considering both the initial function value and the final error, whereas there are instances where SGDM fails to converge with any learning rate. These insights distinctly differentiate Adam and SGDM regarding their convergence rates. Additionally, by introducing a novel stopping-time based technique, we further prove that if we consider the minimum gradient norm during iterations, the corresponding convergence rate can match the lower bounds across all problem hyperparameters. The technique can also help proving that Adam with a specific hyperparameter scheduler is parameter-agnostic, which hence can be of independent interest.

Rui Zhang, Qi Meng, Han Wan et al.

Computational fluid dynamics (CFD) drives progress in numerous scientific and engineering fields, yet high-fidelity simulations remain computationally prohibitive. While machine learning approaches offer computing acceleration, they typically specialize in single physical systems or require extensive training data, hindering their applicability in highly nonlinear and 3D flow scenarios. To overcome these limitations, we propose OmniFluids, a pure physics pre-trained model that captures fundamental fluid dynamics laws and adapts efficiently to diverse downstream tasks with minimal data. We develop a training framework combining physics-only pre-training, coarse-grid operator distillation, and few-shot fine-tuning. This enables OmniFluids to retain broad physics knowledge while delivering fast and accurate predictions. Architecturally, OmniFluids integrates a mixture of operators, a multi-frame decoder, and factorized Fourier layers, seamlessly incorporating physics-based supervision while allowing efficient and scalable modeling of diverse tasks. Extensive tests on a broad range of 2D and 3D benchmarks show that OmniFluids outperforms state-of-the-art AI-driven methods in terms of flow field prediction and turbulence statistics. It delivers 10--100$\times$ speedups over traditional solvers while maintaining a comparable accuracy and accurately identifies unknown physical parameters from sparse, noisy data. This work demonstrates the potential of training a unified CFD solver exclusively from physics knowledge, offering a new approach for efficient and generalizable modeling across complex fluid systems.

Jiawen Wu, Bingguang Chen, Yuyi Zhou et al.

Stochastic interpolants are efficient generative models that bridge two arbitrary probability density functions in finite time, enabling flexible generation from the source to the target distribution or vice versa. These models are primarily developed in Euclidean space, and are therefore limited in their application to many distribution learning problems defined on Riemannian manifolds in real-world scenarios. In this work, we introduce the Riemannian Neural Geodesic Interpolant (RNGI) model, which interpolates between two probability densities on a Riemannian manifold along the stochastic geodesics, and then samples from one endpoint as the final state using the continuous flow originating from the other endpoint. We prove that the temporal marginal density of RNGI solves a transport equation on the Riemannian manifold. After training the model's the neural velocity and score fields, we propose the Embedding Stochastic Differential Equation (E-SDE) algorithm for stochastic sampling of RNGI. E-SDE significantly improves the sampling quality by reducing the accumulated error caused by the excessive intrinsic discretization of Riemannian Brownian motion in the classical Geodesic Random Walk (GRW) algorithm. We also provide theoretical bounds on the generative bias measured in terms of KL-divergence. Finally, we demonstrate the effectiveness of the proposed RNGI and E-SDE through experiments conducted on both collected and synthetic distributions on S2 and SO(3).

Bohan Wang, Huishuai Zhang, Jieyu Zhang et al.

Recently, the information-theoretical framework has been proven to be able to obtain non-vacuous generalization bounds for large models trained by Stochastic Gradient Langevin Dynamics (SGLD) with isotropic noise. In this paper, we optimize the information-theoretical generalization bound by manipulating the noise structure in SGLD. We prove that with constraint to guarantee low empirical risk, the optimal noise covariance is the square root of the expected gradient covariance if both the prior and the posterior are jointly optimized. This validates that the optimal noise is quite close to the empirical gradient covariance. Technically, we develop a new information-theoretical bound that enables such an optimization analysis. We then apply matrix analysis to derive the form of optimal noise covariance. Presented constraint and results are validated by the empirical observations.

Weitao Du, He Zhang, Yuanqi Du et al.

