Search

In this work, we explore the intersection of sparse coding theory and deep learning to enhance our understanding of feature extraction capabilities in advanced neural network architectures. We begin by introducing a novel class of Deep Sparse Coding (DSC) models and establish a thorough theoretical analysis of their uniqueness and stability properties. By applying iterative algorithms to these DSC models, we derive convergence rates for convolutional neural networks (CNNs) in their ability to extract sparse features. This provides a strong theoretical foundation for the use of CNNs in sparse feature learning tasks. We additionally extend this convergence analysis to more general neural network architectures, including those with diverse activation functions, as well as self-attention and transformer-based models. This broadens the applicability of our findings to a wide range of deep learning methods for deep sparse feature extraction. Inspired by the strong connection between sparse coding and CNNs, we also explore training strategies to encourage neural networks to learn more sparse features. Through numerical experiments, we demonstrate the effectiveness of these approaches, providing valuable insights for the design of efficient and interpretable deep learning models.

In this paper, we explore the approximation theory of functions defined on graphs. Our study builds upon the approximation results derived from the $K$-functional. We establish a theoretical framework to assess the lower bounds of approximation for target functions using Graph Convolutional Networks (GCNs) and examine the over-smoothing phenomenon commonly observed in these networks. Initially, we introduce the concept of a $K$-functional on graphs, establishing its equivalence to the modulus of smoothness. We then analyze a typical type of GCN to demonstrate how the high-frequency energy of the output decays, an indicator of over-smoothing. This analysis provides theoretical insights into the nature of over-smoothing within GCNs. Furthermore, we establish a lower bound for the approximation of target functions by GCNs, which is governed by the modulus of smoothness of these functions. This finding offers a new perspective on the approximation capabilities of GCNs. In our numerical experiments, we analyze several widely applied GCNs and observe the phenomenon of energy decay. These observations corroborate our theoretical results on exponential decay order.

Commonsense reasoning, the ability to make logical assumptions about daily scenes, is one core intelligence of human beings. In this work, we present a novel task and dataset for evaluating the ability of text-to-image generative models to conduct commonsense reasoning, which we call PAINTaboo. Given a description with few visual clues of one object, the goal is to generate images illustrating the object correctly. The dataset was carefully hand-curated and covered diverse object categories to analyze model performance comprehensively. Our investigation of several prevalent text-to-image generative models reveals that these models are not proficient in commonsense reasoning, as anticipated. We trust that PAINTaboo can improve our understanding of the reasoning abilities of text-to-image generative models., Comment: It is an incomplete work

The twig edge states in graphene-like structures are viewed as the fourth states complementary to their zigzag, bearded, and armchair counterparts. In this work, we study a rod-in-plasma system in honeycomb lattice with twig edge truncation under external magnetic fields and lattice scaling and show that twig edge states can exist in different phases of the system, such as quantum Hall phase, quantum spin Hall phase and insulating phase. The twig edge states in the negative permittivity background exhibit robust one-way transmission property immune to backscattering and thus provide a novel avenue for solving the plasma communication blackout problem. Moreover, we demonstrate that corner and edge states can exist within the shrunken structure by modulating the on-site potential of the twig edges. Especially, helical edge states with the unique feature of pseudospin-momentum locking that could be excited by chiral sources are demonstrated at the twig edges. Our results show that the twig edges and interface engineering can bring new opportunities for more flexible manipulation of electromagnetic waves.

The nature of heterophilous graphs is significantly different from that of homophilous graphs, which causes difficulties in early graph neural network models and suggests aggregations beyond the 1-hop neighborhood. In this paper, we develop a new way to implement multi-scale extraction via constructing Haar-type graph framelets with desired properties of permutation equivariance, efficiency, and sparsity, for deep learning tasks on graphs. We further design a graph framelet neural network model PEGFAN (Permutation Equivariant Graph Framelet Augmented Network) based on our constructed graph framelets. The experiments are conducted on a synthetic dataset and 9 benchmark datasets to compare performance with other state-of-the-art models. The result shows that our model can achieve the best performance on certain datasets of heterophilous graphs (including the majority of heterophilous datasets with relatively larger sizes and denser connections) and competitive performance on the remaining.

Plasma photonic crystals designed in this paper are composed of gas discharge tubes to control the flow of electromagnetic waves. The band structures calculated by the finite element method are consistent with the experimental results in that two distinct attenuation peaks appear in the ranges of 1 GHz to 2.5 GHz and 5 GHz to 6 GHz. Electromagnetic parameters of the plasma are extracted by the Nicolson-Ross-Weir method and effective medium theory. The measured electron density is between 1e11 cm-3 and 1e12 cm-3, which verifies the correctness of the parameter used in the simulation, and the collision frequency is near 1.5e10 Hz. As the band structures are corroborated by the measured scattering parameters, we introduce the concept of photonic topological insulator based on the quantum Valley Hall effect into the plasma photonic crystal. A valley-dependent plasma photonic crystal with hexagonal lattice is constructed, and the phase transition of the valley K(K') occurs by breaking the spatial inversion symmetry. Valley-spin locked topological edge states are generated and excited by chiral sources. The frequency of the non-bulk state can be dynamically regulated by the electron density. This concept paves the way for novel, tunable topological edge states. More interestingly, the Dirac cone is broken when the electron density increases to 3.1e12 cm-3, which distinguishes from the methods of applying a magnetic field and changing the symmetry of the point group.

