Search

The Arellano-Bond estimator is a fundamental method for dynamic panel data models, widely used in practice. However, the estimator is severely biased when the data's time series dimension $T$ is long due to the large degree of overidentification. We show that weak dependence along the panel's time series dimension naturally implies approximate sparsity of the most informative moment conditions, motivating the following approach to remove the bias: First, apply LASSO to the cross-section data at each time period to construct most informative (and cross-fitted) instruments, using lagged values of suitable covariates. This step relies on approximate sparsity to select the most informative instruments. Second, apply a linear instrumental variable estimator after first differencing the dynamic structural equation using the constructed instruments. Under weak time series dependence, we show the new estimator is consistent and asymptotically normal under much weaker conditions on $T$'s growth than the Arellano-Bond estimator. Our theory covers models with high dimensional covariates, including multiple lags of the dependent variable, common in modern applications. We illustrate our approach by applying it to weekly county-level panel data from the United States to study opening K-12 schools and other mitigation policies' short and long-term effects on COVID-19's spread.

Numerous studies use regression discontinuity design (RDD) for panel data by assuming that the treatment effects are homogeneous across all individuals/groups and pooling the data together. It is unclear how to test for the significance of treatment effects when the treatments vary across individuals/groups and the error terms may exhibit complicated dependence structures. This paper examines the estimation and inference of multiple treatment effects when the errors are not independent and identically distributed, and the treatment effects vary across individuals/groups. We derive a simple analytical expression for approximating the variance-covariance structure of the treatment effect estimators under general dependence conditions and propose two test statistics, one is to test for the overall significance of the treatment effect and the other for the homogeneity of the treatment effects. We find that in the Gaussian approximations to the test statistics, the dependence structures in the data can be safely ignored due to the localized nature of the statistics. This has the important implication that the simulated critical values can be easily obtained. Simulations demonstrate our tests have superb size control and reasonable power performance in finite samples regardless of the presence of strong cross-section dependence or/and weak serial dependence in the data. We apply our tests to two datasets and find significant overall treatment effects in each case.

Video generation necessitates both global coherence and local realism. This work presents a novel non-autoregressive method GLOBER, which first generates global features to obtain comprehensive global guidance and then synthesizes video frames based on the global features to generate coherent videos. Specifically, we propose a video auto-encoder, where a video encoder encodes videos into global features, and a video decoder, built on a diffusion model, decodes the global features and synthesizes video frames in a non-autoregressive manner. To achieve maximum flexibility, our video decoder perceives temporal information through normalized frame indexes, which enables it to synthesize arbitrary sub video clips with predetermined starting and ending frame indexes. Moreover, a novel adversarial loss is introduced to improve the global coherence and local realism between the synthesized video frames. Finally, we employ a diffusion-based video generator to fit the global features outputted by the video encoder for video generation. Extensive experimental results demonstrate the effectiveness and efficiency of our proposed method, and new state-of-the-art results have been achieved on multiple benchmarks.

In this paper, we propose a Vision-Audio-Language Omni-peRception pretraining model (VALOR) for multi-modal understanding and generation. Different from widely-studied vision-language pretraining models, VALOR jointly models relationships of vision, audio and language in an end-to-end manner. It contains three separate encoders for single modality representations, and a decoder for multimodal conditional text generation. We design two pretext tasks to pretrain VALOR model, including Multimodal Grouping Alignment (MGA) and Multimodal Grouping Captioning (MGC). MGA projects vision, language and audio to the same common space, building vision-language, audio-language and audiovisual-language alignment simultaneously. MGC learns how to generate text tokens in conditions of vision, audio or their both. To promote vision-audio-language pretraining research, we construct a large-scale high-quality tri-modality dataset named VALOR-1M, which contains 1M audiable videos with human annotated audiovisual captions. Extensive experiments show that VALOR can learn strong multimodal correlations and be generalized to various downstream tasks (e.g., retrieval, captioning and question answering), with different input modalities (e.g., vision-language, audio-language and audiovisual-language). VALOR achieves new state-of-the-art performances on series of public cross-modality benchmarks. Code and data are available at project page https://casia-iva-group.github.io/projects/VALOR., Comment: Preprint version w/o audio files embeded in PDF. Audio embeded version can be found on project page or github

