Overbooking with bounded loss Freund, Daniel; Zhao, Jiayu Kamessi
arXiv (Cornell University),
04/2022
Paper, Journal Article
Odprti dostop
We study a classical problem in revenue management: quantity-based single-resource revenue management with no-shows. In this problem, a firm observes a sequence of \(T\) customers requesting a ...service. Each arrival is drawn independently from a known distribution of \(k\) different types, and the firm needs to decide irrevocably whether to accept or reject requests in an online fashion. The firm has a capacity of resources \(B\), and wants to maximize its profit. Each accepted service request yields a type-dependent revenue and has a type-dependent probability of requiring a resource once all arrivals have occurred (or, be a no-show). If the number of accepted arrivals that require a resource at the end of the horizon is greater than \(B\), the firm needs to pay a fixed compensation for each service request that it cannot fulfill. With a clairvoyant, that knows all arrivals ahead of time, as a benchmark, we provide an algorithm with a uniform additive loss bound, i.e., its expected loss is independent of \(T\). This improves upon prior works achieving \(\Omega(\sqrt{T})\) guarantees.
Abstract Background aims It is unclear whether the plastic-adherent multipotent mesenchymal stromal cells (MSC) isolated from human bone marrow (BM) represent a uniform cell population or are ...heterogeneous in terms of cell-surface constituents and hence functionality. Methods We investigated the expression profile of certain biofunctional lipids by plastic-adherent MSC, focusing particularly on two membrane microdomain (lipid raft)-associated monosialogangliosides, GM1 and GM3, using indirect confocal laser scanning fluorescence microscopy and flow cytometry. Results Phenotypically, we observed a differential expression where certain MSC subsets exhibited GM1, GM3 or both at the plasma membrane. Furthermore, disialoganglioside GD2 detection increased the complexity of the expression patterns, giving rise to seven identifiable cell phenotypes. Variation of standard culture conditions, such as the number of cell passage and period in culture, as well as donors, did not influence the heterologous ganglioside expression profile. In contrast, the binding of various lectins appeared homogeneous throughout the MSC population, indicating that the general glycosylation pattern remained common. Morphologically, the expression of a given ganglioside-based phenotype was not related to a cell with particular size or shape. Interestingly, a segregation of GM1 and GM3 clusters was observed, GM3 being mostly excluded from the highly curved plasma membrane protrusions. Conclusions These data highlight the phenotypic heterogeneity of plastic-adherent MSC in terms of certain lipid constituents of the plasma membrane, and the presence and/or absence of distinct ganglioside-based membrane microdomains suggest their potential functional diversity.
Queueing systems are widely applicable stochastic models with use cases in communication networks, healthcare, service systems, etc. Although their optimal control has been extensively studied, most ...existing approaches assume perfect knowledge of the system parameters. Of course, this assumption rarely holds in practice where there is parameter uncertainty, thus motivating a recent line of work on bandit learning for queueing systems. This nascent stream of research focuses on the asymptotic performance of the proposed algorithms. In this paper, we argue that an asymptotic metric, which focuses on late-stage performance, is insufficient to capture the intrinsic statistical complexity of learning in queueing systems which typically occurs in the early stage. Instead, we propose the Cost of Learning in Queueing (CLQ), a new metric that quantifies the maximum increase in time-averaged queue length caused by parameter uncertainty. We characterize the CLQ of a single queue multi-server system, and then extend these results to multi-queue multi-server systems and networks of queues. In establishing our results, we propose a unified analysis framework for CLQ that bridges Lyapunov and bandit analysis, provides guarantees for a wide range of algorithms, and could be of independent interest.
Establishment of a defined cell culture system that facilitates ex vivo expansion of isolated hematopoietic stem and progenitor cells (HSPCs) is a crucial issue in hematology and stem cell ...transplantation. Here we have evaluated the capacity of primary human multipotent mesenchymal stromal cells (MSCs) to support the ex vivo expansion of peripheral CD34(+)-enriched HSPCs. We observed that HSPCs co-cultured on MSCs showed a substantially higher total expansion rate compared to those growing without. Moreover, in addition to the expansion of CD34(+)CD133(+) and CD34(+)CD133(-) cells, a third population of CD133(+)CD34(-) stem cells became detectable after expansion. Direct contact between HSPCs and the feeder layer appears beneficial for the expansion of HSPCs harboring CD133(+) phenotype, i.e., CD34(+)CD133(+) and CD133(+)CD34(-), in contrast to CD34(+)CD133(-) cells. Interestingly, electron microscopy and immunofluorescence analyses revealed that adherent HSPCs display various morphologies; they are either round with, in some cases, the appearance of a microvillar pole or exhibit several distinct types of plasma membrane protrusions such as lamellipodium and magnupodium. CD133 is selectively concentrated therein, whereas CD34 is randomly distributed over the entire surface of HSPCs. Together, this co-culture offers a unique experimental system to further characterize the biology and role of markers of rare stem cell populations.
We study decentralized multi-agent learning in bipartite queueing systems, a standard model for service systems. In particular, N agents request service from K servers in a fully decentralized way, ...i.e, by running the same algorithm without communication. Previous decentralized algorithms are restricted to symmetric systems, have performance that is degrading exponentially in the number of servers, require communication through shared randomness and unique agent identities, and are computationally demanding. In contrast, we provide a simple learning algorithm that, when run decentrally by each agent, leads the queueing system to have efficient performance in general asymmetric bipartite queueing systems while also having additional robustness properties. Along the way, we provide the first provably efficient UCB-based algorithm for the centralized case of the problem.
