A recent line of work, starting with Beigman and Vohra [3] and Zadimoghaddam and Roth [28], has addressed the problem of learning a utility function from revealed preference data. The goal here is to make use of past data describing the purchases of a utility maximizing agent when faced with certain prices and budget constraints in order to produce a hypothesis function that can accurately forecast the future behavior of the agent.In this work we advance this line of work by providing sample complexity guarantees and efficient algorithms for a number of important classes. By drawing a connection to recent advances in multi-class learning, we provide a computationally efficient algorithm with tight sample complexity guarantees (Θ(d/ǫ) for the case of d goods) for learning linear utility functions under a linear price model. This solves an open question in Zadimoghaddam and Roth [28]. Our technique yields numerous generalizations including the ability to learn other well-studied classes of utility functions, to deal with a misspecified model, and with non-linear prices.

The Domain Adaptation problem in machine learning occurs when the distribution generating the test data differs from the one that generates the training data. A common approach to this issue is to train a standard learner for the learning task with the available training sample (generated by a distribution that is different from the test distribution). In this work we address this approach, investigating whether there exist successful learning methods for which learning of a target task can be achieved by substituting the standard target-distribution generated sample by a (possibly larger) sample generated by a different distribution without worsening the error guarantee on the learned classifier. We give a positive answer, showing that this is possible when using a Nearest Neighbor algorithm. We show this under the assumptions of covariate shift as well as a bound on the ratio of the probability weights between the source (training) and target (test) distribution. We further show that these assumptions are not always sufficient to allow such a replacement of the training sample: For proper learning, where the output classifier has to come from a predefined class, we prove that any learner needs access to data generated from the target distribution.

No abstract

Machine learning (ML) as a predictive methodology can potentially reduce the configuration cost and workload of inkjet printing. Inkjet printing has many advantages for additive manufacturing and printed electronics including low cost, scalability, non-contact printing and on-demand customization. Inkjet generates droplets with a piezoelectric dispenser controlled through frequency, voltage pulse and timing parameters. A major challenge is the design of jettable inks and the rapid optimization of stable jetting conditions whilst preventing common problems (no ejection, perturbation, satellite drop, multiple drops, drop breaking, nozzle clogging). Material consuming trial and error experiments are replaced here with a machine learning based jetting window. A data set of machine and material properties is created from literature and experimental data. After exploratory data analysis and feature identification, various (linear and non-linear) regression models are compared in detail. The models are trained on 80% of the data and root mean square error (RMSE) is calculated on 20% test data. Simple polynomial relationships between the input and output features yield coarse prediction. Instead, small ensembles of decision trees (boosted decision trees and random forests) have improved predictive power for drop velocity and radius with RMSE of 0.39 m/s and 2.21 µm respectively. The mean absolute percentage error (MAPE) is 3.87%. The models are validated with experimentally collected data for a novel ink where no data points with this ink were included in the training set. Additionally, several classification algorithms are utilized to categorize ink and printer parameters by jetting regime (‘single drop’, ‘multiple drops’, ‘no ejection’). Categorization and regression models are combined to improve overall model prediction. Machine learning enables efficient material and printing parameter selection speeding up the development of novel ink materials for printed electronics by eliminating jetting experiments that are money, time and material intensive.

Despite nearly ubiquitous access to wireless networks, many users still engage in risky behaviors, make bad choices, or are seemingly indifferent to the concerns that security and privacy researchers work diligently to address. At present, research on user attitudes toward security and privacy on public Wi-Fi networks is rare. This paper explores Wi-Fi security and privacy by analyzing users' current actions and reluctance to change. Through interviews and concrete demonstrations of vulnerability, we show that users make security choices based on (often mistaken) analogy to the physical world. Moreover, despite increased awareness of vulnerability, users remain ingenuous, failing to develop a realistic view of risk. We argue that our data present a picture of users engaged in a form of naïve security. We believe our results will be beneficial to researchers in the area of security-tool design, in particular with respect to better informing user choices.