Deep neural networks (DNNs) have garnered significant attention in various fields of science and technology in recent years. Activation functions define how neurons in DNNs process incoming signals for them. They are essential for learning non-linear transformations and for performing diverse computations among successive neuron layers. In the last few years, researchers have investigated the approximation ability of DNNs to explain their power and success. In this paper, we explore the approximation ability of DNNs using a different activation function, called SignReLU. Our theoretical results demonstrate that SignReLU networks outperform rational and ReLU networks in terms of approximation performance. Numerical experiments are conducted comparing SignReLU with the existing activations such as ReLU, Leaky ReLU, and ELU, which illustrate the competitive practical performance of SignReLU.

Deep learning based on deep neural networks has been very successful in many practical applications, but it lacks enough theoretical understanding due to the network architectures and structures. In this paper we establish some analysis for linear feature extraction by a deep multi-channel convolutional neural networks (CNNs), which demonstrates the power of deep learning over traditional linear transformations, like Fourier, wavelets, redundant dictionary coding methods. Moreover, we give an exact construction presenting how linear features extraction can be conducted efficiently with multi-channel CNNs. It can be applied to lower the essential dimension for approximating a high dimensional function. Rates of function approximation by such deep networks implemented with channels and followed by fully-connected layers are investigated as well. Harmonic analysis for factorizing linear features into multi-resolution convolutions plays an essential role in our work. Nevertheless, a dedicate vectorization of matrices is constructed, which bridges 1D CNN and 2D CNN and allows us to have corresponding 2D analysis.

Spherical image processing has been widely applied in many important fields, such as omnidirectional vision for autonomous cars, global climate modelling, and medical imaging. It is non-trivial to extend an algorithm developed for flat images to the spherical ones. In this work, we focus on the challenging task of spherical image inpainting with deep learning-based regularizer. Instead of a naive application of existing models for planar images, we employ a fast directional spherical Haar framelet transform and develop a novel optimization framework based on a sparsity assumption of the framelet transform. Furthermore, by employing progressive encoder-decoder architecture, a new and better-performed deep CNN denoiser is carefully designed and works as an implicit regularizer. Finally, we use a plug-and-play method to handle the proposed optimization model, which can be implemented efficiently by training the CNN denoiser prior. Numerical experiments are conducted and show that the proposed algorithms can greatly recover damaged spherical images and achieve the best performance over purely using deep learning denoiser and plug-and-play model.

In this paper, we develop a general theoretical framework for constructing Haar-type tight framelets on any compact set with a hierarchical partition. In particular, we construct a novel area-regular hierarchical partition on the 2-sphere and establish its corresponding spherical Haar tight framelets with directionality. We conclude by evaluating and illustrating the effectiveness of our area-regular spherical Haar tight framelets in several denoising experiments. Furthermore, we propose a convolutional neural network (CNN) model for spherical signal denoising which employs the fast framelet decomposition and reconstruction algorithms. Experiment results show that our proposed CNN model outperforms threshold methods, and processes strong generalization and robustness properties.

Transition metal dichalcogenides (TMDs) are known to support complex excitonic states. Revealing the differences in relaxation dynamics among different excitonic species and elucidating the transition dynamics between them may provide important guidelines for designing TMD-based excitonic devices. Combining photoluminescence (PL) and reflectance contrast measurements with ultrafast pump-probe spectroscopy under cryogenic temperatures, we herein study the relaxation dynamics of neutral and charged excitons in a back-gate-controlled monolayer device. Pump-probe results reveal quite different relaxation dynamics of excitonic states under different interfacial conditions: while neutral excitons experience much longer lifetime than trions in monolayer WS2, the opposite is true in the WS2/h-BN heterostructure. It is found that the insertion of h-BN layer between the TMD monolayer and the substrate has a great influence on the lifetimes of different excitonic states. The h-BN flakes can not only screen the effects of impurities and defects at the interface, but also help establish a non-radiative transition from neutral excitons to trions to be the dominant relaxation pathway, under cryogenic temperature. Our findings highlight the important role interface may play in governing the transient properties of carriers in 2D semiconductors, and may also have implications for designing light-emitting and photo-detecting devices based on TMDs., Comment: 12 pages

The sub-bandgap absorption and ultrafast relaxation in a GaAs/ErAs/GaAs heterostructure are reported. The infrared absorption and 1550 nm-excited ultrafast photo-response are studied by Fourier transform infrared (FTIR) spectrometry and time-domain pump-probe technique. The two absorption peaks located at 2.0 um (0.62 eV) and 2.7 um (0.45 eV) are originated from the ErAs/GaAs interfacial Schottky states and sub-bandgap transition within GaAs, respectively. The photo-induced carrier lifetime, excited using 1550 nm light, is measured to be as low as 190 fs for the GaAs/ErAs/GaAs heterostructure, making it a promising material for 1550-nm-technology-compatible, high critical-breakdown-field THz devices. The relaxation mechanism is proposed and the functionality of ErAs is revealed.