In this paper we show how to formulate and solve robust portfolio selection problems. The objective of these formulations is systematically combat the sensitivity optimal statistical modeling errors in estimates relevant market parameters. We introduce “uncertainty structures” for parameters that problems corresponding uncertainty structures can be reformulated as secondorder cone programs and, therefore, computational effort required them comparable solving convex quadratic programs....

In this paper we propose a robust formulation for discrete time dynamic programming (DP). The objective of the is to systematically mitigate sensitivity DP optimal policy ambiguity in underlying transition probabilities. modeled by associating set conditional measures with each state-action pair. Consequently, has associated it. We prove that when certain “rectangularity” property, all main results finite and infinite horizon extend natural counterparts. discuss techniques from Nilim El...

Online Allocation of Reusable Resources: New Algorithms and Analytical Tools In the paper “Asymptotically Optimal Competitive Ratio for Resources,” authors develop novel algorithms analysis techniques online allocation reusable resources. Their approach leads to an algorithm with highest possible competitive ratio, a result that was previously out reach are used in classic settings which resources nonreusable. More generally, their LP-free is useful analyzing performance various other...

A flexible product is a menu of two or more alternatives products serving the same market. Purchasers are assigned to one at later date. Gallego and Phillips show that capacitated suppliers, such as airlines hotels, can potentially improve revenue by offering in addition traditional specific products. In this paper, we extend concept networks. We study network management problem with different settings: where demand for each independently exogenously generated; other driven consumer choice...

Systemic risk refers to the of collapse an entire complex system as a result actions taken by individual component entities or agents that comprise system. is issue great concern in modern financial markets well as, more broadly, management business and engineering systems. We propose axiomatic framework for measurement systemic based on simultaneous analysis outcomes across over scenarios nature. Our defines broad class measures accomodate rich set regulatory preferences. This general...

The authors propose a quantitative approach for describing entertainment products, in way that allows improving the predictive performance of consumer choice models these products. Their is based on media psychology literature, which suggests people’s consumption products influenced by psychological themes featured They classify basis “character strengths” taxonomy from positive literature (Peterson and Seligman 2004). develop natural language processing tool, guided latent Dirichlet...

We construct an economic framework for understanding the incentives of participants a permissioned blockchain supply chains and other related industries. Our study aims to determine whether adoption is socially beneficial such arises in equilibrium. find that reduces information asymmetry consumers, thereby enhancing consumer welfare. Consumer welfare gains can be sufficiently large beneficial; nonetheless, we does not arise This situation because costs are borne by manufacturers,...

One of the most interesting application scenarios in anomaly detection is when sequential data are targeted. For example, a safety-critical environment, it crucial to have an automatic system screen streaming gathered by monitoring sensors and report abnormal observations if detected real-time. Oftentimes, stakes much higher these potential anomalies intentional or goal-oriented. We propose end-to-end framework for using inverse reinforcement learning (IRL), whose objective determine...

We study the nonparametric least squares estimator (LSE) of a multivariate convex regression function. The LSE, given as solution to quadratic program with O(n2) linear constraints (n being sample size), is difficult compute for large problems. Exploiting problem specific structure, we propose scalable algorithmic framework based on augmented Lagrangian method LSE. develop novel approach obtain smooth approximations fitted (piecewise affine) LSE and provide formal bounds quality...

We consider an online assortment optimization problem where we have n substitutable products with fixed reusable capacities [Formula: see text]. In each period t, a user some preferences (potentially adversarially chosen) who offers subset of products, S t , from the set available arrives at seller’s platform. The selects product text] probability given by preference model and uses it for random number periods, text], that is distributed i.i.d. according to distribution depends only on j...

We examine a supply chain with single risk-averse manufacturer who purchases from suppliers and sells to consumers. Within this context, we focus on two channels that drive blockchain adoption by the manufacturer: risk aversion consumer information asymmetry. Regarding first channel, enables efficient tracing of defective products so can selectively recall rather than conducting full recall. This ability reduces involved in purchasing multiple thereby leads endogenously diversify across when...

Inferring connectivity in neuronal networks remains a key challenge statistical neuroscience. The “common input” problem presents major roadblock: it is difficult to reliably distinguish causal connections between pairs of observed neurons versus correlations induced by common input from unobserved neurons. Available techniques allow us simultaneously record, with sufficient temporal resolution, only small fraction the network. Consequently, naive estimators that neglect these effects are...

We introduce a novel linear bandit problem with partially observable features, resulting in partial reward information and spurious estimates. Without proper address for latent part, regret possibly grows linearly decision horizon $T$, as their influence on rewards are unknown. To tackle this, we propose analysis to handle the features an algorithm that achieves sublinear regret. The core of our involves (i) augmenting basis vectors orthogonal observed feature space, (ii) introducing...

We study a mechanism design problem in which an indivisible good is auctioned to multiple bidders for each of whom it has private value that unknown the seller and other bidders. The agents perceive ensemble all bidder values as random vector governed by ambiguous probability distribution, belongs commonly known ambiguity set. aims revenue-maximizing not only immunized against values, but also uncertainty about bidders’ attitude toward ambiguity. argue achieves this goal maximizing...

We propose a first-order augmented Lagrangian (FAL) algorithm for solving the basis pursuit problem. FAL computes solution to this problem by inexactly sequence of $\ell_1$-regularized least squares subproblems. These subproblems are solved using an infinite memory proximal gradient wherein each update reduces "shrinkage" or constrained "shrinkage." show that converges optimal whenever is unique, which case with very high probability compressed sensing problems. construct parameter such...

We propose a distributed first-order augmented Lagrangian (DFAL) algorithm to minimize the sum of composite convex functions, where each term in is private cost function belonging node, and only nodes connected by an edge can directly communicate with other. This optimization model abstracts number applications sensing machine learning. show that any limit point DFAL iterates optimal; for $ε>0$, $ε$-optimal $ε$-feasible solution be computed within $\mathcal{O}(\log(ε^{-1}))$ iterations,...