arXiv subjects

Pei Sun

Publications and source records attributed to Pei Sun.

At least 37 records · Page 2Linked to original sources

arXiv2022

Off-diagonal Bethe Ansatz for the $D^{(1)}_3$ model

The exact solutions of the $D^{(1)}_3$ model (or the $so(6)$ quantum spin chain) with either periodic or general integrable open boundary conditions are obtained by using the off-diagonal Bethe Ansatz. From the fusion, the complete operator product identities are obtained, which are sufficient to enable us to determine spectrum of the system. Eigenvalues of the fused transfer matrices are constructed by the $T-Q$ relations for the periodic case and by the inhomogeneous $T-Q$ one for the non-diagonal boundary reflection case. The present method can be generalized to deal with the $D^{(1)}_{n}$ model directly.

math-ph↗

arXiv2022

Statistical Physics of Deep Neural Networks: Initialization toward Optimal Channels

In deep learning, neural networks serve as noisy channels between input data and its representation. This perspective naturally relates deep learning with the pursuit of constructing channels with optimal performance in information transmission and representation. While considerable efforts are concentrated on realizing optimal channel properties during network optimization, we study a frequently overlooked possibility that neural networks can be initialized toward optimal channels. Our theory, consistent with experimental validation, identifies primary mechanics underlying this unknown possibility and suggests intrinsic connections between statistical physics and deep learning. Unlike the conventional theories that characterize neural networks applying the classic mean-filed approximation, we offer analytic proof that this extensively applied simplification scheme is not valid in studying neural networks as information channels. To fill this gap, we develop a corrected mean-field framework applicable for characterizing the limiting behaviors of information propagation in neural networks without strong assumptions on inputs. Based on it, we propose an analytic theory to prove that mutual information maximization is realized between inputs and propagated signals when neural networks are initialized at dynamic isometry, a case where information transmits via norm-preserving mappings. These theoretical predictions are validated by experiments on real neural networks, suggesting the robustness of our theory against finite-size effects. Finally, we analyze our findings with information bottleneck theory to confirm the precise relations among dynamic isometry, mutual information maximization, and optimal channel properties in deep learning.

cs.LG↗

arXiv2022

LidarAugment: Searching for Scalable 3D LiDAR Data Augmentations

Data augmentations are important in training high-performance 3D object detectors for point clouds. Despite recent efforts on designing new data augmentations, perhaps surprisingly, most state-of-the-art 3D detectors only use a few simple data augmentations. In particular, different from 2D image data augmentations, 3D data augmentations need to account for different representations of input data and require being customized for different models, which introduces significant overhead. In this paper, we resort to a search-based approach, and propose LidarAugment, a practical and effective data augmentation strategy for 3D object detection. Unlike previous approaches where all augmentation policies are tuned in an exponentially large search space, we propose to factorize and align the search space of each data augmentation, which cuts down the 20+ hyperparameters to 2, and significantly reduces the search complexity. We show LidarAugment can be customized for different model architectures with different input representations by a simple 2D grid search, and consistently improve both convolution-based UPillars/StarNet/RSN and transformer-based SWFormer. Furthermore, LidarAugment mitigates overfitting and allows us to scale up 3D detectors to much larger capacity. In particular, by combining with latest 3D detectors, our LidarAugment achieves a new state-of-the-art 74.8 mAPH L2 on Waymo Open Dataset.

cs.CV↗

arXiv2022

SWFormer: Sparse Window Transformer for 3D Object Detection in Point Clouds

3D object detection in point clouds is a core component for modern robotics and autonomous driving systems. A key challenge in 3D object detection comes from the inherent sparse nature of point occupancy within the 3D scene. In this paper, we propose Sparse Window Transformer (SWFormer ), a scalable and accurate model for 3D object detection, which can take full advantage of the sparsity of point clouds. Built upon the idea of window-based Transformers, SWFormer converts 3D points into sparse voxels and windows, and then processes these variable-length sparse windows efficiently using a bucketing scheme. In addition to self-attention within each spatial window, our SWFormer also captures cross-window correlation with multi-scale feature fusion and window shifting operations. To further address the unique challenge of detecting 3D objects accurately from sparse features, we propose a new voxel diffusion technique. Experimental results on the Waymo Open Dataset show our SWFormer achieves state-of-the-art 73.36 L2 mAPH on vehicle and pedestrian for 3D object detection on the official test set, outperforming all previous single-stage and two-stage models, while being much more efficient.

