A. Taylan Cemgil

ICML Conference 2024 Conference Paper

Evaluating Model Bias Requires Characterizing its Mistakes

• Isabela Albuquerque
• Jessica Schrouff
• David Warde-Farley
• A. Taylan Cemgil
• Sven Gowal
• Olivia Wiles

The ability to properly benchmark model performance in the face of spurious correlations is important to both build better predictors and increase confidence that models are operating as intended. We demonstrate that characterizing (as opposed to simply quantifying) model mistakes across subgroups is pivotal to properly reflect model biases, which are ignored by standard metrics such as worst-group accuracy or accuracy gap. Inspired by the hypothesis testing framework, we introduce SkewSize, a principled and flexible metric that captures bias from mistakes in a model’s predictions. It can be used in multi-class settings or generalised to the open vocabulary setting of generative models. SkewSize is an aggregation of the effect size of the interaction between two categorical variables: the spurious variable representing the bias attribute the model’s prediction. We demonstrate the utility of SkewSize in multiple settings including: standard vision models trained on synthetic data, vision models trained on ImageNet, and large scale vision-and-language models from the BLIP-2 family. In each case, the proposed SkewSize is able to highlight biases not captured by other metrics, while also providing insights on the impact of recently proposed techniques, such as instruction tuning.

ICML Conference 2023 Conference Paper

Transformers Meet Directed Graphs

• Simon Geisler
• Yujia Li 0001
• Daniel J. Mankowitz
• A. Taylan Cemgil
• Stephan Günnemann
• Cosmin Paduraru

Transformers were originally proposed as a sequence-to-sequence model for text but have become vital for a wide range of modalities, including images, audio, video, and undirected graphs. However, transformers for directed graphs are a surprisingly underexplored topic, despite their applicability to ubiquitous domains, including source code and logic circuits. In this work, we propose two direction- and structure-aware positional encodings for directed graphs: (1) the eigenvectors of the Magnetic Laplacian — a direction-aware generalization of the combinatorial Laplacian; (2) directional random walk encodings. Empirically, we show that the extra directionality information is useful in various downstream tasks, including correctness testing of sorting networks and source code understanding. Together with a data-flow-centric graph construction, our model outperforms the prior state of the art on the Open Graph Benchmark Code2 relatively by 14. 7%.

ICLR Conference 2022 Conference Paper

A Fine-Grained Analysis on Distribution Shift

• Olivia Wiles
• Sven Gowal
• Florian Stimberg
• Sylvestre-Alvise Rebuffi
• Ira Ktena
• Krishnamurthy Dvijotham
• A. Taylan Cemgil

Robustness to distribution shifts is critical for deploying machine learning models in the real world. Despite this necessity, there has been little work in defining the underlying mechanisms that cause these shifts and evaluating the robustness of algorithms across multiple, different distribution shifts. To this end, we introduce a framework that enables fine-grained analysis of various distribution shifts. We provide a holistic analysis of current state-of-the-art methods by evaluating 19 distinct methods grouped into five categories across both synthetic and real-world datasets. Overall, we train more than 85K models. Our experimental framework can be easily extended to include new methods, shifts, and datasets. We find, unlike previous work (Gulrajani & Lopez-Paz, 2021), that progress has been made over a standard ERM baseline; in particular, pretraining and augmentations (learned or heuristic) offer large gains in many cases. However, the best methods are not consistent over different datasets and shifts. We will open source our experimental framework, allowing future work to evaluate new methods over multiple shifts to obtain a more complete picture of a method's effectiveness. Code is available at github.com/deepmind/distribution_shift_framework.

ICML Conference 2022 Conference Paper

Evaluating the Adversarial Robustness of Adaptive Test-time Defenses

• Francesco Croce
• Sven Gowal
• Thomas Brunner
• Evan Shelhamer
• Matthias Hein 0001
• A. Taylan Cemgil

Adaptive defenses, which optimize at test time, promise to improve adversarial robustness. We categorize such adaptive test-time defenses, explain their potential benefits and drawbacks, and evaluate a representative variety of the latest adaptive defenses for image classification. Unfortunately, none significantly improve upon static defenses when subjected to our careful case study evaluation. Some even weaken the underlying static model while simultaneously increasing inference computation. While these results are disappointing, we still believe that adaptive test-time defenses are a promising avenue of research and, as such, we provide recommendations for their thorough evaluation. We extend the checklist of Carlini et al. (2019) by providing concrete steps specific to adaptive defenses.

