Image animation generates a video of a source image following the motion of a driving video. State-of-the-art self-supervised image animation approaches warp the source image according to the motion of the driving video and recover the warping artifacts by inpainting. These approaches mostly use vanilla convolution for inpainting, and vanilla convolution does not distinguish between valid and invalid pixels. As a result, visual artifacts are still noticeable after inpainting. CutMix is a state-of-the-art regularization strategy that cuts and mixes patches of images and is widely studied in different computer vision tasks. Among the remaining computer vision tasks, warp-based image animation is one of the fields that the effects of CutMix have yet to be studied. This paper first presents a preliminary study on the effects of CutMix on warp-based image animation. We observed in our study that CutMix helps improve only pixel values, but disturbs the spatial relationships between pixels. Based on such observation, we propose PriorityCut, a novel augmentation approach that uses the top-k percent occluded pixels of the foreground to regularize warp-based image animation. By leveraging the domain knowledge in warp-based image animation, PriorityCut significantly reduces the warping artifacts in state-of-the-art warp-based image animation models on diverse datasets.
The sharpest known high probability generalization bounds for uniformly stable algorithms (Feldman, Vondr\'{a}k, 2018, 2019), (Bousquet, Klochkov, Zhivotovskiy, 2020) contain a generally inevitable sampling error term of order $\Theta(1/\sqrt{n})$. When applied to excess risk bounds, this leads to suboptimal results in several standard stochastic convex optimization problems. We show that if the so-called Bernstein condition is satisfied, the term $\Theta(1/\sqrt{n})$ can be avoided, and high probability excess risk bounds of order up to $O(1/n)$ are possible via uniform stability. Using this result, we show a high probability excess risk bound with the rate $O(\log n/n)$ for strongly convex and Lipschitz losses valid for \emph{any} empirical risk minimization method. This resolves a question of Shalev-Shwartz, Shamir, Srebro, and Sridharan (2009). We discuss how $O(\log n/n)$ high probability excess risk bounds are possible for projected gradient descent in the case of strongly convex and Lipschitz losses without the usual smoothness assumption.
We present a unified technique for sequential estimation of convex divergences between distributions, including integral probability metrics like the kernel maximum mean discrepancy, $\varphi$-divergences like the Kullback-Leibler divergence, and optimal transport costs, such as powers of Wasserstein distances. This is achieved by observing that empirical convex divergences are (partially ordered) reverse submartingales with respect to the exchangeable filtration, coupled with maximal inequalities for such processes. These techniques appear to be complementary and powerful additions to the existing literature on both confidence sequences and convex divergences. We construct an offline-to-sequential device that converts a wide array of existing offline concentration inequalities into time-uniform confidence sequences that can be continuously monitored, providing valid tests or confidence intervals at arbitrary stopping times. The resulting sequential bounds pay only an iterated logarithmic price over the corresponding fixed-time bounds, retaining the same dependence on problem parameters (like dimension or alphabet size if applicable). These results are also applicable to more general convex functionals -- like the negative differential entropy, suprema of empirical processes, and V-Statistics -- and to more general processes satisfying a key leave-one-out property.
Existing survival analysis techniques heavily rely on strong modelling assumptions and are, therefore, prone to model misspecification errors. In this paper, we develop an inferential method based on ideas from conformal prediction, which can wrap around any survival prediction algorithm to produce calibrated, covariate-dependent lower predictive bounds on survival times. In the Type I right-censoring setting, when the censoring times are completely exogenous, the lower predictive bounds have guaranteed coverage in finite samples without any assumptions other than that of operating on independent and identically distributed data points. Under a more general conditionally independent censoring assumption, the bounds satisfy a doubly robust property which states the following: marginal coverage is approximately guaranteed if either the censoring mechanism or the conditional survival function is estimated well. Further, we demonstrate that the lower predictive bounds remain valid and informative for other types of censoring. The validity and efficiency of our procedure are demonstrated on synthetic data and real COVID-19 data from the UK Biobank.
The R package DoubleML implements the double/debiased machine learning framework of Chernozhukov et al. (2018). It provides functionalities to estimate parameters in causal models based on machine learning methods. The double machine learning framework consist of three key ingredients: Neyman orthogonality, high-quality machine learning estimation and sample splitting. Estimation of nuisance components can be performed by various state-of-the-art machine learning methods that are available in the mlr3 ecosystem. DoubleML makes it possible to perform inference in a variety of causal models, including partially linear and interactive regression models and their extensions to instrumental variable estimation. The object-oriented implementation of DoubleML enables a high flexibility for the model specification and makes it easily extendable. This paper serves as an introduction to the double machine learning framework and the R package DoubleML. In reproducible code examples with simulated and real data sets, we demonstrate how DoubleML users can perform valid inference based on machine learning methods.