cs.CV↗

arXiv2022

LidarNAS: Unifying and Searching Neural Architectures for 3D Point Clouds

Developing neural models that accurately understand objects in 3D point clouds is essential for the success of robotics and autonomous driving. However, arguably due to the higher-dimensional nature of the data (as compared to images), existing neural architectures exhibit a large variety in their designs, including but not limited to the views considered, the format of the neural features, and the neural operations used. Lack of a unified framework and interpretation makes it hard to put these designs in perspective, as well as systematically explore new ones. In this paper, we begin by proposing a unified framework of such, with the key idea being factorizing the neural networks into a series of view transforms and neural layers. We demonstrate that this modular framework can reproduce a variety of existing works while allowing a fair comparison of backbone designs. Then, we show how this framework can easily materialize into a concrete neural architecture search (NAS) space, allowing a principled NAS-for-3D exploration. In performing evolutionary NAS on the 3D object detection task on the Waymo Open Dataset, not only do we outperform the state-of-the-art models, but also report the interesting finding that NAS tends to discover the same macro-level architecture concept for both the vehicle and pedestrian classes.

cs.CV↗

arXiv2022

Bridging the information and dynamics attributes of neural activities

The brain works as a dynamic system to process information. Various challenges remain in understanding the connection between information and dynamics attributes in the brain. The present research pursues exploring how the characteristics of neural information functions are linked to neural dynamics. We attempt to bridge dynamics (e.g., Kolmogorov-Sinai entropy) and information (e.g., mutual information and Fisher information) metrics on the stimulus-triggered stochastic dynamics in neural populations. On the one hand, our unified analysis identifies various essential features of the information-processing-related neural dynamics. We discover spatiotemporal differences in the dynamic randomness and chaotic degrees of neural dynamics during neural information processing. On the other hand, our framework reveals the fundamental role of neural dynamics in shaping neural information processing. The neural dynamics creates an oppositely directed variation of encoding and decoding properties under specific conditions, and it determines the neural representation of stimulus distribution. Overall, our findings demonstrate a potential direction to explain the emergence of neural information processing from neural dynamics and help understand the intrinsic connections between the informational and the physical brain.

q-bio.NC↗

arXiv2022

A unified theory of information transfer and causal relation

Information transfer between coupled stochastic dynamics, measured by transfer entropy and information flow, is suggested as a physical process underlying the causal relation of systems. While information transfer analysis has booming applications in both science and engineering fields, critical mysteries about its foundations remain unsolved. Fundamental yet difficult questions concern how information transfer and causal relation originate, what they depend on, how they differ from each other, and if they are created by a unified and general quantity. These questions essentially determine the validity of causal relation measurement via information transfer. Here we pursue to lay a complete theoretical basis of information transfer and causal relation. Beyond the well-known relations between these concepts that conditionally hold, we demonstrate that information transfer and causal relation universally originate from specific information synergy and redundancy phenomena characterized by high-order mutual information. More importantly, our theory analytically explains the mechanisms for information transfer and causal relation to originate, vanish, and differ from each other. Moreover, our theory naturally defines the effect sizes of information transfer and causal relation based on high-dimensional coupling events. These results may provide a unified view of information, synergy, and causal relation to bridge Pearl's causal inference theory in computer science and information transfer analysis in physics.

cond-mat.stat-mech↗

arXiv2022

Self-organized critical dynamics of RNA virus evolution

RNA virus (e.g., SARS-CoV-2) evolves in a complex manner. Studying RNA virus evolution is vital for understanding molecular evolution and medicine development. Scientists lack, however, general frameworks to characterize the dynamics of RNA virus evolution directly from empirical data and identify potential physical laws. To fill this gap, we present a theory to characterize the RNA virus evolution as a physical system with absorbing states and avalanche behaviors. This approach maps accessible biological data (e.g., phylogenetic tree and infection) to a general stochastic process of RNA virus infection and evolution, enabling researchers to verify potential self-organized criticality underlying RNA virus evolution. We apply our framework to SARS-CoV-2, the virus accounting for the global epidemic of COVID-19. We find that SARS-CoV-2 exhibits scale-invariant avalanches as mean-field theory predictions. The observed scaling relation, universal collapse, and slowly decaying auto-correlation suggest a self-organized critical dynamics of SARS-CoV-2 evolution. Interestingly, the lineages that emerge from critical evolution processes coincidentally match with threatening lineages of SARS-CoV-2 (e.g., the Delta virus). We anticipate our approach to be a general formalism to portray RNA virus evolution and help identify potential virus lineages to be concerned.

