Bernhard Schoelkopf

Justus Mattern, Zhijing Jin, Benjamin Weggenmann et al.

To protect the privacy of individuals whose data is being shared, it is of high importance to develop methods allowing researchers and companies to release textual data while providing formal privacy guarantees to its originators. In the field of NLP, substantial efforts have been directed at building mechanisms following the framework of local differential privacy, thereby anonymizing individual text samples before releasing them. In practice, these approaches are often dissatisfying in terms of the quality of their output language due to the strong noise required for local differential privacy. In this paper, we approach the problem at hand using global differential privacy, particularly by training a generative language model in a differentially private manner and consequently sampling data from it. Using natural language prompts and a new prompt-mismatch loss, we are able to create highly accurate and fluent textual datasets taking on specific desired attributes such as sentiment or topic and resembling statistical properties of the training data. We perform thorough experiments indicating that our synthetic datasets do not leak information from our original data and are of high language quality and highly suitable for training models for further analysis on real-world data. Notably, we also demonstrate that training classifiers on private synthetic data outperforms directly training classifiers on real data with DP-SGD.

Yiwen Ding, Jiarui Liu, Zhiheng Lyu et al.

While several previous studies have analyzed gender bias in research, we are still missing a comprehensive analysis of gender differences in the AI community, covering diverse topics and different development trends. Using the AI Scholar dataset of 78K researchers in the field of AI, we identify several gender differences: (1) Although female researchers tend to have fewer overall citations than males, this citation difference does not hold for all academic-age groups; (2) There exist large gender homophily in co-authorship on AI papers; (3) Female first-authored papers show distinct linguistic styles, such as longer text, more positive emotion words, and more catchy titles than male first-authored papers. Our analysis provides a window into the current demographic trends in our AI community, and encourages more gender equality and diversity in the future. Our code and data are at https://github.com/causalNLP/ai-scholar-gender.

Zhiheng Lyu, Zhijing Jin, Justus Mattern et al.

NLP datasets are richer than just input-output pairs; rather, they carry causal relations between the input and output variables. In this work, we take sentiment classification as an example and look into the causal relations between the review (X) and sentiment (Y). As psychology studies show that language can affect emotion, different psychological processes are evoked when a person first makes a rating and then self-rationalizes their feeling in a review (where the sentiment causes the review, i.e., Y -> X), versus first describes their experience, and weighs the pros and cons to give a final rating (where the review causes the sentiment, i.e., X -> Y ). Furthermore, it is also a completely different psychological process if an annotator infers the original rating of the user by theory of mind (ToM) (where the review causes the rating, i.e., X -ToM-> Y ). In this paper, we verbalize these three causal mechanisms of human psychological processes of sentiment classification into three different causal prompts, and study (1) how differently they perform, and (2) what nature of sentiment classification data leads to agreement or diversity in the model responses elicited by the prompts. We suggest future work raise awareness of different causal structures in NLP tasks. Our code and data are at https://github.com/cogito233/psych-causal-prompt

Dominik Janzing, Bernhard Schoelkopf

We consider linear models where $d$ potential causes $X_1,...,X_d$ are correlated with one target quantity $Y$ and propose a method to infer whether the association is causal or whether it is an artifact caused by overfitting or hidden common causes. We employ the idea that in the former case the vector of regression coefficients has 'generic' orientation relative to the covariance matrix $Σ_{XX}$ of $X$. Using an ICA based model for confounding, we show that both confounding and overfitting yield regression vectors that concentrate mainly in the space of low eigenvalues of $Σ_{XX}$.

Paul K. Rubenstein, Bernhard Schoelkopf, Ilya Tolstikhin

We study the role of latent space dimensionality in Wasserstein auto-encoders (WAEs). Through experimentation on synthetic and real datasets, we argue that random encoders should be preferred over deterministic encoders. We highlight the potential of WAEs for representation learning with promising results on a benchmark disentanglement task.

Behzad Tabibian, Utkarsh Upadhyay, Abir De et al.

Spaced repetition is a technique for efficient memorization which uses repeated, spaced review of content to improve long-term retention. Can we find the optimal reviewing schedule to maximize the benefits of spaced repetition? In this paper, we introduce a novel, flexible representation of spaced repetition using the framework of marked temporal point processes and then address the above question as an optimal control problem for stochastic differential equations with jumps. For two well-known human memory models, we show that the optimal reviewing schedule is given by the recall probability of the content to be learned. As a result, we can then develop a simple, scalable online algorithm, Memorize, to sample the optimal reviewing times. Experiments on both synthetic and real data gathered from Duolingo, a popular language-learning online platform, show that our algorithm may be able to help learners memorize more effectively than alternatives.

Jooyeon Kim, Behzad Tabibian, Alice Oh et al.