As a combination of visual and audio signals, video is inherently multi-modal. However, existing video generation methods are primarily intended for the synthesis of visual frames, whereas audio signals in realistic videos are disregarded. In this work, we concentrate on a rarely investigated problem of text guided sounding video generation and propose the Sounding Video Generator (SVG), a unified framework for generating realistic videos along with audio signals. Specifically, we present the SVG-VQGAN to transform visual frames and audio melspectrograms into discrete tokens. SVG-VQGAN applies a novel hybrid contrastive learning method to model inter-modal and intra-modal consistency and improve the quantized representations. A cross-modal attention module is employed to extract associated features of visual frames and audio signals for contrastive learning. Then, a Transformer-based decoder is used to model associations between texts, visual frames, and audio signals at token level for auto-regressive sounding video generation. AudioSetCap, a human annotated text-video-audio paired dataset, is produced for training SVG. Experimental results demonstrate the superiority of our method when compared with existing textto-video generation methods as well as audio generation methods on Kinetics and VAS datasets.

Motion, scene and object are three primary visual components of a video. In particular, objects represent the foreground, scenes represent the background, and motion traces their dynamics. Based on this insight, we propose a two-stage MOtion, Scene and Object decomposition framework (MOSO) for video prediction, consisting of MOSO-VQVAE and MOSO-Transformer. In the first stage, MOSO-VQVAE decomposes a previous video clip into the motion, scene and object components, and represents them as distinct groups of discrete tokens. Then, in the second stage, MOSO-Transformer predicts the object and scene tokens of the subsequent video clip based on the previous tokens and adds dynamic motion at the token level to the generated object and scene tokens. Our framework can be easily extended to unconditional video generation and video frame interpolation tasks. Experimental results demonstrate that our method achieves new state-of-the-art performance on five challenging benchmarks for video prediction and unconditional video generation: BAIR, RoboNet, KTH, KITTI and UCF101. In addition, MOSO can produce realistic videos by combining objects and scenes from different videos., Comment: Accepted by CVPR 2023

We propose an inference method for detecting multiple change points in high-dimensional time series, targeting dense or spatially clustered signals. Our method aggregates moving sum (MOSUM) statistics cross-sectionally by an $\ell^2$-norm and maximizes them over time. We further introduce a novel Two-Way MOSUM, which utilizes spatial-temporal moving regions to search for breaks, with the added advantage of enhancing testing power when breaks occur in only a few groups. The limiting distribution of an $\ell^2$-aggregated statistic is established for testing break existence by extending a high-dimensional Gaussian approximation theorem to spatial-temporal non-stationary processes. Simulation studies exhibit promising performance of our test in detecting non-sparse weak signals. Two applications, analyzing equity returns and COVID-19 cases in the United States, showcase the real-world relevance of our proposed algorithms., Comment: 111 pages, 10 figures

Beta-sorted portfolios -- portfolios comprised of assets with similar covariation to selected risk factors -- are a popular tool in empirical finance to analyze models of (conditional) expected returns. Despite their widespread use, little is known of their statistical properties in contrast to comparable procedures such as two-pass regressions. We formally investigate the properties of beta-sorted portfolio returns by casting the procedure as a two-step nonparametric estimator with a nonparametric first step and a beta-adaptive portfolios construction. Our framework rationalize the well-known estimation algorithm with precise economic and statistical assumptions on the general data generating process and characterize its key features. We study beta-sorted portfolios for both a single cross-section as well as for aggregation over time (e.g., the grand mean), offering conditions that ensure consistency and asymptotic normality along with new uniform inference procedures allowing for uncertainty quantification and testing of various relevant hypotheses in financial applications. We also highlight some limitations of current empirical practices and discuss what inferences can and cannot be drawn from returns to beta-sorted portfolios for either a single cross-section or across the whole sample. Finally, we illustrate the functionality of our new procedures in an empirical application.

