Although recent advances in machine learning have shown its success to learn from independent and identically distributed (IID) data, it is vulnerable to out-of-distribution (OOD) data in an open world. Domain generalization (DG) deals with such an issue and it aims to learn a model from multiple source domains that can be generalized to unseen target domains. Existing studies on DG have largely focused on stationary settings with homogeneous source domains. However, in many applications, domains may evolve along a specific direction (e.g., time, space). Without accounting for such non-stationary patterns, models trained with existing methods may fail to generalize on OOD data. In this paper, we study domain generalization in non-stationary environment. We first examine the impact of environmental non-stationarity on model performance and establish the theoretical upper bounds for the model error at target domains. Then, we propose a novel algorithm based on adaptive invariant representation learning, which leverages the non-stationary pattern to train a model that attains good performance on target domains. Experiments on both synthetic and real data validate the proposed algorithm.
Offline-to-online reinforcement learning has recently been shown effective in reducing the online sample complexity by first training from offline collected data. However, this additional data source may also invite new poisoning attacks that target offline training. In this work, we reveal such vulnerabilities in critic-regularized offline RL by proposing a novel data poisoning attack method, which is stealthy in the sense that the performance during the offline training remains intact, but the online fine-tuning stage will suffer a significant performance drop. Our method leverages the techniques from bi-level optimization to promote the over-estimation/distribution shift under offline-to-online reinforcement learning. Experiments on four environments confirm the satisfaction of the new stealthiness requirement, and can be effective in attacking with only a small budget and without having white-box access to the victim model.
Propensity scores are commonly used to estimate treatment effects from observational data. We argue that the probabilistic output of a learned propensity score model should be calibrated-i.e., a predictive treatment probability of 90% should correspond to 90% individuals being assigned the treatment group-and we propose simple recalibration techniques to ensure this property. We prove that calibration is a necessary condition for unbiased treatment effect estimation when using popular inverse propensity weighted and doubly robust estimators. We derive error bounds on causal effect estimates that directly relate to the quality of uncertainties provided by the probabilistic propensity score model and show that calibration strictly improves this error bound while also avoiding extreme propensity weights. We demonstrate improved causal effect estimation with calibrated propensity scores in several tasks including high-dimensional image covariates and genome-wide association studies (GWASs). Calibrated propensity scores improve the speed of GWAS analysis by more than two-fold by enabling the use of simpler models that are faster to train.
Uncertainty estimation in deep models is essential in many real-world applications and has benefited from developments over the last several years. Recent evidence [Farquhar et al., 2020] suggests that existing solutions dependent on simple Gaussian formulations may not be sufficient. However, moving to other distributions necessitates Monte Carlo (MC) sampling to estimate quantities such as the KL divergence: it could be expensive and scales poorly as the dimensions of both the input data and the model grow. This is directly related to the structure of the computation graph, which can grow linearly as a function of the number of MC samples needed. Here, we construct a framework to describe these computation graphs, and identify probability families where the graph size can be independent or only weakly dependent on the number of MC samples. These families correspond directly to large classes of distributions. Empirically, we can run a much larger number of iterations for MC approximations for larger architectures used in computer vision with gains in performance measured in confident accuracy, stability of training, memory and training time.
Panel data involving longitudinal measurements of the same set of participants taken over multiple time points is common in studies to understand childhood development and disease modeling. Deep hybrid models that marry the predictive power of neural networks with physical simulators such as differential equations, are starting to drive advances in such applications. The task of modeling not just the observations but the hidden dynamics that are captured by the measurements poses interesting statistical/computational questions. We propose a probabilistic model called ME-NODE to incorporate (fixed + random) mixed effects for analyzing such panel data. We show that our model can be derived using smooth approximations of SDEs provided by the Wong-Zakai theorem. We then derive Evidence Based Lower Bounds for ME-NODE, and develop (efficient) training algorithms using MC based sampling methods and numerical ODE solvers. We demonstrate ME-NODE's utility on tasks spanning the spectrum from simulations and toy data to real longitudinal 3D imaging data from an Alzheimer's disease (AD) study, and study its performance in terms of accuracy of reconstruction for interpolation, uncertainty estimates and personalized prediction.