Group equivariance (e.g. SE(3) equivariance) is a critical physical symmetry in science, from classical and quantum physics to computational biology. It enables robust and accurate prediction under arbitrary reference transformations. In light of this, great efforts have been put on encoding this symmetry into deep neural networks, which has been shown to improve the generalization performance and data efficiency for downstream tasks. Constructing an equivariant neural network generally brings high computational costs to ensure expressiveness. Therefore, how to better trade-off the expressiveness and computational efficiency plays a core role in the design of the equivariant deep learning models. In this paper, we propose a framework to construct SE(3) equivariant graph neural networks that can approximate the geometric quantities efficiently. Inspired by differential geometry and physics, we introduce equivariant local complete frames to graph neural networks, such that tensor information at given orders can be projected onto the frames. The local frame is constructed to form an orthonormal basis that avoids direction degeneration and ensure completeness. Since the frames are built only by cross product operations, our method is computationally efficient. We evaluate our method on two tasks: Newton mechanics modeling and equilibrium molecule conformation generation. Extensive experimental results demonstrate that our model achieves the best or competitive performance in two types of datasets.

Bohan Wang, Qi Meng, Huishuai Zhang et al.

The momentum acceleration technique is widely adopted in many optimization algorithms. However, there is no theoretical answer on how the momentum affects the generalization performance of the optimization algorithms. This paper studies this problem by analyzing the implicit regularization of momentum-based optimization. We prove that on the linear classification problem with separable data and exponential-tailed loss, gradient descent with momentum (GDM) converges to the L2 max-margin solution, which is the same as vanilla gradient descent. That means gradient descent with momentum acceleration still converges to a low-complexity model, which guarantees their generalization. We then analyze the stochastic and adaptive variants of GDM (i.e., SGDM and deterministic Adam) and show they also converge to the L2 max-margin solution. Technically, to overcome the difficulty of the error accumulation in analyzing the momentum, we construct new potential functions to analyze the gap between the model parameter and the max-margin solution. Numerical experiments are conducted and support our theoretical results.

Sang-gil Lee, Heeseung Kim, Chaehun Shin et al.

Denoising diffusion probabilistic models have been recently proposed to generate high-quality samples by estimating the gradient of the data density. The framework defines the prior noise as a standard Gaussian distribution, whereas the corresponding data distribution may be more complicated than the standard Gaussian distribution, which potentially introduces inefficiency in denoising the prior noise into the data sample because of the discrepancy between the data and the prior. In this paper, we propose PriorGrad to improve the efficiency of the conditional diffusion model for speech synthesis (for example, a vocoder using a mel-spectrogram as the condition) by applying an adaptive prior derived from the data statistics based on the conditional information. We formulate the training and sampling procedures of PriorGrad and demonstrate the advantages of an adaptive prior through a theoretical analysis. Focusing on the speech synthesis domain, we consider the recently proposed diffusion-based speech generative models based on both the spectral and time domains and show that PriorGrad achieves faster convergence and inference with superior performance, leading to an improved perceptual quality and robustness to a smaller network capacity, and thereby demonstrating the efficiency of a data-dependent adaptive prior.

Shiqi Gong, Qi Meng, Yue Wang et al.

Learning dynamics governed by differential equations is crucial for predicting and controlling the systems in science and engineering. Neural Ordinary Differential Equation (NODE), a deep learning model integrated with differential equations, is popular in learning dynamics recently due to its robustness to irregular samples and its flexibility to high-dimensional input. However, the training of NODE is sensitive to the precision of the numerical solver, which makes the convergence of NODE unstable, especially for ill-conditioned dynamical systems. In this paper, to reduce the reliance on the numerical solver, we propose to enhance the supervised signal in the training of NODE. Specifically, we pre-train a neural differential operator (NDO) to output an estimation of the derivatives to serve as an additional supervised signal. The NDO is pre-trained on a class of basis functions and learns the mapping between the trajectory samples of these functions to their derivatives. To leverage both the trajectory signal and the estimated derivatives from NDO, we propose an algorithm called NDO-NODE, in which the loss function contains two terms: the fitness on the true trajectory samples and the fitness on the estimated derivatives that are outputted by the pre-trained NDO. Experiments on various kinds of dynamics show that our proposed NDO-NODE can consistently improve the forecasting accuracy with one pre-trained NDO. Especially for the stiff ODEs, we observe that NDO-NODE can capture the transitions in the dynamics more accurately compared with other regularization methods.

Ziming Liu, Bohan Wang, Qi Meng et al.