Natural language segmentation (NLS), or text segmentation, refers to the process of dividing written text into meaningful units. Sentence segmentation, a subfield of text segmentation, is the problem of dividing a string of natural language text into its component sentences. Current methods of sentence segmentation are often either hardcoded-they require manual implementation of fixed grammar and segmentation rules-or require extensive training on labeled corpora and are not explainable-they are "black box" algorithms that cannot be understood by humans. In this paper, we present a novel explainable sentence segmentation method capable of separating bodies of text into grammatically valid sentences solely based on the grammatical relationships between individual words or tokens. The proposed NLS architecture can both automate the input query parsing and semantic query execution components of voice-activated question answering and information retrieval systems as well as enable automatic summarization, entity extraction, sentiment identification, and a variety of other natural language processing (NLP) algorithms that operate at the sentential level.
Under mild conditions on the noise level of the measurements, rotation averaging satisfies strong duality, which enables global solutions to be obtained via semidefinite programming (SDP) relaxation. However, generic solvers for SDP are rather slow in practice, even on rotation averaging instances of moderate size, thus developing specialised algorithms is vital. In this paper, we present a fast algorithm that achieves global optimality called rotation coordinate descent (RCD). Unlike block coordinate descent (BCD) which solves SDP by updating the semidefinite matrix in a row-by-row fashion, RCD directly maintains and updates all valid rotations throughout the iterations. This obviates the need to store a large dense semidefinite matrix. We mathematically prove the convergence of our algorithm and empirically show its superior efficiency over state-of-the-art global methods on a variety of problem configurations. Maintaining valid rotations also facilitates incorporating local optimisation routines for further speed-ups. Moreover, our algorithm is simple to implement; see supplementary material for a demonstration program.
In this work, we study the image transformation problem, which targets at learning the underlying transformations (e.g., the transition of seasons) from a collection of unlabeled images. However, there could be countless of transformations in the real world, making such a task incredibly challenging, especially under the unsupervised setting. To tackle this obstacle, we propose a novel learning framework built on generative adversarial networks (GANs), where the discriminator and the generator share a transformation space. After the model gets fully optimized, any two points within the shared space are expected to define a valid transformation. In this way, at the inference stage, we manage to adequately extract the variation factor between a customizable image pair by projecting both images onto the transformation space. The resulting transformation vector can further guide the image synthesis, facilitating image editing with continuous semantic change (e.g., altering summer to winter with fall as the intermediate step). Noticeably, the learned transformation space supports not only transferring image styles (e.g., changing day to night), but also manipulating image contents (e.g., adding clouds in the sky). In addition, we make in-depth analysis on the properties of the transformation space to help understand how various transformations are organized. Project page is at https://genforce.github.io/trgan/.
Confidence intervals based on the central limit theorem (CLT) are a cornerstone of classical statistics. Despite being only asymptotically valid, they are ubiquitous because they permit statistical inference under weak assumptions and can often be applied to problems even when nonasymptotic inference is impossible. This paper introduces time-uniform analogues of such asymptotic confidence intervals, adding to the literature on confidence sequences (CS) -- sequences of confidence intervals that are uniformly valid over time -- which provide valid inference at arbitrary stopping times and incur no penalties for "peeking" at the data, unlike classical confidence intervals which require the sample size to be fixed in advance. Existing CSs in the literature are nonasymptotic, enjoying finite-sample guarantees but not the aforementioned broad applicability of asymptotic confidence intervals. This work provides a definition for "asymptotic CSs" and a general recipe for deriving them. Asymptotic CSs forgo nonasymptotic validity for CLT-like versatility and (asymptotic) time-uniform guarantees. While the CLT approximates the distribution of a sample average by that of a Gaussian for a fixed sample size, we use strong invariance principles (stemming from the seminal 1960s work of Strassen) to uniformly approximate the entire sample average process by an implicit Gaussian process. As an illustration, we derive asymptotic CSs for the average treatment effect in observational studies (for which nonasymptotic bounds are essentially impossible to derive even in the fixed-time regime) as well as randomized experiments, enabling causal inference in sequential environments.
