Machine Learning

---

Mediation Challenges and Socio-Technical Gaps for Explainable Deep Learning Applications (1907.07178v1)

Rafael Brandão, Joel Carbonera, Clarisse de Souza, Juliana Ferreira, Bernardo Gonçalves, Carla Leitão

2019-07-16

The presumed data owners' right to explanations brought about by the General Data Protection Regulation in Europe has shed light on the social challenges of explainable artificial intelligence (XAI). In this paper, we present a case study with Deep Learning (DL) experts from a research and development laboratory focused on the delivery of industrial-strength AI technologies. Our aim was to investigate the social meaning (i.e. meaning to others) that DL experts assign to what they do, given a richly contextualized and familiar domain of application. Using qualitative research techniques to collect and analyze empirical data, our study has shown that participating DL experts did not spontaneously engage into considerations about the social meaning of machine learning models that they build. Moreover, when explicitly stimulated to do so, these experts expressed expectations that, with real-world DL application, there will be available mediators to bridge the gap between technical meanings that drive DL work, and social meanings that AI technology users assign to it. We concluded that current research incentives and values guiding the participants' scientific interests and conduct are at odds with those required to face some of the scientific challenges involved in advancing XAI, and thus responding to the alleged data owners' right to explanations or similar societal demands emerging from current debates. As a concrete contribution to mitigate what seems to be a more general problem, we propose three preliminary XAI Mediation Challenges with the potential to bring together technical and social meanings of DL applications, as well as to foster much needed interdisciplinary collaboration among AI and the Social Sciences researchers.

Outlier Robust Extreme Learning Machine for Multi-Target Regression (1905.09368v2)

Bruno Légora Souza da Silva, Fernando Kentaro Inaba, Evandro Ottoni Teatini Salles, Patrick Marques Ciarelli

2019-05-22

The popularity of algorithms based on Extreme Learning Machine (ELM), which can be used to train Single Layer Feedforward Neural Networks (SLFN), has increased in the past years. They have been successfully applied to a wide range of classification and regression tasks. The most commonly used methods are the ones based on minimizing the norm of the error, which is not suitable to deal with outliers, essentially in regression tasks. The use of norm was proposed in Outlier Robust ELM (OR-ELM), which is defined to one-dimensional outputs. In this paper, we generalize OR-ELM to deal with multi-target regression problems, using the error norm and the Elastic Net theory, which can result in a more sparse network, resulting in our method, Generalized Outlier Robust ELM (GOR-ELM). We use Alternating Direction Method of Multipliers (ADMM) to solve the resulting optimization problem. An incremental version of GOR-ELM is also proposed. We chose 15 public real-world multi-target regression datasets to test our methods. Our conducted experiments show that they are statistically better than other ELM-based techniques, when considering data contaminated with outliers, and equivalent to them, otherwise.

Natural Adversarial Examples (1907.07174v1)

Dan Hendrycks, Kevin Zhao, Steven Basart, Jacob Steinhardt, Dawn Song

2019-07-16

We introduce natural adversarial examples -- real-world, unmodified, and naturally occurring examples that cause classifier accuracy to significantly degrade. We curate 7,500 natural adversarial examples and release them in an ImageNet classifier test set that we call ImageNet-A. This dataset serves as a new way to measure classifier robustness. Like l_p adversarial examples, ImageNet-A examples successfully transfer to unseen or black-box classifiers. For example, on ImageNet-A a DenseNet-121 obtains around 2% accuracy, an accuracy drop of approximately 90%. Recovering this accuracy is not simple because ImageNet-A examples exploit deep flaws in current classifiers including their over-reliance on color, texture, and background cues. We observe that popular training techniques for improving robustness have little effect, but we show that some architectural changes can enhance robustness to natural adversarial examples. Future research is required to enable robust generalization to this hard ImageNet test set.

On the ''steerability" of generative adversarial networks (1907.07171v1)

Ali Jahanian, Lucy Chai, Phillip Isola

2019-07-16

An open secret in contemporary machine learning is that many models work beautifully on standard benchmarks but fail to generalize outside the lab. This has been attributed to training on biased data, which provide poor coverage over real world events. Generative models are no exception, but recent advances in generative adversarial networks (GANs) suggest otherwise -- these models can now synthesize strikingly realistic and diverse images. Is generative modeling of photos a solved problem? We show that although current GANs can fit standard datasets very well, they still fall short of being comprehensive models of the visual manifold. In particular, we study their ability to fit simple transformations such as camera movements and color changes. We find that the models reflect the biases of the datasets on which they are trained (e.g., centered objects), but that they also exhibit some capacity for generalization: by "steering" in latent space, we can shift the distribution while still creating realistic images. We hypothesize that the degree of distributional shift is related to the breadth of the training data distribution, and conduct experiments that demonstrate this. Code is released on our project page: https://ali-design.github.io/gan_steerability/

Fast, Provably convergent IRLS Algorithm for p-norm Linear Regression (1907.07167v1)

Deeksha Adil, Richard Peng, Sushant Sachdeva

2019-07-16

Linear regression in -norm is a canonical optimization problem that arises in several applications, including sparse recovery, semi-supervised learning, and signal processing. Generic convex optimization algorithms for solving -regression are slow in practice. Iteratively Reweighted Least Squares (IRLS) is an easy to implement family of algorithms for solving these problems that has been studied for over 50 years. However, these algorithms often diverge for p > 3, and since the work of Osborne (1985), it has been an open problem whether there is an IRLS algorithm that is guaranteed to converge rapidly for p > 3. We propose p-IRLS, the first IRLS algorithm that provably converges geometrically for any Our algorithm is simple to implement and is guaranteed to find a -approximate solution in iterations. Our experiments demonstrate that it performs even better than our theoretical bounds, beats the standard Matlab/CVX implementation for solving these problems by 10--50x, and is the fastest among available implementations in the high-accuracy regime.