q-bio.PE↗

arXiv2022

Information Evolution in Complex Networks

Many biological phenomena or social events critically depend on how information evolves in complex networks. However, a general theory to characterize information evolution is yet absent. Consequently, numerous unknowns remain about the mechanisms underlying information evolution. Among these unknowns, a fundamental problem, being a seeming paradox, lies in the coexistence of local randomness, manifested as the stochastic distortion of information content during individual-individual diffusion, and global regularity, illustrated by specific non-random patterns of information content on the network scale. Here, we attempt to formalize information evolution and explain the coexistence of randomness and regularity in complex networks. Applying network dynamics and information theory, we discover that a certain amount of information, determined by the selectivity of networks to the input information, frequently survives from random distortion. Other information will inevitably experience distortion or dissipation, whose speeds are shaped by the diversity of information selectivity in networks. The discovered laws exist irrespective of noise, but the noise accounts for the intensification. We further demonstrate the ubiquity of our discovered laws by analyzing the emergence of neural tuning properties in the primary visual and medial temporal cortices of animal brains and the emergence of extreme opinions in social networks.

physics.bio-ph↗

arXiv2022

Exact solution of the quantum integrable model associated with the twisted $D^{(2)}_3$ algebra

We generalize the nested off-diagonal Bethe ansatz method to study the quantum chain associated with the twisted $D^{(2)}_3$ algebra (or the $D^{(2)}_3$ model) with either periodic or integrable open boundary conditions. We obtain the intrinsic operator product identities among the fused transfer matrices and find a way to close the recursive fusion relations, which makes it possible to determinate eigenvalues of transfer matrices with an arbitrary anisotropic parameter $η$. Based on them, and the asymptotic behaviors and values at certain points, we construct eigenvalues of transfer matrices in terms of homogeneous $T-Q$ relations for the periodic case and inhomogeneous ones for the open case with some off-diagonal boundary reflections. The associated Bethe ansatz equations are also given. The method and results in this paper can be generalized to the $D^{(2)}_{n+1}$ model and other high rank integrable models.

math-ph↗

arXiv2022

Thermodynamic limit and boundary energy of the spin-1 Heisenberg chain with non-diagonal boundary fields

The thermodynamic limit and boundary energy of the isotropic spin-1 Heisenberg chain with non-diagonal boundary fields are studied. The finite size scaling properties of the inhomogeneous term in the $ T-Q $ relation at the ground state are calculated by the density matrix renormalization group. Based on our findings, the boundary energy of the system in the thermodynamic limit can be obtained from Bethe ansatz equations of a related model with parallel boundary fields. These results can be generalized to the $SU(2)$ symmetric high spin Heisenberg model directly.

hep-th↗

arXiv2021

Scattering matrix of elementary excitations in the antiperiodic XXZ spin chain with η=iπ/3

We study the thermodynamic limit of the antiperiodic XXZ spin chain with the anisotropic parameter $η=\frac{πi}{3}$. We parameterize eigenvalues of the transfer matrix by their zero points instead of Bethe roots. We obtain patterns of the distribution of zero points. Based on them, we calculate the ground state energy and the elementary excitations in the thermodynamic limit. We also obtain the two-body scattering matrix of elementary excitations. Two types of elementary excitations and three types of scattering processes are discussed in detailed.

math-ph↗

arXiv2021

Thermodynamics of Encoding and Encoders

Non-isolated systems have diverse coupling relations with the external environment. These relations generate complex thermodynamics and information transmission between the system and its environment. The framework depicted in the current research attempts to glance at the critical role of the internal orders inside the non-isolated system in shaping the information thermodynamics coupling. We characterize the coupling as a generalized encoding process, where the system acts as an information thermodynamics encoder to encode the external information based on thermodynamics. We formalize the encoding process in the context of the nonequilibrium second law of thermodynamics, revealing an intrinsic difference in information thermodynamics characteristics between information thermodynamics encoders with and without internal correlations. During the information encoding process of an external source $\mathsf{Y}$, specific sub-systems in an encoder $\mathsf{X}$ with internal correlations can exceed the information thermodynamics bound on $\left(\mathsf{X},\mathsf{Y}\right)$ and encode more information than system $\mathsf{X}$ works as a whole. We computationally verify this theoretical finding in an Ising model with a random external field and a neural data set of the human brain during visual perception and recognition. Our analysis demonstrates that the stronger internal correlation inside these systems implies a higher possibility for specific sub-systems to encode more information than the global one. These findings may suggest a new perspective in studying information thermodynamics in diverse physical and biological systems.

cond-mat.stat-mech↗

arXiv2021

Surface energy of the one-dimensional supersymmetric $t-J$ model with general integrable boundary terms in the antiferromagnetic sector