This paper reexamines the seminal Lagrange multiplier test for cross-section independence in a large panel model where both the number of cross-sectional units n and the number of time series observations T can be large. The first contribution of the paper is an enlargement of the test with two extensions: firstly the new asymptotic normality is derived in a simultaneous limiting scheme where the two dimensions (n, T) tend to infinity with comparable magnitudes; second, the result is valid for general error distribution (not necessarily normal). The second contribution of the paper is a new test statistic based on the sum of the fourth powers of cross-section correlations from OLS residuals, instead of their squares used in the Lagrange multiplier statistic. This new test is generally more powerful, and the improvement is particularly visible against alternatives with weak or sparse cross-section dependence. Both simulation study and real data analysis are proposed to demonstrate the advantages of the enlarged Lagrange multiplier test and the power enhanced test in comparison with the existing procedures.
An important issue when using Machine Learning algorithms in recent research is the lack of interpretability. Although these algorithms provide accurate point predictions for various learning problems, uncertainty estimates connected with point predictions are rather sparse. A contribution to this gap for the Random Forest Regression Learner is presented here. Based on its Out-of-Bag procedure, several parametric and non-parametric prediction intervals are provided for Random Forest point predictions and theoretical guarantees for its correct coverage probability is delivered. In a second part, a thorough investigation through Monte-Carlo simulation is conducted evaluating the performance of the proposed methods from three aspects: (i) Analyzing the correct coverage rate of the proposed prediction intervals, (ii) Inspecting interval width and (iii) Verifying the competitiveness of the proposed intervals with existing methods. The simulation yields that the proposed prediction intervals are robust towards non-normal residual distributions and are competitive by providing correct coverage rates and comparably narrow interval lengths, even for comparably small samples.
The boundary of tumors (hepatocellular carcinoma, or HCC) contains rich semantics: capsular invasion, visibility, smoothness, folding and protuberance, etc. Capsular invasion on tumor boundary has proven to be clinically correlated with the prognostic indicator, microvascular invasion (MVI). Investigating tumor boundary semantics has tremendous clinical values. In this paper, we propose the first and novel computational framework that disentangles the task into two components: spatial vertex localization and sequential semantic classification. (1) A HCC tumor segmentor is built for tumor mask boundary extraction, followed by polar transform representing the boundary with radius and angle. Vertex generator is used to produce fixed-length boundary vertices where vertex features are sampled on the corresponding spatial locations. (2) The sampled deep vertex features with positional embedding are mapped into a sequential space and decoded by a multilayer perceptron (MLP) for semantic classification. Extensive experiments on tumor capsule semantics demonstrate the effectiveness of our framework. Mining the correlation between the boundary semantics and MVI status proves the feasibility to integrate this boundary semantics as a valid HCC prognostic biomarker.
One of the central puzzles in modern machine learning is the ability of heavily overparametrized models to generalize well. Although the low-dimensional structure of typical datasets is key to this behavior, most theoretical studies of overparametrization focus on isotropic inputs. In this work, we instead consider an analytically tractable model of structured data, where the input covariance is built from independent blocks allowing us to tune the saliency of low-dimensional structures and their alignment with respect to the target function. Using methods from statistical physics, we derive a precise asymptotic expression for the train and test error achieved by random feature models trained to classify such data, which is valid for any convex loss function. We study in detail how the data structure affects the double descent curve, and show that in the over-parametrized regime, its impact is greater for logistic loss than for mean-squared loss: the easier the task, the wider the gap in performance at the advantage of the logistic loss. Our insights are confirmed by numerical experiments on MNIST and CIFAR10.
We study the potential scenarios from a Susceptible-Infected-Recovered-Asymptomatic-Symptomatic-Dead (SIRASD) model. As a novelty, we consider populations that differ in their degree of compliance with social distancing policies following socioeconomic attributes that are observed in emerging and developing countries. Considering epidemiological parameters estimated from data of the propagation of SARS-CoV-2 in Brazil -- where there is a significant stake of the population making their living in the informal economy and thus prone to not follow self-isolation -- we assert that if the confinement measures are lifted too soon, namely as much as one week of consecutive declining numbers of new cases, it is very likely the appearance of a second peak. Our approach should be valid for any country where the number of people involved in the informal economy is a large proportion of the total labor force. In summary, our results point out the crucial relevance of target policies for supporting people in the informal economy to properly comply with preventive measures during the pandemic.
Several approaches were proposed to describe the geomorphology of drainage networks and the abiotic/biotic factors determining their morphology. There is an intrinsic complexity of the explicit qualification of the morphological variations in response to various types of control factors and the difficulty of expressing the cause-effect links. Traditional methods of drainage network classification are based on the manual extraction of key characteristics, then applied as pattern recognition schemes. These approaches, however, have low predictive and uniform ability. We present a different approach, based on the data-driven supervised learning by images, extended also to extraterrestrial cases. With deep learning models, the extraction and classification phase is integrated within a more objective, analytical, and automatic framework. Despite the initial difficulties, due to the small number of training images available, and the similarity between the different shapes of the drainage samples, we obtained successful results, concluding that deep learning is a valid way for data exploration in geomorphology and related fields.