Effective load balancing is at the heart of many applications in operations. Often tackled via the balls-into-bins paradigm, seminal results have shown that a limited amount of flexibility goes a ...long way in order to maintain (approximately) balanced loads throughout the decision-making horizon. This paper is motivated by the fact that balance across time is too stringent a requirement for some applications; rather, the only desideratum is approximate balance at the end of the horizon. In this work we design ``limited-flexibility'' algorithms for three instantiations of the end-of-horizon balance problem: the balls-into-bins problem, opaque selling strategies for inventory management, and parcel delivery for e-commerce fulfillment. For the balls-into-bins model, we show that a simple policy which begins exerting flexibility toward the end of the time horizon (i.e., when \(\Theta\left(\sqrt{T\log T}\right)\) periods remain), suffices to achieve an approximately balanced load (i.e., a maximum load within \({O}(1)\) of the average load). Moreover, with just a small amount of adaptivity, a threshold policy achieves the same result, while only exerting flexibility in \({O}\left(\sqrt{T}\right)\) periods, matching a natural lower bound. We then adapt these algorithms to develop order-wise optimal policies for the opaque selling problem. Finally, we show via a data-driven case study that the adaptive policy designed for the balls-into-bins model can be modified to (i) achieve approximate balance at the end of the horizon and (ii) yield significant cost savings relative to policies which either never exert flexibility, or exert flexibility aggressively enough to achieve anytime balance. The unifying motivation behind our algorithms is the observation that exerting flexibility at the beginning of the horizon is likely wasted when system balance is only evaluated at the end.
Overbooking with Bounded Loss Freund, Daniel; Zhao, Jiayu (Kamessi)
Proceedings of the 22nd ACM Conference on Economics and Computation,
07/2021
Conference Proceeding
Odprti dostop
We study a classical problem in revenue management: quantity-based single-resource revenue management with no-shows. In this problem, a firm observes a sequence of T customers requesting a service. ...Each arrival is drawn independently from a known distribution of k different types, and the firm needs to decide irrevocably whether to accept or reject requests in an online fashion. The firm has a capacity of resources B, and wants to maximize its profit. Each accepted service request yields a type-dependent revenue and has a type-dependent probability of requiring a resource once all arrivals have occurred (or, be a no-show). If the number of accepted arrivals that require a resource at the end of the horizon is greater than B, the firm needs to pay a fixed compensation for each service request that it cannot fulfill. With a clairvoyant, that knows all arrivals ahead of time, as a benchmark, we provide an algorithm with a uniform additive loss bound, i.e., its expected loss is independent of B and T. This improves upon prior works achieving $Ømega(\sqrtT )$ guarantees.
During an online search, I recently came across an Employee Benefit News opinion article from last year in which Anne Burkett, national HR technology practice leader for USI Insurance Services, made ...the case against a no-cost employee benefits administration solution. “An employer that relies on a free benefits administration solution also may struggle to sever ties with the owner of the technology if a decision is made to partner with other carriers or even change consultants in the future. Since a consultant or carrier in such an arrangement owns the technology, the company can lose electronic access to their benefits administration solution and all the current and historical data, requiring an employer to collect all new data from employees,” Burkett says. “Automated data file feeds between payroll/HR systems and the free benefits administration solution are not widely available or often limited, which can impact data integrity,” Burkett says.
During an online search, I recently came across an Employee Benefit News opinion article from last year in which Anne Burkett, national HR technology practice leader for USI Insurance Services, made ...the case against a no-cost employee benefits administration solution. “An employer that relies on a free benefits administration solution also may struggle to sever ties with the owner of the technology if a decision is made to partner with other carriers or even change consultants in the future. Since a consultant or carrier in such an arrangement owns the technology, the company can lose electronic access to their benefits administration solution and all the current and historical data, requiring an employer to collect all new data from employees,” Burkett says. “Automated data file feeds between payroll/HR systems and the free benefits administration solution are not widely available or often limited, which can impact data integrity,” Burkett says.
Due to the fact that electric vehicles have not broadly entered the vehicle market there are many attempts to convince producers to integrate technologies that utilise embedded batteries for purposes ...different from driving. The vehicle-to-grid technology, for instance, literally turns electric vehicles into a mobile battery, enabling new areas of applications (e.g., to provide regulatory energy, to do grid-load balancing, or to buffer surpluses of energy) and business perspectives. Utilising a vehicle’s battery, however is not without a price—in this case: the driver’s mobility. Given this dependency, it is interesting that most available works consider the application of electric vehicles for energy and grid-related problems in isolation, that is, detached from mobility-related issues. The
distributed artificial intelligence laboratory
, or
DAI-Lab
, is a third-party funded research lab at Technische Universität Berlin and integrates the chair for
agent technologies in business applications and telecommunication
. The DAI-Lab has engaged in a large number of both, past and upcoming projects concerned with two aspects of managing electric vehicles, namely: energy and mobility. This article aims to summarise experiences that were collected during the last years and to present developed solutions which consider energy and mobility-related problems jointly.