A foundational tool for making causal inferences is the emulation of randomized control trials via variable interventions. This approach has been applied to a wide variety of contexts, from health to economics [4, 7]. Variable interventions have long been studied in independent and identically distributed (iid) data contexts, but recently non-iid settings, such as networks with interacting agents [9, 20, 32] have attracted interest. In this paper, we propose a type of structural intervention [14] relevant in network contexts: the network intervention. Rather than estimating the effect of changing variables, we consider changes to social network structure resulting from creation or severance of ties between agents. We define the individual participant and average bystander effects for these interventions and describe identification criteria. We then prove a series of theoretical results that show existing identification theory obtains minimally KL-divergent distributions corresponding to network interventions. Finally, we demonstrate estimation of effects of network interventions via a simulation study.
Scientific models describe natural phenomena at different levels of abstraction. Abstract descriptions can provide the basis for interventions on the system and explanation of observed phenomena at a level of granularity that is coarser than the most fundamental account of the system. Beckers and Halpern (2019), building on work of Rubenstein et al. (2017), developed an account of abstraction for causal models that is exact. Here we extend this account to the more realistic case where an abstract causal model offers only an approximation of the underlying system. We show how the resulting account handles the discrepancy that can arise between low- and high-level causal models of the same system, and in the process provide an account of how one causal model approximates another, a topic of independent interest. Finally, we extend the account of approximate abstractions to probabilistic causal models, indicating how and where uncertainty can enter into an approximate abstraction.
Domain generalization (DG) aims to incorporate knowledge from multiple source domains into a single model that could generalize well on unseen target domains. This problem is ubiquitous in practice since the distributions of the target data may rarely be identical to those of the source data. In this paper, we propose Multidomain Discriminant Analysis (MDA) to address DG of classification tasks in general situations. MDA learns a domain-invariant feature transformation that aims to achieve appealing properties, including a minimal divergence among domains within each class, a maximal separability among classes, and overall maximal compactness of all classes. Furthermore, we provide the bounds on excess risk and generalization error by learning theory analysis. Comprehensive experiments on synthetic and real benchmark datasets demonstrate the effectiveness of MDA.
Missing data is a pervasive problem in data analyses, resulting in datasets that contain censored realizations of a target distribution. Many approaches to inference on the target distribution using censored observed data, rely on missing data models represented as a factorization with respect to a directed acyclic graph. In this paper we consider the identifiability of the target distribution within this class of models, and show that the most general identification strategies proposed so far retain a significant gap in that they fail to identify a wide class of identifiable distributions. To address this gap, we propose a new algorithm that significantly generalizes the types of manipulations used in the ID algorithm [14, 16], developed in the context of causal inference, in order to obtain identification.
Classical causal and statistical inference methods typically assume the observed data consists of independent realizations. However, in many applications this assumption is inappropriate due to a network of dependences between units in the data. Methods for estimating causal effects have been developed in the setting where the structure of dependence between units is known exactly [10, 36, 20], but in practice there is often substantial uncertainty about the precise network structure. This is true, for example, in trial data drawn from vulnerable communities where social ties are difficult to query directly. In this paper we combine techniques from the structure learning and interference literatures in causal inference, proposing a general method for estimating causal effects under data dependence when the structure of this dependence is not known a priori. We demonstrate the utility of our method on synthetic datasets which exhibit network dependence.
There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are benefits to a system that is not only accurate but also has a sense for when it is not. Existing proposals center around Bayesian interpretations of modern deep architectures - these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We show that while uncertainty is quite useful by itself in computer vision and machine learning, we also demonstrate that it can play a key role in enabling statistical analysis with deep networks in neuroimaging studies with normative modeling methods. To our knowledge, this is the first result describing sampling free uncertainty estimation for powerful sequential models such as GRUs.