Energy conservation is a basic physics principle, the breakdown of which often implies new physics. This paper presents a method for data-driven "new physics" discovery. Specifically, given a trajectory governed by unknown forces, our Neural New-Physics Detector (NNPhD) aims to detect new physics by decomposing the force field into conservative and non-conservative components, which are represented by a Lagrangian Neural Network (LNN) and a universal approximator network (UAN), respectively, trained to minimize the force recovery error plus a constant $λ$ times the magnitude of the predicted non-conservative force. We show that a phase transition occurs at $λ$=1, universally for arbitrary forces. We demonstrate that NNPhD successfully discovers new physics in toy numerical experiments, rediscovering friction (1493) from a damped double pendulum, Neptune from Uranus' orbit (1846) and gravitational waves (2017) from an inspiraling orbit. We also show how NNPhD coupled with an integrator outperforms previous methods for predicting the future of a damped double pendulum.

Zhen Wu, Lijun Wu, Qi Meng et al.

Transformer architecture achieves great success in abundant natural language processing tasks. The over-parameterization of the Transformer model has motivated plenty of works to alleviate its overfitting for superior performances. With some explorations, we find simple techniques such as dropout, can greatly boost model performance with a careful design. Therefore, in this paper, we integrate different dropout techniques into the training of Transformer models. Specifically, we propose an approach named UniDrop to unites three different dropout techniques from fine-grain to coarse-grain, i.e., feature dropout, structure dropout, and data dropout. Theoretically, we demonstrate that these three dropouts play different roles from regularization perspectives. Empirically, we conduct experiments on both neural machine translation and text classification benchmark datasets. Extensive results indicate that Transformer with UniDrop can achieve around 1.5 BLEU improvement on IWSLT14 translation tasks, and better accuracy for the classification even using strong pre-trained RoBERTa as backbone.

Bohan Wang, Qi Meng, Wei Chen et al.

Despite their overwhelming capacity to overfit, deep neural networks trained by specific optimization algorithms tend to generalize well to unseen data. Recently, researchers explained it by investigating the implicit regularization effect of optimization algorithms. A remarkable progress is the work (Lyu&Li, 2019), which proves gradient descent (GD) maximizes the margin of homogeneous deep neural networks. Except GD, adaptive algorithms such as AdaGrad, RMSProp and Adam are popular owing to their rapid training process. However, theoretical guarantee for the generalization of adaptive optimization algorithms is still lacking. In this paper, we study the implicit regularization of adaptive optimization algorithms when they are optimizing the logistic loss on homogeneous deep neural networks. We prove that adaptive algorithms that adopt exponential moving average strategy in conditioner (such as Adam and RMSProp) can maximize the margin of the neural network, while AdaGrad that directly sums historical squared gradients in conditioner can not. It indicates superiority on generalization of exponential moving average strategy in the design of the conditioner. Technically, we provide a unified framework to analyze convergent direction of adaptive optimization algorithms by constructing novel adaptive gradient flow and surrogate margin. Our experiments can well support the theoretical findings on convergent direction of adaptive optimization algorithms.

Qi Meng, Shiqi Gong, Wei Chen et al.

Stochastic gradient descent (SGD) and its variants are mainstream methods to train deep neural networks. Since neural networks are non-convex, more and more works study the dynamic behavior of SGD and the impact to its generalization, especially the escaping efficiency from local minima. However, these works take the over-simplified assumption that the covariance of the noise in SGD is (or can be upper bounded by) constant, although it is actually state-dependent. In this work, we conduct a formal study on the dynamic behavior of SGD with state-dependent noise. Specifically, we show that the covariance of the noise of SGD in the local region of the local minima is a quadratic function of the state. Thus, we propose a novel power-law dynamic with state-dependent diffusion to approximate the dynamic of SGD. We prove that, power-law dynamic can escape from sharp minima exponentially faster than flat minima, while the previous dynamics can only escape sharp minima polynomially faster than flat minima. Our experiments well verified our theoretical results. Inspired by our theory, we propose to add additional state-dependent noise into (large-batch) SGD to further improve its generalization ability. Experiments verify that our method is effective.

Juanping Zhu, Qi Meng, Wei Chen et al.

Based on basis path set, G-SGD algorithm significantly outperforms conventional SGD algorithm in optimizing neural networks. However, how the inner mechanism of basis paths work remains mysterious. From the aspect of graph theory, this paper defines basis path, investigates structure properties of basis paths in regular fully connected neural network and interprets the graph representation of basis path set. Moreover, we propose hierarchical algorithm HBPS to find basis path set B in fully connected neural network by decomposing the network into several independent and parallel substructures. Algorithm HBPS demands that there doesn't exist shared edges between any two independent substructure paths.

Ling Pan, Qingpeng Cai, Qi Meng et al.