In this paper, we study the surface energy of the one-dimensional supersymmetric $t-J$ model with unparallel boundary magnetic fields, which is a typical $U(1)$-symmetry broken quantum integrable strongly correlated electron system. It is shown that at the ground state, the contribution of inhomogeneous term in the Bethe ansatz solution of eigenvalues of transfer matrix satisfies the finite size scaling law $L^β$ where $β<0$. Based on it, the physical quantities of the system in the thermodynamic limit are calculated. We obtain the patterns of Bethe roots and the analytical expressions of density of states, ground state energy and surface energy. We also find that there exist the stable boundary bound states if the boundary fields satisfy some constraints.

math-ph↗

arXiv2021

Fourier-domain transfer entropy spectrum

We propose the Fourier-domain transfer entropy spectrum, a novel generalization of transfer entropy, as a model-free metric of causality. For arbitrary systems, this approach systematically quantifies the causality among their different system components rather than merely analyze systems as entireties. The generated spectrum offers a rich-information representation of time-varying latent causal relations, efficiently dealing with non-stationary processes and high-dimensional conditions. We demonstrate its validity in the aspects of parameter dependence, statistic significance test, and sensibility. An open-source multi-platform implementation of this metric is developed and computationally applied on neuroscience data sets and diffusively coupled logistic oscillators.

physics.data-an↗

arXiv2021

RSN: Range Sparse Net for Efficient, Accurate LiDAR 3D Object Detection

The detection of 3D objects from LiDAR data is a critical component in most autonomous driving systems. Safe, high speed driving needs larger detection ranges, which are enabled by new LiDARs. These larger detection ranges require more efficient and accurate detection models. Towards this goal, we propose Range Sparse Net (RSN), a simple, efficient, and accurate 3D object detector in order to tackle real time 3D object detection in this extended detection regime. RSN predicts foreground points from range images and applies sparse convolutions on the selected foreground points to detect objects. The lightweight 2D convolutions on dense range images results in significantly fewer selected foreground points, thus enabling the later sparse convolutions in RSN to efficiently operate. Combining features from the range image further enhance detection accuracy. RSN runs at more than 60 frames per second on a 150m x 150m detection region on Waymo Open Dataset (WOD) while being more accurate than previously published detectors. As of 11/2020, RSN is ranked first in the WOD leaderboard based on the APH/LEVEL 1 metrics for LiDAR-based pedestrian and vehicle detection, while being several times faster than alternatives.

cs.CV↗

arXiv2021

To the Point: Efficient 3D Object Detection in the Range Image with Graph Convolution Kernels

3D object detection is vital for many robotics applications. For tasks where a 2D perspective range image exists, we propose to learn a 3D representation directly from this range image view. To this end, we designed a 2D convolutional network architecture that carries the 3D spherical coordinates of each pixel throughout the network. Its layers can consume any arbitrary convolution kernel in place of the default inner product kernel and exploit the underlying local geometry around each pixel. We outline four such kernels: a dense kernel according to the bag-of-words paradigm, and three graph kernels inspired by recent graph neural network advances: the Transformer, the PointNet, and the Edge Convolution. We also explore cross-modality fusion with the camera image, facilitated by operating in the perspective range image view. Our method performs competitively on the Waymo Open Dataset and improves the state-of-the-art AP for pedestrian detection from 69.7% to 75.5%. It is also efficient in that our smallest model, which still outperforms the popular PointPillars in quality, requires 180 times fewer FLOPS and model parameters

cs.CV↗

arXiv2021

Large Scale Interactive Motion Forecasting for Autonomous Driving : The Waymo Open Motion Dataset

As autonomous driving systems mature, motion forecasting has received increasing attention as a critical requirement for planning. Of particular importance are interactive situations such as merges, unprotected turns, etc., where predicting individual object motion is not sufficient. Joint predictions of multiple objects are required for effective route planning. There has been a critical need for high-quality motion data that is rich in both interactions and annotation to develop motion planning models. In this work, we introduce the most diverse interactive motion dataset to our knowledge, and provide specific labels for interacting objects suitable for developing joint prediction models. With over 100,000 scenes, each 20 seconds long at 10 Hz, our new dataset contains more than 570 hours of unique data over 1750 km of roadways. It was collected by mining for interesting interactions between vehicles, pedestrians, and cyclists across six cities within the United States. We use a high-accuracy 3D auto-labeling system to generate high quality 3D bounding boxes for each road agent, and provide corresponding high definition 3D maps for each scene. Furthermore, we introduce a new set of metrics that provides a comprehensive evaluation of both single agent and joint agent interaction motion forecasting models. Finally, we provide strong baseline models for individual-agent prediction and joint-prediction. We hope that this new large-scale interactive motion dataset will provide new opportunities for advancing motion forecasting models.

cs.CV↗