Unlike classical causal inference, where the goal is to estimate average causal effects within a population, in settings such as personalized medicine, the goal is to map a unit's characteristics to a treatment tailored to maximize the expected outcome for that unit. Obtaining high-quality mappings of this type is the goal of the dynamic treatment regime literature. In healthcare settings, optimizing policies with respect to a particular causal pathway is often of interest as well. In the context of average treatment effects, estimation of effects associated with causal pathways is considered in the mediation analysis literature. In this paper, we combine mediation analysis and dynamic treatment regime ideas and consider how unit characteristics may be used to tailor a treatment strategy that maximizes an effect along specified sets of causal pathways. In particular, we define counterfactual responses to such policies, give a general identification algorithm for these counterfactuals, and prove completeness of the algorithm for unrestricted policies. A corollary of our results is that the identification algorithm for responses to policies given in [16] is complete for arbitrary policies.
The conditional independence structure induced on the observed marginal distribution by a hidden variable directed acyclic graph (DAG) may be represented by a graphical model represented by mixed graphs called maximal ancestral graphs (MAGs). This model has a number of desirable properties, in particular the set of Gaussian distributions can be parameterized by viewing the graph as a path diagram. Models represented by MAGs have been used for causal discovery [22], and identification theory for causal effects [28]. In addition to ordinary conditional independence constraints, hidden variable DAGs also induce generalized independence constraints. These constraints form the nested Markov property [20]. We first show that acyclic linear SEMs obey this property. Further we show that a natural parameterization for all Gaussian distributions obeying the nested Markov property arises from a generalization of maximal ancestral graphs that we call maximal arid graphs (MArG). We show that every nested Markov model can be associated with a MArG; viewed as a path diagram this MArG parametrizes the Gaussian nested Markov model. This leads directly to methods for ML fitting and computing BIC scores for Gaussian nested models.
The goal of personalized decision making is to map a unit's characteristics to an action tailored to maximize the expected outcome for that unit. Obtaining high-quality mappings of this type is the goal of the dynamic regime literature. In healthcare settings, optimizing policies with respect to a particular causal pathway may be of interest as well. For example, we may wish to maximize the chemical effect of a drug given data from an observational study where the chemical effect of the drug on the outcome is entangled with the indirect effect mediated by differential adherence. In such cases, we may wish to optimize the direct effect of a drug, while keeping the indirect effect to that of some reference treatment. [15] shows how to combine mediation analysis and dynamic treatment regime ideas to defines policies associated with causal pathways and counterfactual responses to these policies. In this paper, we derive a variety of methods for learning high quality policies of this type from data, in a causal model corresponding to a longitudinal setting of practical importance. We illustrate our methods via a dataset of HIV patients undergoing therapy, gathered in the Nigerian PEPFAR program.
We study the $L_1$-regularized maximum likelihood estimator/estimation (MLE) problem for discrete Markov random fields (MRFs), where efficient and scalable learning requires both sparse regularization and approximate inference. To address these challenges, we consider a stochastic learning framework called stochastic proximal gradient (SPG; Honorio 2012a, Atchade et al. 2014,Miasojedow and Rejchel 2016). SPG is an inexact proximal gradient algorithm [Schmidtet al., 2011], whose inexactness stems from the stochastic oracle (Gibbs sampling) for gradient approximation - exact gradient evaluation is infeasible in general due to the NP-hard inference problem for discrete MRFs [Koller and Friedman, 2009]. Theoretically, we provide novel verifiable bounds to inspect and control the quality of gradient approximation. Empirically, we propose the tighten asymptotically (TAY) learning strategy based on the verifiable bounds to boost the performance of SPG.
