University of Wisconsin-Madison

About: University of Wisconsin-Madison is a education organization based out in Madison, Wisconsin, United States. It is known for research contribution in the topics: Population & Gene. The organization has 108707 authors who have published 237594 publications receiving 11883575 citations.

Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures

Abstract: Machine-learning (ML) algorithms are increasingly utilized in privacy-sensitive applications such as predicting lifestyle choices, making medical diagnoses, and facial recognition. In a model inversion attack, recently introduced in a case study of linear classifiers in personalized medicine by Fredrikson et al., adversarial access to an ML model is abused to learn sensitive genomic information about individuals. Whether model inversion attacks apply to settings outside theirs, however, is unknown. We develop a new class of model inversion attack that exploits confidence values revealed along with predictions. Our new attacks are applicable in a variety of settings, and we explore two in depth: decision trees for lifestyle surveys as used on machine-learning-as-a-service systems and neural networks for facial recognition. In both cases confidence values are revealed to those with the ability to make prediction queries to models. We experimentally show attacks that are able to estimate whether a respondent in a lifestyle survey admitted to cheating on their significant other and, in the other context, show how to recover recognizable images of people's faces given only their name and access to the ML model. We also initiate experimental exploration of natural countermeasures, investigating a privacy-aware decision tree training algorithm that is a simple variant of CART learning, as well as revealing only rounded confidence values. The lesson that emerges is that one can avoid these kinds of MI attacks with negligible degradation to utility.

The Great Psychotherapy Debate: Models, Methods and Findings

Abstract: Contents: Foreword. Preface. Competing Meta-Models: The Medical Model Versus the Contextual Model. Differential Hypotheses and Evidentiary Rules. Absolute Efficacy: The Benefits of Psychotherapy Established by Meta-Analysis. Relative Efficacy: The Dodo Bird Was Smarter Than We Have Been Led to Believe. Specific Effects: Weak Empirical Evidence That Benefits of Psychotherapy Are Derived From Specific Ingredients. General Effects: The Alliance as a Case in Point. Allegiance and Adherence: Further Evidence for the Contextual Model. Therapist Effects: An Ignored but Critical Factor. Implications of Rejecting the Medical Model.

Development of Depression From Preadolescence to Young Adulthood: Emerging Gender Differences in a 10-Year Longitudinal Study

Abstract: The authors investigated the emergence of gender differences in clinical depression and the overall development of depression from preadolescence to young adulthood among members of a complete birth cohort using a prospective longitudinal approach with structured diagnostic interviews administered 5 times over the course of 10 years. Small gender differences in depression (females greater than males) first began to emerge between the ages of 13 and 15. However, the greatest increase in this gender difference occurred between ages 15 and 18. Depression rates and accompanying gender differences for a university student subsample were no different than for a nonuniversity subsample. There was no gender difference for depression recurrence or for depression symptom severity. The peak increase in both overall rates of depression and new cases of depression occurred between the ages of 15 and 18. Results suggest that middle-to-late adolescence (ages 15-18) may be a critical time for studying vulnerability to depression because of the higher depression rates and the greater risk for depression onset and dramatic increase in gender differences in depression during this period.

Advances and open problems in federated learning

Abstract: The term Federated Learning was coined as recently as 2016 to describe a machine learning setting where multiple entities collaborate in solving a machine learning problem, under the coordination of a central server or service provider. Each client’s raw data is stored locally and not exchanged or transferred; instead, focused updates intended for immediate aggregation are used to achieve the learning objective. Since then, the topic has gathered much interest across many different disciplines and the realization that solving many of these interdisciplinary problems likely requires not just machine learning but techniques from distributed optimization, cryptography, security, differential privacy, fairness, compressed sensing, systems, information theory, statistics, and more. This monograph has contributions from leading experts across the disciplines, who describe the latest state-of-the art from their perspective. These contributions have been carefully curated into a comprehensive treatment that enables the reader to understand the work that has been done and get pointers to where effort is required to solve many of the problems before Federated Learning can become a reality in practical systems. Researchers working in the area of distributed systems will find this monograph an enlightening read that may inspire them to work on the many challenging issues that are outlined. This monograph will get the reader up to speed quickly and easily on what is likely to become an increasingly important topic: Federated Learning.

Foldamers: A Manifesto

Abstract: Nature relies on large molecules to carry out sophisticated chemical operations, such as catalysis, tight and specific binding, directed flow of electrons, or controlled crystallization of inorganic phases. The polymers entrusted with these crucial tasks, mostly proteins but sometimes RNA, are unique relative to other biological and synthetic polymers in that they adopt specific compact conformations that are thermodynamically and kinetically stable. These folding patterns generate “active sites” via precise three-dimensional arrangement of functional groups. In terms of covalent connectivity, the groups that comprise the active site are often widely spaced along the polymer backbone. The remarkable range of chemical capabilities that evolution has elicited from proteins suggests that it might be possible to design analogous capabilities into unnatural polymers that fold into compact and specific conformations. Since biological evolution has operated under many constraints, the functional properties of proteins and RNA should be viewed as merely exemplifying the potential of compactly folded polymers. The chemist’s domain includes all possible combinations of the elements, and the biological realm, vast and complex though it may be, is only a small part of that domain. Therefore, realization of the potential of folding polymers may be limited more by the human imagination than by physical barriers. I use the term “foldamer” to describe any polymer with a strong tendency to adopt a specific compact conformation. Among proteins, the term “compact” is associated with tertiary structure, and there is as yet no synthetic polymer that displays a specific tertiary structure. Protein tertiary structure arises from the assembly of elements of regular secondary structure (helices, sheets, and turns). The first step in foldamer design must therefore be to identify new backbones with well-defined secondary structural preferences. “Well-defined” in this case means that the conformational preference should be displayed in solution by oligomers of modest length, and I will designate as a foldamer any oligomer that meets this criterion. Within the past decade, a handful of research groups have described unnatural oligomers with interesting conformational propensities. The motivations behind such efforts are varied, but these studies suggest a collective, emerging realization that control over oligomer and polymer folding could lead to new types of molecules with useful properties. The purpose of this “manifesto” is to introduce a large audience to the broad research horizons offered by the concept of synthetic foldamers. The path to creating useful foldamers involves several daunting steps. (i) One must identify new polymeric backbones with suitable folding propensities. This goal includes developing a predictively useful understanding of the relationship between the repetitive features of monomer structure and conformational properties at the polymer level. (ii) One must endow the resulting foldamers with interesting chemical functions, by design, by randomization and screening (“evolution”), or by some combination of these two approaches. (iii) For technological utility, one must be able to produce a foldamer efficiently, which will generally include preparation of the constituent monomers in stereochemically pure form and optimization of heteropolymer synthesis. Each of these steps involves fascinating chemical challenges; the first step is the focus of this Account.