Many social sciences such as psychology and economics try to learn the behaviour of complex agents such as humans, organisations and countries. The current statistical methods used for learning this behaviour try to infer generally valid behaviour, but can only learn from one type of study at a time. Furthermore, only data from carefully designed studies can be used, as the phenomenon of interest has to be isolated and confounding factors accounted for. These restrictions limit the robustness and accuracy of insights that can be gained from social/economic systems. Here we present the online platform TreeOfKnowledge which implements a new methodology specifically designed for learning complex behaviours from complex systems: agent-based behaviour learning. With agent-based behaviour learning it is possible to gain more accurate and robust insights as it does not have the restriction of conventional statistics. It learns agent behaviour from many heterogenous datasets and can learn from these datasets even if the phenomenon of interest is not directly observed, but appears deep within complex systems. This new methodology shows how the internet and advances in computational power allow for more accurate and powerful mathematical models.
We investigate upper bounds on the length of cost optimal plans that are valid for problems with 0-cost actions. We employ these upper bounds as horizons for a SAT-based encoding of planning with costs. Given an initial upper bound on the cost of the optimal plan, we experimentally show that this SAT-based approach is able to compute plans with better costs, and in many cases it can match the optimal cost. Also, in multiple instances, the approach is successful in proving that a certain cost is the optimal plan cost.
We investigate the asymptotic risk of a general class of overparameterized likelihood models, including deep models. The recent empirical success of large-scale models has motivated several theoretical studies to investigate a scenario wherein both the number of samples, $n$, and parameters, $p$, diverge to infinity and derive an asymptotic risk at the limit. However, these theorems are only valid for linear-in-feature models, such as generalized linear regression, kernel regression, and shallow neural networks. Hence, it is difficult to investigate a wider class of nonlinear models, including deep neural networks with three or more layers. In this study, we consider a likelihood maximization problem without the model constraints and analyze the upper bound of an asymptotic risk of an estimator with penalization. Technically, we combine a property of the Fisher information matrix with an extended Marchenko-Pastur law and associate the combination with empirical process techniques. The derived bound is general, as it describes both the double descent and the regularized risk curves, depending on the penalization. Our results are valid without the linear-in-feature constraints on models and allow us to derive the general spectral distributions of a Fisher information matrix from the likelihood. We demonstrate that several explicit models, such as parallel deep neural networks, ensemble learning, and residual networks, are in agreement with our theory. This result indicates that even large and deep models have a small asymptotic risk if they exhibit a specific structure, such as divisibility. To verify this finding, we conduct a real-data experiment with parallel deep neural networks. Our results expand the applicability of the asymptotic risk analysis, and may also contribute to the understanding and application of deep learning.
The reliability of software that has a Deep Neural Network (DNN) as a component is urgently important today given the increasing number of critical applications being deployed with DNNs. The need for reliability raises a need for rigorous testing of the safety and trustworthiness of these systems. In the last few years, there have been a number of research efforts focused on testing DNNs. However the test generation techniques proposed so far lack a check to determine whether the test inputs they are generating are valid, and thus invalid inputs are produced. To illustrate this situation, we explored three recent DNN testing techniques. Using deep generative model based input validation, we show that all the three techniques generate significant number of invalid test inputs. We further analyzed the test coverage achieved by the test inputs generated by the DNN testing techniques and showed how invalid test inputs can falsely inflate test coverage metrics. To overcome the inclusion of invalid inputs in testing, we propose a technique to incorporate the valid input space of the DNN model under test in the test generation process. Our technique uses a deep generative model-based algorithm to generate only valid inputs. Results of our empirical studies show that our technique is effective in eliminating invalid tests and boosting the number of valid test inputs generated.
As facial interaction systems are prevalently deployed, security and reliability of these systems become a critical issue, with substantial research efforts devoted. Among them, face anti-spoofing emerges as an important area, whose objective is to identify whether a presented face is live or spoof. Recently, a large-scale face anti-spoofing dataset, CelebA-Spoof which comprised of 625,537 pictures of 10,177 subjects has been released. It is the largest face anti-spoofing dataset in terms of the numbers of the data and the subjects. This paper reports methods and results in the CelebA-Spoof Challenge 2020 on Face AntiSpoofing which employs the CelebA-Spoof dataset. The model evaluation is conducted online on the hidden test set. A total of 134 participants registered for the competition, and 19 teams made valid submissions. We will analyze the top ranked solutions and present some discussion on future work directions.