Value function estimation is an important task in reinforcement learning, i.e., prediction. The Boltzmann softmax operator is a natural value estimator and can provide several benefits. However, it does not satisfy the non-expansion property, and its direct use may fail to converge even in value iteration. In this paper, we propose to update the value function with dynamic Boltzmann softmax (DBS) operator, which has good convergence property in the setting of planning and learning. Experimental results on GridWorld show that the DBS operator enables better estimation of the value function, which rectifies the convergence issue of the softmax operator. Finally, we propose the DBS-DQN algorithm by applying dynamic Boltzmann softmax updates in deep Q-network, which outperforms DQN substantially in 40 out of 49 Atari games.

Mingyang Yi, Qi Meng, Wei Chen et al.

It was empirically confirmed by Keskar et al.\cite{SharpMinima} that flatter minima generalize better. However, for the popular ReLU network, sharp minimum can also generalize well \cite{SharpMinimacan}. The conclusion demonstrates that the existing definitions of flatness fail to account for the complex geometry of ReLU neural networks because they can't cover the Positively Scale-Invariant (PSI) property of ReLU network. In this paper, we formalize the PSI causes problem of existing definitions of flatness and propose a new description of flatness - \emph{PSI-flatness}. PSI-flatness is defined on the values of basis paths \cite{GSGD} instead of weights. Values of basis paths have been shown to be the PSI-variables and can sufficiently represent the ReLU neural networks which ensure the PSI property of PSI-flatness. Then we study the relation between PSI-flatness and generalization theoretically and empirically. First, we formulate a generalization bound based on PSI-flatness which shows generalization error decreasing with the ratio between the largest basis path value and the smallest basis path value. That is to say, the minimum with balanced values of basis paths will more likely to be flatter and generalize better. Finally. we visualize the PSI-flatness of loss surface around two learned models which indicates the minimum with smaller PSI-flatness can indeed generalize better.

Yue Wang, Qi Meng, Wei Cheng et al.

Q-learning is one of the most popular methods in Reinforcement Learning (RL). Transfer Learning aims to utilize the learned knowledge from source tasks to help new tasks to improve the sample complexity of the new tasks. Considering that data collection in RL is both more time and cost consuming and Q-learning converges slowly comparing to supervised learning, different kinds of transfer RL algorithms are designed. However, most of them are heuristic with no theoretical guarantee of the convergence rate. Therefore, it is important for us to clearly understand when and how will transfer learning help RL method and provide the theoretical guarantee for the improvement of the sample complexity. In this paper, we propose to transfer the Q-function learned in the source task to the target of the Q-learning in the new task when certain safe conditions are satisfied. We call this new transfer Q-learning method target transfer Q-Learning. The safe conditions are necessary to avoid the harm to the new tasks and thus ensure the convergence of the algorithm. We study the convergence rate of the target transfer Q-learning. We prove that if the two tasks are similar with respect to the MDPs, the optimal Q-functions in the source and new RL tasks are similar which means the error of the transferred target Q-function in new MDP is small. Also, the convergence rate analysis shows that the target transfer Q-Learning will converge faster than Q-learning if the error of the transferred target Q-function is smaller than the current Q-function in the new task. Based on our theoretical results, we design the safe condition as the Bellman error of the transferred target Q-function is less than the current Q-function. Our experiments are consistent with our theoretical founding and verified the effectiveness of our proposed target transfer Q-learning method.

Shuxin Zheng, Qi Meng, Huishuai Zhang et al.

Recently, path norm was proposed as a new capacity measure for neural networks with Rectified Linear Unit (ReLU) activation function, which takes the rescaling-invariant property of ReLU into account. It has been shown that the generalization error bound in terms of the path norm explains the empirical generalization behaviors of the ReLU neural networks better than that of other capacity measures. Moreover, optimization algorithms which take path norm as the regularization term to the loss function, like Path-SGD, have been shown to achieve better generalization performance. However, the path norm counts the values of all paths, and hence the capacity measure based on path norm could be improperly influenced by the dependency among different paths. It is also known that each path of a ReLU network can be represented by a small group of linearly independent basis paths with multiplication and division operation, which indicates that the generalization behavior of the network only depends on only a few basis paths. Motivated by this, we propose a new norm \emph{Basis-path Norm} based on a group of linearly independent paths to measure the capacity of neural networks more accurately. We establish a generalization error bound based on this basis path norm, and show it explains the generalization behaviors of ReLU networks more accurately than previous capacity measures via extensive experiments. In addition, we develop optimization algorithms which minimize the empirical risk regularized by the basis-path norm. Our experiments on benchmark datasets demonstrate that the proposed regularization method achieves clearly better performance on the test set than the previous regularization approaches.