ICLR Conference 2022 Conference Paper

Learning Optimal Conformal Classifiers

• David Stutz
• Krishnamurthy Dvijotham
• A. Taylan Cemgil
• Arnaud Doucet

Modern deep learning based classifiers show very high accuracy on test data but this does not provide sufficient guarantees for safe deployment, especially in high-stake AI applications such as medical diagnosis. Usually, predictions are obtained without a reliable uncertainty estimate or a formal guarantee. Conformal prediction (CP) addresses these issues by using the classifier's predictions, e.g., its probability estimates, to predict confidence sets containing the true class with a user-specified probability. However, using CP as a separate processing step after training prevents the underlying model from adapting to the prediction of confidence sets. Thus, this paper explores strategies to differentiate through CP during training with the goal of training model with the conformal wrapper end-to-end. In our approach, conformal training (ConfTr), we specifically "simulate" conformalization on mini-batches during training. Compared to standard training, ConfTr reduces the average confidence set size (inefficiency) of state-of-the-art CP methods applied after training. Moreover, it allows to "shape" the confidence sets predicted at test time, which is difficult for standard CP. On experiments with several datasets, we show ConfTr can influence how inefficiency is distributed across classes, or guide the composition of confidence sets in terms of the included classes, while retaining the guarantees offered by CP.

UAI Conference 2021 Conference Paper

Unbiased gradient estimation for variational auto-encoders using coupled Markov chains

• Francisco J. R. Ruiz
• Michalis K. Titsias
• A. Taylan Cemgil
• Arnaud Doucet

The variational auto-encoder (VAE) is a deep latent variable model that has two neural networks in an autoencoder-like architecture; one of them parameterizes the model’s likelihood. Fitting its parameters via maximum likelihood (ML) is challenging since the computation of the marginal likelihood involves an intractable integral over the latent space; thus the VAE is trained instead by maximizing a variational lower bound. Here, we develop a ML training scheme for VAEs by introducing unbiased estimators of the log-likelihood gradient. We obtain the estimators by augmenting the latent space with a set of importance samples, similarly to the importance weighted auto-encoder (IWAE), and then constructing a Markov chain Monte Carlo coupling procedure on this augmented space. We provide the conditions under which the estimators can be computed in finite time and with finite variance. We show experimentally that VAEs fitted with unbiased estimators exhibit better predictive performance.

ICLR Conference 2020 Conference Paper

Adversarially Robust Representations with Smooth Encoders

• A. Taylan Cemgil
• Sumedh Ghaisas
• Krishnamurthy Dvijotham
• Pushmeet Kohli

This paper studies the undesired phenomena of over-sensitivity of representations learned by deep networks to semantically-irrelevant changes in data. We identify a cause for this shortcoming in the classical Variational Auto-encoder (VAE) objective, the evidence lower bound (ELBO). We show that the ELBO fails to control the behaviour of the encoder out of the support of the empirical data distribution and this behaviour of the VAE can lead to extreme errors in the learned representation. This is a key hurdle in the effective use of representations for data-efficient learning and transfer. To address this problem, we propose to augment the data with specifications that enforce insensitivity of the representation with respect to families of transformations. To incorporate these specifications, we propose a regularization method that is based on a selection mechanism that creates a fictive data point by explicitly perturbing an observed true data point. For certain choices of parameters, our formulation naturally leads to the minimization of the entropy regularized Wasserstein distance between representations. We illustrate our approach on standard datasets and experimentally show that significant improvements in the downstream adversarial accuracy can be achieved by learning robust representations completely in an unsupervised manner, without a reference to a particular downstream task and without a costly supervised adversarial training procedure.

ICML Conference 2018 Conference Paper

Asynchronous Stochastic Quasi-Newton MCMC for Non-Convex Optimization

• Umut Simsekli
• Çagatay Yildiz
• Thanh Huy Nguyen 0001
• A. Taylan Cemgil
• Gaël Richard

Recent studies have illustrated that stochastic gradient Markov Chain Monte Carlo techniques have a strong potential in non-convex optimization, where local and global convergence guarantees can be shown under certain conditions. By building up on this recent theory, in this study, we develop an asynchronous-parallel stochastic L-BFGS algorithm for non-convex optimization. The proposed algorithm is suitable for both distributed and shared-memory settings. We provide formal theoretical analysis and show that the proposed method achieves an ergodic convergence rate of ${\cal O}(1/\sqrt{N})$ ($N$ being the total number of iterations) and it can achieve a linear speedup under certain conditions. We perform several experiments on both synthetic and real datasets. The results support our theory and show that the proposed algorithm provides a significant speedup over the recently proposed synchronous distributed L-BFGS algorithm.