Online social networking sites are experimenting with the following crowd-powered procedure to reduce the spread of fake news and misinformation: whenever a user is exposed to a story through her feed, she can flag the story as misinformation and, if the story receives enough flags, it is sent to a trusted third party for fact checking. If this party identifies the story as misinformation, it is marked as disputed. However, given the uncertain number of exposures, the high cost of fact checking, and the trade-off between flags and exposures, the above mentioned procedure requires careful reasoning and smart algorithms which, to the best of our knowledge, do not exist to date. In this paper, we first introduce a flexible representation of the above procedure using the framework of marked temporal point processes. Then, we develop a scalable online algorithm, Curb, to select which stories to send for fact checking and when to do so to efficiently reduce the spread of misinformation with provable guarantees. In doing so, we need to solve a novel stochastic optimal control problem for stochastic differential equations with jumps, which is of independent interest. Experiments on two real-world datasets gathered from Twitter and Weibo show that our algorithm may be able to effectively reduce the spread of fake news and misinformation.

Ilya Tolstikhin, Olivier Bousquet, Sylvain Gelly et al.

We propose the Wasserstein Auto-Encoder (WAE)---a new algorithm for building a generative model of the data distribution. WAE minimizes a penalized form of the Wasserstein distance between the model distribution and the target distribution, which leads to a different regularizer than the one used by the Variational Auto-Encoder (VAE). This regularizer encourages the encoded training distribution to match the prior. We compare our algorithm with several other techniques and show that it is a generalization of adversarial auto-encoders (AAE). Our experiments show that WAE shares many of the properties of VAEs (stable training, encoder-decoder architecture, nice latent manifold structure) while generating samples of better quality, as measured by the FID score.

Paul K. Rubenstein, Ilya Tolstikhin, Philipp Hennig et al.

We consider the problem of learning the functions computing children from parents in a Structural Causal Model once the underlying causal graph has been identified. This is in some sense the second step after causal discovery. Taking a probabilistic approach to estimating these functions, we derive a natural myopic active learning scheme that identifies the intervention which is optimally informative about all of the unknown functions jointly, given previously observed data. We test the derived algorithms on simple examples, to demonstrate that they produce a structured exploration policy that significantly improves on unstructured base-lines.

Olivier Bousquet, Sylvain Gelly, Ilya Tolstikhin et al.

We study unsupervised generative modeling in terms of the optimal transport (OT) problem between true (but unknown) data distribution $P_X$ and the latent variable model distribution $P_G$. We show that the OT problem can be equivalently written in terms of probabilistic encoders, which are constrained to match the posterior and prior distributions over the latent space. When relaxed, this constrained optimization problem leads to a penalized optimal transport (POT) objective, which can be efficiently minimized using stochastic gradient descent by sampling from $P_X$ and $P_G$. We show that POT for the 2-Wasserstein distance coincides with the objective heuristically employed in adversarial auto-encoders (AAE) (Makhzani et al., 2016), which provides the first theoretical justification for AAEs known to the authors. We also compare POT to other popular techniques like variational auto-encoders (VAE) (Kingma and Welling, 2014). Our theoretical results include (a) a better understanding of the commonly observed blurriness of images generated by VAEs, and (b) establishing duality between Wasserstein GAN (Arjovsky and Bottou, 2017) and POT for the 1-Wasserstein distance.

Dominik Janzing, Bernhard Schoelkopf

We study a model where one target variable Y is correlated with a vector X:=(X_1,...,X_d) of predictor variables being potential causes of Y. We describe a method that infers to what extent the statistical dependences between X and Y are due to the influence of X on Y and to what extent due to a hidden common cause (confounder) of X and Y. The method relies on concentration of measure results for large dimensions d and an independence assumption stating that, in the absence of confounding, the vector of regression coefficients describing the influence of each X on Y typically has `generic orientation' relative to the eigenspaces of the covariance matrix of X. For the special case of a scalar confounder we show that confounding typically spoils this generic orientation in a characteristic way that can be used to quantitatively estimate the amount of confounding.

Paul K. Rubenstein, Stephan Bongers, Bernhard Schoelkopf et al.

Structural Causal Models are widely used in causal modelling, but how they relate to other modelling tools is poorly understood. In this paper we provide a novel perspective on the relationship between Ordinary Differential Equations and Structural Causal Models. We show how, under certain conditions, the asymptotic behaviour of an Ordinary Differential Equation under non-constant interventions can be modelled using Dynamic Structural Causal Models. In contrast to earlier work, we study not only the effect of interventions on equilibrium states; rather, we model asymptotic behaviour that is dynamic under interventions that vary in time, and include as a special case the study of static equilibria.