This paper develops a novel method for policy choice in a dynamic setting where the available data is a multivariate time series. Building on the statistical treatment choice framework, we propose Time-series Empirical Welfare Maximization (T-EWM) methods to estimate an optimal policy rule for the current period or over multiple periods by maximizing an empirical welfare criterion constructed using nonparametric potential outcome time-series. We characterize conditions under which T-EWM consistently learns a policy choice that is optimal in terms of conditional welfare given the time-series history. We then derive a nonasymptotic upper bound for conditional welfare regret and its minimax lower bound. To illustrate the implementation and uses of T-EWM, we perform simulation studies and apply the method to estimate optimal monetary policy rules from macroeconomic time-series data.

Although face swapping has attracted much attention in recent years, it remains a challenging problem. The existing methods leverage a large number of data samples to explore the intrinsic properties of face swapping without taking into account the semantic information of face images. Moreover, the representation of the identity information tends to be fixed, leading to suboptimal face swapping. In this paper, we present a simple yet efficient method named FaceSwapper, for one-shot face swapping based on Generative Adversarial Networks. Our method consists of a disentangled representation module and a semantic-guided fusion module. The disentangled representation module is composed of an attribute encoder and an identity encoder, which aims to achieve the disentanglement of the identity and the attribute information. The identity encoder is more flexible and the attribute encoder contains more details of the attributes than its competitors. Benefiting from the disentangled representation, FaceSwapper can swap face images progressively. In addition, semantic information is introduced into the semantic-guided fusion module to control the swapped area and model the pose and expression more accurately. The experimental results show that our method achieves state-of-the-art results on benchmark datasets with fewer training samples. Our code is publicly available at https://github.com/liqi-casia/FaceSwapper.

This paper presents a new approach for the estimation and inference of the regression parameters in a panel data model with interactive fixed effects. It relies on the assumption that the factor loadings can be expressed as an unknown smooth function of the time average of covariates plus an idiosyncratic error term. Compared to existing approaches, our estimator has a simple partial least squares form and does neither require iterative procedures nor the previous estimation of factors. We derive its asymptotic properties by finding out that the limiting distribution has a discontinuity, depending on the explanatory power of our basis functions which is expressed by the variance of the error of the factor loadings. As a result, the usual ``plug-in" methods based on estimates of the asymptotic covariance are only valid pointwise and may produce either over- or under-coverage probabilities. We show that uniformly valid inference can be achieved by using the cross-sectional bootstrap. A Monte Carlo study indicates good performance in terms of mean squared error. We apply our methodology to analyze the determinants of growth rates in OECD countries.

We propose a dynamic network quantile regression model to investigate the quantile connectedness using a predetermined network information. We extend the existing network quantile autoregression model of Zhu et al. (2019b) by explicitly allowing the contemporaneous network effects and controlling for the common factors across quantiles. To cope with the endogeneity issue due to simultaneous network spillovers, we adopt the instrumental variable quantile regression (IVQR) estimation and derive the consistency and asymptotic normality of the IVQR estimator using the near epoch dependence property of the network process. Via Monte Carlo simulations, we confirm the satisfactory performance of the IVQR estimator across different quantiles under the different network structures. Finally, we demonstrate the usefulness of our proposed approach with an application to the dataset on the stocks traded in NYSE and NASDAQ in 2016.

Compared with image scene parsing, video scene parsing introduces temporal information, which can effectively improve the consistency and accuracy of prediction. In this paper, we propose a Spatial-Temporal Semantic Consistency method to capture class-exclusive context information. Specifically, we design a spatial-temporal consistency loss to constrain the semantic consistency in spatial and temporal dimensions. In addition, we adopt an pseudo-labeling strategy to enrich the training dataset. We obtain the scores of 59.84% and 58.85% mIoU on development (test part 1) and testing set of VSPW, respectively. And our method wins the 1st place on VSPW challenge at ICCV2021., Comment: 1st Place technical report for "The 1st Video Scene Parsing in the Wild Challenge" at ICCV2021

In this paper, we propose an Omni-perception Pre-Trainer (OPT) for cross-modal understanding and generation, by jointly modeling visual, text and audio resources. OPT is constructed in an encoder-decoder framework, including three single-modal encoders to generate token-based embeddings for each modality, a cross-modal encoder to encode the correlations among the three modalities, and two cross-modal decoders to generate text and image respectively. For the OPT's pre-training, we design a multi-task pretext learning scheme to model multi-modal resources from three different data granularities, \ie, token-, modality-, and sample-level modeling, through which OPT learns to align and translate among different modalities. The pre-training task is carried out on a large amount of image-text-audio triplets from Open Images. Experimental results show that OPT can learn strong image-text-audio multi-modal representations and achieve promising results on a variety of cross-modal understanding and generation tasks.

