Classifiers can be trained with data-dependent constraints to satisfy fairness goals, reduce churn, achieve a targeted false positive rate, or other policy goals. We study the generalization performance for such constrained optimization problems, in terms of how well are satisfied at evaluation time, given that they training time. To improve performance, we frame problem as two-player game where one player optimizes model parameters on dataset, and enforces an independent validation dataset....

We propose learning deep models that are monotonic with respect to a user-specified set of inputs by alternating layers linear embeddings, ensembles lattices, and calibrators (piecewise functions), appropriate constraints for monotonicity, jointly training the resulting network. implement projections new computational graph nodes in TensorFlow use ADAM optimizer batched stochastic gradients. Experiments on benchmark real-world datasets show six-layer lattice networks achieve state-of-the art...

We reverse-engineer the frequency dynamics with general primary control and show that it is a distributed algorithm to solve well-defined optimization problem. further investigate role of deadband in control, if aggregated uncontrolled load deviation nonzero frequencies will be synchronized, however zero may oscillate but within deadband. The model does not only provide way characterize equilibrium establish convergence dynamics, also suggests principled engineer control. By leveraging...

We show that many machine learning goals, such as improved fairness metrics, can be expressed constraints on the model's predictions, which we call rate constraints. study problem of training non-convex models subject to these (or any and non-differentiable constraints). In setting, standard approach Lagrange multipliers may fail. Furthermore, if are non-differentiable, then one cannot optimize Lagrangian with gradient-based methods. To solve issues, introduce proxy-Lagrangian formulation....

In this paper, we propose a coordinated localization algorithm for mobile sensor networks with camera sensors to operate under Global Positioning System (GPS) denied areas or indoor environments. Mobile robots are partitioned into two groups. One group moves within the field of views remaining stationary robots. The moving tracked by and their trajectories used as spatiotemporal features. From these features, relative poses computed using multiview geometry is localized respect reference...

The optimal power flow (OPF) problem is fundamental to system planing and operation. It a non-convex optimization the semidefinite programming (SDP) relaxation has been proposed recently. However, SDP may give an infeasible solution original OPF problem. In this paper, we apply alternating direction method of multiplier recover feasible when Specifically, procedure iterates between convex problem, with rank constraint. By exploiting special structure constraint, obtain closed form based on...

This paper introduces the notion of localizable distributed systems. These are systems for which a controller exists that limits effect each disturbance to some local subset entire plant, akin spatio-temporal dead-beat control. We characterize localizing state-feedback in terms feasibility set linear equations. then show when feasible solution exists, it can be found way, and used localized synthesis implementation controllers lead desired closed loop response. In particular, by allowing...

The weighted sum-rate maximization in a general multiple-input multiple-output (MIMO) interference network has known to be challenging non-convex problem, mainly due the between different links. In this paper, by exploring special structure of function being difference concave functions, we apply convex-concave procedure handle non-convexity. With introduction certain damping term, establish monotonie convergence proposed algorithm. Numerical examples show that introduced term slows down our...

This paper presents a scalable method to design large-scale Kalman-like filters for class of linear systems. In particular, we consider systems which both the propagation dynamics through plant and exchange information between estimators/sensors is subject delays. Under suitable assumptions on these delays, our proposed filter has following desirable properties: (1) each local estimator only needs collect within localized region estimate its state, (2) can be designed by solving optimization...

In this paper, we consider the signal-anticipating behavior in local volt/var control distribution systems. We define a voltage game, and show that is best response algorithm of game. further game has unique Nash equilibrium, characterize it as optimum global optimization problem, establish its asymptotic stability. then introduce notion price (PoSA) to impact control, use gap cost between network equilibrium signal-taking metric for PoSA. how PoSA scales with size, topology, heterogeneity...

In this paper, we discuss a large-scale fleet management problem in multi-objective setting. We aim to seek receding horizon taxi dispatch solution that serves as many ride requests possible while minimizing the cost of relocating vehicles. To obtain desired solution, first convert into network flow problem, which can be solved using classical minimum maximum (MCMF) algorithm. show obtained MCMF algorithm is integer-valued; thus, it does not require any additional rounding procedure may...

MIMO interference network optimization is important for increasingly crowded wireless communication networks. This paper presents a new algorithm, named Dual Link Algorithm, weighted sum-rate maximization where the efficiently managed. We consider general channels with Gaussian input and total power constraint. Two of previous state-of-the-art algorithms are WMMSE algorithm polite water-filling (PWF) algorithm. The provably convergent, while PWF takes advantage optimal transmit signal...

This paper presents a new, elementary proof for the Generalized Kalman-Yakubovich-Popov lemma based on Lagrangian duality, and new sufficient Linear Matrix Inequality test bandpass type frequency bound. Numerical experiments have failed to find gap, so it is possible that LMI may be necessary.

Distributed systems are comprised of multiple subsystems that interact in two distinct ways: (1) physical interactions and (2) cyber interactions; i.e. sensors, actuators computers controlling these subsystems, the network over which they communicate. A broad class cyber-physical (CPS) described by such interactions, as smart grid, platoons autonomous vehicles sensorimotor system. This paper will survey recent progress developing a coherent mathematical framework describes rich CPS “design...

This paper proposes a direct, and simple approach to the H infinity norm calculation in more general settings. In contrast method based on Kalman-Yakubovich-Popov lemma, our does not require controllability assumption, returns sinusoidal input that achieves of system including its frequency. addition, using semidefinite programming duality, we present new proof Kalman- Yakubovich-Popov make connection between strong duality controllability. Finally, generalize towards generalized which...

In this paper, we propose a cooperative localization algorithm for mobile sensor networks with camera sensors to operate under GPS denied areas or indoor environments. Mobile robots are partitioned into two groups. One group moves within the field of views remaining stationary robots. The moving tracked by and their trajectories used as spatio-temporal features. From these features, relative poses computed using multi-view geometry. order provide respect reference coordinate system, take...

This paper reformulates and streamlines the core tools of robust stability performance for LTI systems using now-standard methods in convex optimization. In particular, robustness analysis can be formulated directly as a primal (semidefinite program or SDP) optimization problem sets Gramians whose closure is semidefinite cone. allows various constraints such structured uncertainty to included directly, worst-case disturbances perturbations constructed from variables. Well known results KYP...

Short-term demand forecasting models commonly combine convolutional and recurrent layers to extract complex spatiotemporal patterns in data. Long-term histories are also used consider periodicity seasonality as time series In this study, we propose an efficient architecture, Temporal-Guided Network (TGNet), which utilizes graph networks temporal-guided embedding. Graph invariant features permutations of adjacent regions instead layers. Temporal-guided embedding explicitly learns temporal...

We take a new perspective on the weighted sum-rate maximization in multiple-input multiple-output (MIMO) interference networks, by formulating an equivalent max-min problem. This seemingly trivial reformulation has significant implications: Lagrangian duality of problem provides elegant way to establish between network and its reciprocal when such exists, more importantly, suggests novel iterative minimax algorithm for maximization. Moreover, design convergence proof use only general convex...

We consider the signal-anticipating behavior in local Volt/Var control for distribution systems. Such a makes interaction among nodes game. characterize Nash equilibrium of game as optimum global optimization problem and establish its asymptotic stability. also show that voltage has less restrictive convergence condition than signal-taking control. then introduce notion Price Signal-Anticipation (PoSA) to impact control, use gap cost between network metric PoSA. how PoSA scales with size,...