Philipp Geiger, Lucian Carata, Bernhard Schoelkopf

Cloud computing involves complex technical and economical systems and interactions. This brings about various challenges, two of which are: (1) debugging and control to optimize the performance of computing systems, with the help of sandbox experiments, and (2) privacy-preserving prediction of the cost of ``spot'' resources for decision making of cloud clients. In this paper, we formalize debugging by counterfactual probabilities and control by post-(soft-)interventional probabilities. We prove that counterfactuals can approximately be calculated from a ``stochastic'' graphical causal model (while they are originally defined only for ``deterministic'' functional causal models), and based on this sketch a data-driven approach to address problem (1). To address problem (2), we formalize bidding by post-(soft-)interventional probabilities and present a simple mathematical result on approximate integration of ``incomplete'' conditional probability distributions. We show how this can be used by cloud clients to trade off privacy against predictability of the outcome of their bidding actions in a toy scenario. We report experiments on simulated and real data.

Krikamol Muandet, Bernhard Schoelkopf

We propose one-class support measure machines (OCSMMs) for group anomaly detection which aims at recognizing anomalous aggregate behaviors of data points. The OCSMMs generalize well-known one-class support vector machines (OCSVMs) to a space of probability measures. By formulating the problem as quantile estimation on distributions, we can establish an interesting connection to the OCSVMs and variable kernel density estimators (VKDEs) over the input space on which the distributions are defined, bridging the gap between large-margin methods and kernel density estimators. In particular, we show that various types of VKDEs can be considered as solutions to a class of regularization problems studied in this paper. Experiments on Sloan Digital Sky Survey dataset and High Energy Particle Physics dataset demonstrate the benefits of the proposed framework in real-world applications.

Joris Mooij, Dominik Janzing, Bernhard Schoelkopf

We show how, and under which conditions, the equilibrium states of a first-order Ordinary Differential Equation (ODE) system can be described with a deterministic Structural Causal Model (SCM). Our exposition sheds more light on the concept of causality as expressed within the framework of Structural Causal Models, especially for cyclic models.

Hadi Daneshmand, Manuel Gomez-Rodriguez, Le Song et al.

Information spreads across social and technological networks, but often the network structures are hidden from us and we only observe the traces left by the diffusion processes, called cascades. Can we recover the hidden network structures from these observed cascades? What kind of cascades and how many cascades do we need? Are there some network structures which are more difficult than others to recover? Can we design efficient inference algorithms with provable guarantees? Despite the increasing availability of cascade data and methods for inferring networks from these data, a thorough theoretical understanding of the above questions remains largely unexplored in the literature. In this paper, we investigate the network structure inference problem for a general family of continuous-time diffusion models using an $l_1$-regularized likelihood maximization framework. We show that, as long as the cascade sampling process satisfies a natural incoherence condition, our framework can recover the correct network structure with high probability if we observe $O(d^3 \log N)$ cascades, where $d$ is the maximum number of parents of a node and $N$ is the total number of nodes. Moreover, we develop a simple and efficient soft-thresholding inference algorithm, which we use to illustrate the consequences of our theoretical results, and show that our framework outperforms other alternatives in practice.

Mikhail Langovoy, Michael Habeck, Bernhard Schoelkopf

Cryo-electron microscopy (cryo-EM) is an emerging experimental method to characterize the structure of large biomolecular assemblies. Single particle cryo-EM records 2D images (so-called micrographs) of projections of the three-dimensional particle, which need to be processed to obtain the three-dimensional reconstruction. A crucial step in the reconstruction process is particle picking which involves detection of particles in noisy 2D micrographs with low signal-to-noise ratios of typically 1:10 or even lower. Typically, each picture contains a large number of particles, and particles have unknown irregular and nonconvex shapes.

Eleni Sgouritsa, Dominik Janzing, Jonas Peters et al.

We propose a kernel method to identify finite mixtures of nonparametric product distributions. It is based on a Hilbert space embedding of the joint distribution. The rank of the constructed tensor is equal to the number of mixture components. We present an algorithm to recover the components by partitioning the data points into clusters such that the variables are jointly conditionally independent given the cluster. This method can be used to identify finite confounders.

Manuel Gomez Rodriguez, Jure Leskovec, Bernhard Schoelkopf

Networks provide a skeleton for the spread of contagions, like, information, ideas, behaviors and diseases. Many times networks over which contagions diffuse are unobserved and need to be inferred. Here we apply survival theory to develop general additive and multiplicative risk models under which the network inference problems can be solved efficiently by exploiting their convexity. Our additive risk model generalizes several existing network inference models. We show all these models are particular cases of our more general model. Our multiplicative model allows for modeling scenarios in which a node can either increase or decrease the risk of activation of another node, in contrast with previous approaches, which consider only positive risk increments. We evaluate the performance of our network inference algorithms on large synthetic and real cascade datasets, and show that our models are able to predict the length and duration of cascades in real data.