Face sketch synthesis has made significant progress with the development of deep neural networks in these years. The delicate depiction of sketch portraits facilitates a wide range of applications like digital entertainment and law enforcement. However, accurate and realistic face sketch generation is still a challenging task due to the illumination variations and complex backgrounds in the real scenes. To tackle these challenges, we propose a novel Semantic-Driven Generative Adversarial Network (SDGAN) which embeds global structure-level style injection and local class-level knowledge re-weighting. Specifically, we conduct facial saliency detection on the input face photos to provide overall facial texture structure, which could be used as a global type of prior information. In addition, we exploit face parsing layouts as the semantic-level spatial prior to enforce globally structural style injection in the generator of SDGAN. Furthermore, to enhance the realistic effect of the details, we propose a novel Adaptive Re-weighting Loss (ARLoss) which dedicates to balance the contributions of different semantic classes. Experimentally, our extensive experiments on CUFS and CUFSF datasets show that our proposed algorithm achieves state-of-the-art performance.

We propose employing a debiased-regularized, high-dimensional generalized method of moments (GMM) framework to perform inference on large-scale spatial panel networks. In particular, network structure with a flexible sparse deviation, which can be regarded either as latent or as misspecified from a predetermined adjacency matrix, is estimated using debiased machine learning approach. The theoretical analysis establishes the consistency and asymptotic normality of our proposed estimator, taking into account general temporal and spatial dependency inherent in the data-generating processes. The dimensionality allowance in presence of dependency is discussed. A primary contribution of our study is the development of uniform inference theory that enables hypothesis testing on the parameters of interest, including zero or non-zero elements in the network structure. Additionally, the asymptotic properties for the estimator are derived for both linear and nonlinear moments. Simulations demonstrate superior performance of our proposed approach. Lastly, we apply our methodology to investigate the spatial network effect of stock returns.

Video semantic segmentation requires to utilize the complex temporal relations between frames of the video sequence. Previous works usually exploit accurate optical flow to leverage the temporal relations, which suffer much from heavy computational cost. In this paper, we propose a Temporal Memory Attention Network (TMANet) to adaptively integrate the long-range temporal relations over the video sequence based on the self-attention mechanism without exhaustive optical flow prediction. Specially, we construct a memory using several past frames to store the temporal information of the current frame. We then propose a temporal memory attention module to capture the relation between the current frame and the memory to enhance the representation of the current frame. Our method achieves new state-of-the-art performances on two challenging video semantic segmentation datasets, particularly 80.3% mIoU on Cityscapes and 76.5% mIoU on CamVid with ResNet-50., Comment: ICIP 2021

Image captioning transforms complex visual information into abstract natural language for representation, which can help computers understanding the world quickly. However, due to the complexity of the real environment, it needs to identify key objects and realize their connections, and further generate natural language. The whole process involves a visual understanding module and a language generation module, which brings more challenges to the design of deep neural networks than other tasks. Neural Architecture Search (NAS) has shown its important role in a variety of image recognition tasks. Besides, RNN plays an essential role in the image captioning task. We introduce a AutoCaption method to better design the decoder module of the image captioning where we use the NAS to design the decoder module called AutoRNN automatically. We use the reinforcement learning method based on shared parameters for automatic design the AutoRNN efficiently. The search space of the AutoCaption includes connections between the layers and the operations in layers both, and it can make AutoRNN express more architectures. In particular, RNN is equivalent to a subset of our search space. Experiments on the MSCOCO datasets show that our AutoCaption model can achieve better performance than traditional hand-design methods.