Li He, Qi Meng, Wei Chen et al.

Asynchronous stochastic gradient descent (ASGD) is a popular parallel optimization algorithm in machine learning. Most theoretical analysis on ASGD take a discrete view and prove upper bounds for their convergence rates. However, the discrete view has its intrinsic limitations: there is no characterization of the optimization path and the proof techniques are induction-based and thus usually complicated. Inspired by the recent successful adoptions of stochastic differential equations (SDE) to the theoretical analysis of SGD, in this paper, we study the continuous approximation of ASGD by using stochastic differential delay equations (SDDE). We introduce the approximation method and study the approximation error. Then we conduct theoretical analysis on the convergence rates of ASGD algorithm based on the continuous approximation. There are two methods: moment estimation and energy function minimization can be used to analyze the convergence rates. Moment estimation depends on the specific form of the loss function, while energy function minimization only leverages the convex property of the loss function, and does not depend on its specific form. In addition to the convergence analysis, the continuous view also helps us derive better convergence rates. All of this clearly shows the advantage of taking the continuous view in gradient descent algorithms.

Qi Meng, Shuxin Zheng, Huishuai Zhang et al.

It is well known that neural networks with rectified linear units (ReLU) activation functions are positively scale-invariant. Conventional algorithms like stochastic gradient descent optimize the neural networks in the vector space of weights, which is, however, not positively scale-invariant. This mismatch may lead to problems during the optimization process. Then, a natural question is: \emph{can we construct a new vector space that is positively scale-invariant and sufficient to represent ReLU neural networks so as to better facilitate the optimization process }? In this paper, we provide our positive answer to this question. First, we conduct a formal study on the positive scaling operators which forms a transformation group, denoted as $\mathcal{G}$. We show that the value of a path (i.e. the product of the weights along the path) in the neural network is invariant to positive scaling and prove that the value vector of all the paths is sufficient to represent the neural networks under mild conditions. Second, we show that one can identify some basis paths out of all the paths and prove that the linear span of their value vectors (denoted as $\mathcal{G}$-space) is an invariant space with lower dimension under the positive scaling group. Finally, we design stochastic gradient descent algorithm in $\mathcal{G}$-space (abbreviated as $\mathcal{G}$-SGD) to optimize the value vector of the basis paths of neural networks with little extra cost by leveraging back-propagation. Our experiments show that $\mathcal{G}$-SGD significantly outperforms the conventional SGD algorithm in optimizing ReLU networks on benchmark datasets.

Qi Meng, Wei Chen, Yue Wang et al.

When using stochastic gradient descent to solve large-scale machine learning problems, a common practice of data processing is to shuffle the training data, partition the data across multiple machines if needed, and then perform several epochs of training on the re-shuffled (either locally or globally) data. The above procedure makes the instances used to compute the gradients no longer independently sampled from the training data set. Then does the distributed SGD method have desirable convergence properties in this practical situation? In this paper, we give answers to this question. First, we give a mathematical formulation for the practical data processing procedure in distributed machine learning, which we call data partition with global/local shuffling. We observe that global shuffling is equivalent to without-replacement sampling if the shuffling operations are independent. We prove that SGD with global shuffling has convergence guarantee in both convex and non-convex cases. An interesting finding is that, the non-convex tasks like deep learning are more suitable to apply shuffling comparing to the convex tasks. Second, we conduct the convergence analysis for SGD with local shuffling. The convergence rate for local shuffling is slower than that for global shuffling, since it will lose some information if there's no communication between partitioned data. Finally, we consider the situation when the permutation after shuffling is not uniformly distributed (insufficient shuffling), and discuss the condition under which this insufficiency will not influence the convergence rate. Our theoretical results provide important insights to large-scale machine learning, especially in the selection of data processing methods in order to achieve faster convergence and good speedup. Our theoretical findings are verified by extensive experiments on logistic regression and deep neural networks.

Qi Meng, Guolin Ke, Taifeng Wang et al.