Bernhard Schoelkopf, Dominik Janzing, Jonas Peters et al.

We consider the problem of function estimation in the case where an underlying causal model can be inferred. This has implications for popular scenarios such as covariate shift, concept drift, transfer learning and semi-supervised learning. We argue that causal knowledge may facilitate some approaches for a given problem, and rule out others. In particular, we formulate a hypothesis for when semi-supervised learning can help, and corroborate it with empirical results.

Dominik Janzing, Jonas Peters, Joris Mooij et al.

We propose a method for inferring the existence of a latent common cause ('confounder') of two observed random variables. The method assumes that the two effects of the confounder are (possibly nonlinear) functions of the confounder plus independent, additive noise. We discuss under which conditions the model is identifiable (up to an arbitrary reparameterization of the confounder) from the joint distribution of the effects. We state and prove a theoretical result that provides evidence for the conjecture that the model is generically identifiable under suitable technical conditions. In addition, we propose a practical method to estimate the confounder from a finite i.i.d. sample of the effects and illustrate that the method works well on both simulated and real-world data.

Povilas Daniusis, Dominik Janzing, Joris Mooij et al.

We consider two variables that are related to each other by an invertible function. While it has previously been shown that the dependence structure of the noise can provide hints to determine which of the two variables is the cause, we presently show that even in the deterministic (noise-free) case, there are asymmetries that can be exploited for causal inference. Our method is based on the idea that if the function and the probability density of the cause are chosen independently, then the distribution of the effect will, in a certain sense, depend on the function. We provide a theoretical analysis of this method, showing that it also works in the low noise regime, and link it to information geometry. We report strong empirical results on various real-world data sets from different domains.

Kun Zhang, Bernhard Schoelkopf, Dominik Janzing

In nonlinear latent variable models or dynamic models, if we consider the latent variables as confounders (common causes), the noise dependencies imply further relations between the observed variables. Such models are then closely related to causal discovery in the presence of nonlinear confounders, which is a challenging problem. However, generally in such models the observation noise is assumed to be independent across data dimensions, and consequently the noise dependencies are ignored. In this paper we focus on the Gaussian process latent variable model (GPLVM), from which we develop an extended model called invariant GPLVM (IGPLVM), which can adapt to arbitrary noise covariances. With the Gaussian process prior put on a particular transformation of the latent nonlinear functions, instead of the original ones, the algorithm for IGPLVM involves almost the same computational loads as that for the original GPLVM. Besides its potential application in causal discovery, IGPLVM has the advantage that its estimated latent nonlinear manifold is invariant to any nonsingular linear transformation of the data. Experimental results on both synthetic and realworld data show its encouraging performance in nonlinear manifold learning and causal discovery.

Dominik Janzing, Eleni Sgouritsa, Oliver Stegle et al.

We describe a method that infers whether statistical dependences between two observed variables X and Y are due to a "direct" causal link or only due to a connecting causal path that contains an unobserved variable of low complexity, e.g., a binary variable. This problem is motivated by statistical genetics. Given a genetic marker that is correlated with a phenotype of interest, we want to detect whether this marker is causal or it only correlates with a causal one. Our method is based on the analysis of the location of the conditional distributions P(Y|x) in the simplex of all distributions of Y. We report encouraging results on semi-empirical data.

Kun Zhang, Jonas Peters, Dominik Janzing et al.

Conditional independence testing is an important problem, especially in Bayesian network learning and causal discovery. Due to the curse of dimensionality, testing for conditional independence of continuous variables is particularly challenging. We propose a Kernel-based Conditional Independence test (KCI-test), by constructing an appropriate test statistic and deriving its asymptotic distribution under the null hypothesis of conditional independence. The proposed method is computationally efficient and easy to implement. Experimental results show that it outperforms other methods, especially when the conditioning set is large or the sample size is not very large, in which case other methods encounter difficulties.

Jonas Peters, Joris Mooij, Dominik Janzing et al.

This work addresses the following question: Under what assumptions on the data generating process can one infer the causal graph from the joint distribution? The approach taken by conditional independence-based causal discovery methods is based on two assumptions: the Markov condition and faithfulness. It has been shown that under these assumptions the causal graph can be identified up to Markov equivalence (some arrows remain undirected) using methods like the PC algorithm. In this work we propose an alternative by defining Identifiable Functional Model Classes (IFMOCs). As our main theorem we prove that if the data generating process belongs to an IFMOC, one can identify the complete causal graph. To the best of our knowledge this is the first identifiability result of this kind that is not limited to linear functional relationships. We discuss how the IFMOC assumption and the Markov and faithfulness assumptions relate to each other and explain why we believe that the IFMOC assumption can be tested more easily on given data. We further provide a practical algorithm that recovers the causal graph from finitely many data; experiments on simulated data support the theoretical findings.