Cryptocurrencies, especially Bitcoin (BTC), which comprise a new digital asset class, have drawn extraordinary worldwide attention. The characteristics of the cryptocurrency/BTC include a high level of speculation, extreme volatility and price discontinuity. We propose a pricing mechanism based on a stochastic volatility with a correlated jump (SVCJ) model and compare it to a flexible co-jump model by Bandi and Ren\`o (2016). The estimation results of both models confirm the impact of jumps and co-jumps on options obtained via simulation and an analysis of the implied volatility curve. We show that a sizeable proportion of price jumps are significantly and contemporaneously anti-correlated with jumps in volatility. Our study comprises pioneering research on pricing BTC options. We show how the proposed pricing mechanism underlines the importance of jumps in cryptocurrency markets.

Estimating spot covariance is an important issue to study, especially with the increasing availability of high-frequency financial data. We study the estimation of spot covariance using a kernel method for high-frequency data. In particular, we consider first the kernel weighted version of realized covariance estimator for the price process governed by a continuous multivariate semimartingale. Next, we extend it to the threshold kernel estimator of the spot covariances when the underlying price process is a discontinuous multivariate semimartingale with finite activity jumps. We derive the asymptotic distribution of the estimators for both fixed and shrinking bandwidth. The estimator in a setting with jumps has the same rate of convergence as the estimator for diffusion processes without jumps. A simulation study examines the finite sample properties of the estimators. In addition, we study an application of the estimator in the context of covariance forecasting. We discover that the forecasting model with our estimator outperforms a benchmark model in the literature.

We develop a uniform test for detecting and dating explosive behavior of a strictly stationary GARCH$(r,s)$ (generalized autoregressive conditional heteroskedasticity) process. Namely, we test the null hypothesis of a globally stable GARCH process with constant parameters against an alternative where there is an 'abnormal' period with changed parameter values. During this period, the change may lead to an explosive behavior of the volatility process. It is assumed that both the magnitude and the timing of the breaks are unknown. We develop a double supreme test for the existence of a break, and then provide an algorithm to identify the period of change. Our theoretical results hold under mild moment assumptions on the innovations of the GARCH process. Technically, the existing properties for the QMLE in the GARCH model need to be reinvestigated to hold uniformly over all possible periods of change. The key results involve a uniform weak Bahadur representation for the estimated parameters, which leads to weak convergence of the test statistic to the supreme of a Gaussian Process. In simulations we show that the test has good size and power for reasonably large time series lengths. We apply the test to Apple asset returns and Bitcoin returns.

In this paper, we study estimation of nonlinear models with cross sectional data using two-step generalized estimating equations (GEE) in the quasi-maximum likelihood estimation (QMLE) framework. In the interest of improving efficiency, we propose a grouping estimator to account for the potential spatial correlation in the underlying innovations. We use a Poisson model and a Negative Binomial II model for count data and a Probit model for binary response data to demonstrate the GEE procedure. Under mild weak dependency assumptions, results on estimation consistency and asymptotic normality are provided. Monte Carlo simulations show efficiency gain of our approach in comparison of different estimation methods for count data and binary response data. Finally we apply the GEE approach to study the determinants of the inflow foreign direct investment (FDI) to China.

We consider the estimation and inference in a system of high-dimensional regression equations allowing for temporal and cross-sectional dependency in covariates and error processes, covering rather general forms of weak temporal dependence. A sequence of regressions with many regressors using LASSO (Least Absolute Shrinkage and Selection Operator) is applied for variable selection purpose, and an overall penalty level is carefully chosen by a block multiplier bootstrap procedure to account for multiplicity of the equations and dependencies in the data. Correspondingly, oracle properties with a jointly selected tuning parameter are derived. We further provide high-quality de-biased simultaneous inference on the many target parameters of the system. We provide bootstrap consistency results of the test procedure, which are based on a general Bahadur representation for the $Z$-estimators with dependent data. Simulations demonstrate good performance of the proposed inference procedure. Finally, we apply the method to quantify spillover effects of textual sentiment indices in a financial market and to test the connectedness among sectors.