Explaining Classifiers with Causal Concept Effect (CaCE) (1907.07165v1)

Yash Goyal, Uri Shalit, Been Kim

2019-07-16

How can we understand classification decisions made by deep neural nets? We propose answering this question by using ideas from causal inference. We define the ``Causal Concept Effect'' (CaCE) as the causal effect that the presence or absence of a concept has on the prediction of a given deep neural net. We then use this measure as a mean to understand what drives the network's prediction and what does not. Yet many existing interpretability methods rely solely on correlations, resulting in potentially misleading explanations. We show how CaCE can avoid such mistakes. In high-risk domains such as medicine, knowing the root cause of the prediction is crucial. If we knew that the network's prediction was caused by arbitrary concepts such as the lighting conditions in an X-ray room instead of medically meaningful concept, this would prevent us from disastrous deployment of such models. Estimating CaCE is difficult in situations where we cannot easily simulate the do-operator. As a simple solution, we propose learning a generative model, specifically a Variational AutoEncoder (VAE) on image pixels or image embeddings extracted from the classifier to measure VAE-CaCE. We show that VAE-CaCE is able to correctly estimate the true causal effect as compared to other baselines in controlled settings with synthetic and semi-natural high dimensional images.

The Tradeoff Between Privacy and Accuracy in Anomaly Detection Using Federated XGBoost (1907.07157v1)

Mengwei Yang, Linqi Song, Jie Xu, Congduan Li, Guozhen Tan

2019-07-16

Privacy has raised considerable concerns recently, especially with the advent of information explosion and numerous data mining techniques to explore the information inside large volumes of data. In this context, a new distributed learning paradigm termed federated learning becomes prominent recently to tackle the privacy issues in distributed learning, where only learning models will be transmitted from the distributed nodes to servers without revealing users' own data and hence protecting the privacy of users. In this paper, we propose a horizontal federated XGBoost algorithm to solve the federated anomaly detection problem, where the anomaly detection aims to identify abnormalities from extremely unbalanced datasets and can be considered as a special classification problem. Our proposed federated XGBoost algorithm incorporates data aggregation and sparse federated update processes to balance the tradeoff between privacy and learning performance. In particular, we introduce the virtual data sample by aggregating a group of users' data together at a single distributed node. We compute parameters based on these virtual data samples in the local nodes and aggregate the learning model in the central server. In the learning model upgrading process, we focus more on the wrongly classified data before in the virtual sample and hence to generate sparse learning model parameters. By carefully controlling the size of these groups of samples, we can achieve a tradeoff between privacy and learning performance. Our experimental results show the effectiveness of our proposed scheme by comparing with existing state-of-the-arts.

Efficient Segmentation: Learning Downsampling Near Semantic Boundaries (1907.07156v1)

Dmitrii Marin, Zijian He, Peter Vajda, Priyam Chatterjee, Sam Tsai, Fei Yang, Yuri Boykov

2019-07-16

Many automated processes such as auto-piloting rely on a good semantic segmentation as a critical component. To speed up performance, it is common to downsample the input frame. However, this comes at the cost of missed small objects and reduced accuracy at semantic boundaries. To address this problem, we propose a new content-adaptive downsampling technique that learns to favor sampling locations near semantic boundaries of target classes. Cost-performance analysis shows that our method consistently outperforms the uniform sampling improving balance between accuracy and computational efficiency. Our adaptive sampling gives segmentation with better quality of boundaries and more reliable support for smaller-size objects.

A Two-Stage Approach to Multivariate Linear Regression with Sparsely Mismatched Data (1907.07148v1)

Martin Slawski, Emanuel Ben-David, Ping Li

2019-07-16

A tacit assumption in linear regression is that (response, predictor)-pairs correspond to identical observational units. A series of recent works have studied scenarios in which this assumption is violated under terms such as Unlabeled Sensing andRegression with Unknown Permutation''. In this paper, we study the setup of multiple response variables and a notion of mismatches that generalizes permutations in order to allow for missing matches as well as for one-to-many matches. A two-stage method is proposed under the assumption that most pairs are correctly matched. In the first stage, the regression parameter is estimated by handling mismatches as contaminations, and subsequently the generalized permutation is estimated by a basic variant of matching. The approach is both computationally convenient and equipped with favorable statistical guarantees. Specifically, it is shown that the conditions for permutation recovery become considerably less stringent as the number of responses per observation increase. Particularly, for , the required signal-to-noise ratio does no longer depend on the sample size . Numerical results on synthetic and real data are presented to support the main findings of our analysis.

The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks (1802.08232v3)

Nicholas Carlini, Chang Liu, Úlfar Erlingsson, Jernej Kos, Dawn Song

2018-02-22

This paper describes a testing methodology for quantitatively assessing the risk that rare or unique training-data sequences are unintentionally memorized by generative sequence models---a common type of machine-learning model. Because such models are sometimes trained on sensitive data (e.g., the text of users' private messages), this methodology can benefit privacy by allowing deep-learning practitioners to select means of training that minimize such memorization. In experiments, we show that unintended memorization is a persistent, hard-to-avoid issue that can have serious consequences. Specifically, for models trained without consideration of memorization, we describe new, efficient procedures that can extract unique, secret sequences, such as credit card numbers. We show that our testing strategy is a practical and easy-to-use first line of defense, e.g., by describing its application to quantitatively limit data exposure in Google's Smart Compose, a commercial text-completion neural network trained on millions of users' email messages.

---