Decision tree (and its extensions such as Gradient Boosting Decision Trees and Random Forest) is a widely used machine learning algorithm, due to its practical effectiveness and model interpretability. With the emergence of big data, there is an increasing need to parallelize the training process of decision tree. However, most existing attempts along this line suffer from high communication costs. In this paper, we propose a new algorithm, called \emph{Parallel Voting Decision Tree (PV-Tree)}, to tackle this challenge. After partitioning the training data onto a number of (e.g., $M$) machines, this algorithm performs both local voting and global voting in each iteration. For local voting, the top-$k$ attributes are selected from each machine according to its local data. Then, globally top-$2k$ attributes are determined by a majority voting among these local candidates. Finally, the full-grained histograms of the globally top-$2k$ attributes are collected from local machines in order to identify the best (most informative) attribute and its split point. PV-Tree can achieve a very low communication cost (independent of the total number of attributes) and thus can scale out very well. Furthermore, theoretical analysis shows that this algorithm can learn a near optimal decision tree, since it can find the best attribute with a large probability. Our experiments on real-world datasets show that PV-Tree significantly outperforms the existing parallel decision tree algorithms in the trade-off between accuracy and efficiency.

Qi Meng, Wei Chen, Jingcheng Yu et al.

Regularized empirical risk minimization (R-ERM) is an important branch of machine learning, since it constrains the capacity of the hypothesis space and guarantees the generalization ability of the learning algorithm. Two classic proximal optimization algorithms, i.e., proximal stochastic gradient descent (ProxSGD) and proximal stochastic coordinate descent (ProxSCD) have been widely used to solve the R-ERM problem. Recently, variance reduction technique was proposed to improve ProxSGD and ProxSCD, and the corresponding ProxSVRG and ProxSVRCD have better convergence rate. These proximal algorithms with variance reduction technique have also achieved great success in applications at small and moderate scales. However, in order to solve large-scale R-ERM problems and make more practical impacts, the parallel version of these algorithms are sorely needed. In this paper, we propose asynchronous ProxSVRG (Async-ProxSVRG) and asynchronous ProxSVRCD (Async-ProxSVRCD) algorithms, and prove that Async-ProxSVRG can achieve near linear speedup when the training data is sparse, while Async-ProxSVRCD can achieve near linear speedup regardless of the sparse condition, as long as the number of block partitions are appropriately set. We have conducted experiments on a regularized logistic regression task. The results verified our theoretical findings and demonstrated the practical efficiency of the asynchronous stochastic proximal algorithms with variance reduction.

Qi Meng, Yue Wang, Wei Chen et al.

Many machine learning tasks can be formulated as Regularized Empirical Risk Minimization (R-ERM), and solved by optimization algorithms such as gradient descent (GD), stochastic gradient descent (SGD), and stochastic variance reduction (SVRG). Conventional analysis on these optimization algorithms focuses on their convergence rates during the training process, however, people in the machine learning community may care more about the generalization performance of the learned model on unseen test data. In this paper, we investigate on this issue, by using stability as a tool. In particular, we decompose the generalization error for R-ERM, and derive its upper bound for both convex and non-convex cases. In convex cases, we prove that the generalization error can be bounded by the convergence rate of the optimization algorithm and the stability of the R-ERM process, both in expectation (in the order of $\mathcal{O}((1/n)+\mathbb{E}ρ(T))$, where $ρ(T)$ is the convergence error and $T$ is the number of iterations) and in high probability (in the order of $\mathcal{O}\left(\frac{\log{1/δ}}{\sqrt{n}}+ρ(T)\right)$ with probability $1-δ$). For non-convex cases, we can also obtain a similar expected generalization error bound. Our theorems indicate that 1) along with the training process, the generalization error will decrease for all the optimization algorithms under our investigation; 2) Comparatively speaking, SVRG has better generalization ability than GD and SGD. We have conducted experiments on both convex and non-convex problems, and the experimental results verify our theoretical findings.

Shuxin Zheng, Qi Meng, Taifeng Wang et al.

With the fast development of deep learning, it has become common to learn big neural networks using massive training data. Asynchronous Stochastic Gradient Descent (ASGD) is widely adopted to fulfill this task for its efficiency, which is, however, known to suffer from the problem of delayed gradients. That is, when a local worker adds its gradient to the global model, the global model may have been updated by other workers and this gradient becomes "delayed". We propose a novel technology to compensate this delay, so as to make the optimization behavior of ASGD closer to that of sequential SGD. This is achieved by leveraging Taylor expansion of the gradient function and efficient approximation to the Hessian matrix of the loss function. We call the new algorithm Delay Compensated ASGD (DC-ASGD). We evaluated the proposed algorithm on CIFAR-10 and ImageNet datasets, and the experimental results demonstrate that DC-ASGD outperforms both synchronous SGD and asynchronous SGD, and nearly approaches the performance of sequential SGD.