Discovering causal structure of a dynamical system from observed time series is a traditional and important problem. In many practical applications, observed data are obtained by applying subsampling or temporally aggregation to the original causal processes, making it difficult to discover the underlying causal relations. Subsampling refers to the procedure that for every k consecutive observations, one is kept, the rest being skipped, and recently some advances have been made in causal discovery from such data. With temporal aggregation, the local averages or sums of k consecutive, non-overlapping observations in the causal process are computed as new observations, and causal discovery from such data is even harder. In this paper, we investigate how to recover causal relations at the original causal frequency from temporally aggregated data when k is known. Assuming the time series at the causal frequency follows a vector autoregressive (VAR) model, we show that the causal structure at the causal frequency is identifiable from aggregated time series if the noise terms are independent and non-Gaussian and some other technical conditions hold. We then present an estimation method based on non-Gaussian state-space modeling and evaluate its performance on both synthetic and real data.
Branching processes are a class of continuous-time Markov chains (CTMCs) with ubiquitous applications. A general difficulty in statistical inference under partially observed CTMC models arises in computing transition probabilities when the discrete state space is large or uncountable. Classical methods such as matrix exponentiation are infeasible for large or countably infinite state spaces, and sampling-based alternatives are computationally intensive, requiring integration over all possible hidden events. Recent work has successfully applied generating function techniques to computing transition probabilities for linear multi-type branching processes. While these techniques often require significantly fewer computations than matrix exponentiation, they also become prohibitive in applications with large populations. We propose a compressed sensing framework that significantly accelerates the generating function method, decreasing computational cost up to a logarithmic factor by only assuming the probability mass of transitions is sparse. We demonstrate accurate and efficient transition probability computations in branching process models for blood cell formation and evolution of self-replicating transposable elements in bacterial genomes.
Standard time series structure learning algorithms assume that the measurement timescale is approximately the same as the timescale of the underlying (causal) system. In many scientific contexts, however, this assumption is violated: the measurement timescale can be substantially slower than the system timescale (so intermediate time series datapoints will be missing). This assumption violation can lead to significant learning errors. In this paper, we provide a novel learning algorithm to extract system-timescale structure from measurement data that undersample the underlying system. We employ multiple algorithmic optimizations that exploit the problem structure in order to achieve computational tractability. The resulting algorithm is highly reliable at extracting system-timescale structure from undersampled data.
We present a peptide-spectrum alignment strategy that employs a dynamic Bayesian network (DBN) for the identification of spectra produced by tandem mass spectrometry (MS/MS). Our method is fundamentally generative in that it models peptide fragmentation in MS/MS as a physical process. The model traverses an observed MS/MS spectrum and a peptide-based theoretical spectrum to calculate the best alignment between the two spectra. Unlike all existing state-of-the-art methods for spectrum identification that we are aware of, our method can learn alignment probabilities given a dataset of high-quality peptide-spectrum pairs. The method, moreover, accounts for noise peaks and absent theoretical peaks in the observed spectrum. We demonstrate that our method out-performs, on a majority of datasets, several widely used, state-of-the-art database search tools for spectrum identification. Furthermore, the proposed approach provides an extensible framework for MS/MS analysis and provides useful information that is not produced by other methods, thanks to its generative structure.
The episodic, irregular and asynchronous nature of medical data render them difficult substrates for standard machine learning algorithms. We would like to abstract away this difficulty for the class of time-stamped categorical variables (or events) by modeling them as a renewal process and inferring a probability density over non-parametric longitudinal intensity functions that modulate the process. Several methods exist for inferring such a density over intensity functions, but either their constraints prevent their use with our potentially bursty event streams, or their time complexity renders their use intractable on our long-duration observations of high-resolution events, or both. In this paper we present a new efficient and flexible inference method that uses direct numeric integration and smooth interpolation over Gaussian processes. We demonstrate that our direct method is up to twice as accurate and two orders of magnitude more efficient than the best existing method (thinning). Importantly, our direct method can infer intensity functions over the full range of bursty to memoryless to regular events, which thinning and many other methods cannot do. Finally, we apply the method to clinical event data and demonstrate a simple example application facilitated by the abstraction.