ICRA Conference 2018 Conference Paper

EndoSensorFusion: Particle Filtering-Based Multi-Sensory Data Fusion with Switching State-Space Model for Endoscopic Capsule Robots

• Mehmet Turan
• Yasin Almalioglu
• Hunter B. Gilbert
• Helder Araújo
• A. Taylan Cemgil
• Metin Sitti

A reliable, real time, multi-sensor fusion functionality is crucial for localization of actively controlled capsule endoscopy robots, which are an emerging, minimally invasive diagnostic and therapeutic technology for the gastrointestinal (GI) tract. In this study, we propose a novel multi-sensor fusion approach based on a particle filter that incorporates an online estimation of sensor reliability and a non-linear kinematic model learned by a recurrent neural network. Our method sequentially estimates the true robot pose from noisy pose observations delivered by multiple sensors. We experimentally test the method using 5 degree-of-freedom (5-DoF) absolute pose measurement by a magnetic localization system and a 6-DoF relative pose measurement by visual odometry. In addition, the proposed method is capable of detecting and handling sensor failures by ignoring corrupted data, providing the robustness expected of a medical device. Detailed analyses and evaluations are presented using ex vivo experiments on a porcine stomach model, proving that our system achieves high translational and rotational accuracies for different types of endoscopic capsule robot trajectories.

ICML Conference 2016 Conference Paper

Stochastic Quasi-Newton Langevin Monte Carlo

• Umut Simsekli
• Roland Badeau
• A. Taylan Cemgil
• Gaël Richard

Recently, Stochastic Gradient Markov Chain Monte Carlo (SG-MCMC) methods have been proposed for scaling up Monte Carlo computations to large data problems. Whilst these approaches have proven useful in many applications, vanilla SG-MCMC might suffer from poor mixing rates when random variables exhibit strong couplings under the target densities or big scale differences. In this study, we propose a novel SG-MCMC method that takes the local geometry into account by using ideas from Quasi-Newton optimization methods. These second order methods directly approximate the inverse Hessian by using a limited history of samples and their gradients. Our method uses dense approximations of the inverse Hessian while keeping the time and memory complexities linear with the dimension of the problem. We provide a formal theoretical analysis where we show that the proposed method is asymptotically unbiased and consistent with the posterior expectations. We illustrate the effectiveness of the approach on both synthetic and real datasets. Our experiments on two challenging applications show that our method achieves fast convergence rates similar to Riemannian approaches while at the same time having low computational requirements similar to diagonal preconditioning approaches.

ICML Conference 2013 Conference Paper

Learning the beta-Divergence in Tweedie Compound Poisson Matrix Factorization Models

• Umut Simsekli
• A. Taylan Cemgil
• Yusuf Kenan Yilmaz

In this study, we derive algorithms for estimating mixed β-divergences. Such cost functions are useful for Nonnegative Matrix and Tensor Factorization models with a compound Poisson observation model. Compound Poisson is a particular Tweedie model, an important special case of exponential dispersion models characterized by the fact that the variance is proportional to a power function of the mean. There are several well known matrix and tensor factorization algorithms that minimize the β-divergence; these estimate the mean parameter. The probabilistic interpretation gives us more flexibility and robustness by providing us additional tunable parameters such as power and dispersion. Estimation of the power parameter is useful for choosing a suitable divergence and estimation of dispersion is useful for data driven regularization and weighting in collective/coupled factorization of heterogeneous datasets. We present three inference algorithms for both estimating the factors and the additional parameters of the compound Poisson distribution. The methods are evaluated on two applications: modeling symbolic representations for polyphonic music and lyric prediction from audio features. Our conclusion is that the compound poisson based factorization models can be useful for sparse positive data.

EWRL Workshop 2013 Workshop Paper

Sequentially Interacting Markov Chain Monte Carlo Based Policy Iteration

• Orhan Sönmez
• A. Taylan Cemgil