Polynomial regression under arbitrary product distributions

In recent work, Kalai, Klivans, Mansour, and Servedio (2005) studied a variant of the “Low-Degree (Fourier) Algorithm” for learning under the uniform probability distribution on {0,1} ⁿ . They showed that the L ₁ polynomial regression algorithm yields agnostic (tolerant to arbitrary noise) learning algorithms with respect to the class of threshold functions—under certain restricted instance distributions, including uniform on {0,1} ⁿ and Gaussian on ? ⁿ . In this work we show how all learning results based on the Low-Degree Algorithm can be generalized to give almost identical agnostic guarantees under arbitrary product distributions on instance spaces X ₁×???×X _n . We also extend these results to learning under mixtures of product distributions. The main technical innovation is the use of (Hoeffding) orthogonal decomposition and the extension of the “noise sensitivity method” to arbitrary product spaces. In particular, we give a very simple proof that threshold functions over arbitrary product spaces have δ-noise sensitivity $O(\sqrt{\delta})$ , resolving an open problem suggested by Peres (2004).     

Action selection and task sequence learning for hybrid dynamical cognitive agents

As a foundation for action selection and task-sequencing intelligence, the reactive and deliberative subsystems of a hybrid agent can be unified by a single, shared representation of intention. In this paper, we summarize a framework for hybrid dynamical cognitive agents (HDCAs) that incorporates a representation of dynamical intention into both reactive and deliberative structures of a hybrid dynamical system model, and we present methods for learning in these intention-guided agents. The HDCA framework is based on ideas from spreading activation models and belief–desire–intention (BDI) models. Intentions and other cognitive elements are represented as interconnected, continuously varying quantities, employed by both reactive and deliberative processes. HDCA learning methods—such as Hebbian strengthening of links between co-active elements, and belief–intention learning of task-specific relationships—modify interconnections among cognitive elements, extending the benefits of reactive intelligence by enhancing high-level task sequencing without additional reliance on or modification of deliberation. We also present demonstrations of simulated robots that learned geographic and domain-specific task relationships in an office environment.     

Marine Geospatial Ecology Tools: An integrated framework for ecological geoprocessing with ArcGIS,Python, R,MATLAB, and C++

With the arrival of GPS, satellite remote sensing, and personal computers, the last two decades have witnessed rapid advances in the field of spatially-explicit marine ecological modeling. But with this innovation has come complexity. To keep up, ecologists must master multiple specialized software packages, such as ArcGIS for display and manipulation of geospatial data, R for statistical analysis, and MATLAB for matrix processing. This requires a costly investment of time and energy learning computer programming, a high hurdle for many ecologists. To provide easier access to advanced analytic methods, we developed Marine Geospatial Ecology Tools (MGET), an extensible collection of powerful, easy-to-use, open-source geoprocessing tools that ecologists can invoke from ArcGIS without resorting to computer programming. Internally, MGET integrates Python, R, MATLAB, and C++, bringing the power of these specialized platforms to tool developers without requiring developers to orchestrate the interoperability between them.In this paper, we describe MGET’s software architecture and the tools in the collection. Next, we present an example application: a habitat model for Atlantic spotted dolphin (Stenella frontalis) that predicts dolphin presence using a statistical model fitted with oceanographic predictor variables. We conclude by discussing the lessons we learned engineering a highly integrated tool framework.     

Search advertising placement strategy: Exploring the efficacy of the conventional wisdom

Search advertising is one of the most important forms of electronic commerce. While click-through rates are considered a key measure of search advertising effectiveness by search providers, brand attention can also be a valuable objective. Our paper reports on an experiment that investigated how search advertisement placement affected search users’ brand recall and recognition. The results showed that semantically associated search ads generated significantly higher levels of brand attention than contextually associated ones. Significant interaction effects were found among search ad position, keyword association, and search result quality. Implications for decision makers are discussed.     

Effect of La2CuO4 electrode area on potentiometric NOx sensor response and its implications on sensing mechanism

The intent of this work is to look at the effects of varying the La₂CuO₄ electrode area and the asymmetry between the sensing and counter electrode in a solid state potentiometric sensor with respect to NO_x sensitivity. NO₂ sensitivity was observed at 500-600 °C with a maximum sensitivity of ∼22 mV/decade [NO₂] observed at 500 °C for the sensor with a La₂CuO₄ electrode area of ∼30 mm². The relationship between NO₂ sensitivity and area is nearly parabolic at 500 °C, decreases linearly with increasing electrode area at 600 °C, and was a mixture of parabolic and linear behavior 550 °C. NO sensitivity varied non-linearly with electrode area with a minima (maximum sensitivity) of ∼−22 mV/decade [NO] at 450 °C for the sensor with a La₂CuO₄ electrode area of 16 mm². The behavior at 400 °C was similar to that of 450 °C, but with smaller sensitivities due to a saturation effect. At 500 °C, NO sensitivity decreases linearly with area.We also used electrochemical impedance spectroscopy (EIS) to investigate the electrochemical processes that are affected when the sensing electrode area is changed. Changes in impedance with exposure to NO_x were attributed to either changes in La₂CuO₄ conductivity due to gas adsorption (high frequency impedance) or electrocatalysis occurring at the electrode/electrolyte interface (total electrode impedance). NO₂ caused a decrease in high frequency impedance while NO caused an increase. In contrast, NO₂ and NO both caused a decrease in the total electrode impedance. The effect of area on both the potentiometric and impedance responses show relationships that can be explained through the mechanistic contributions included in differential electrode equilibria.     

A photoluminescence-based quantum semiconductor biosensor for rapid in situ detection of Escherichia coli

This work describes a novel method of detecting Escherichia coli using photoluminescence (PL) emission from III–V quantum semiconductor (QS) devices functionalized with two different antibody-based architectures. The first approach employed self-assembled monolayers of biotinylated polyethylene glycol thiols to immobilize biotinylated antibody via neutravidin. In the second approach, we used QS microstructures coated with a thin layer of Si₃N₄ allowing direct functionalization with E. coli antibodies through hydrofluoric acid etching and glutaraldehyde-based reticulation. Atomic force, optical and fluorescence microscopy measurements were used to assess the immobilization process. Depending on the biosensing architecture, density of the immobilized bacteria was observed in the range of 0.5–0.7 bacteria/100 μm². The detection of E. coli at 10⁴ CFU/ml was achieved within less than 120 min of the bacteria exposure. It is expected that an even better sensitivity threshold could be achieved following further optimization of the method.     

Setpoint regulation for stochastically interacting robots

We present an integral feedback controller that regulates the average copy number of an assembly in a system of stochastically interacting robots. The mathematical model for these robots is a tunable reaction network, which makes this approach applicable to a large class of other systems, including ones that exhibit stochastic self-assembly at various length scales. We prove that this controller works for a range of setpoints and how to compute this range both analytically and experimentally. Finally, we demonstrate these ideas on a physical testbed.