Search

Federated optimization, wherein several agents in a network collaborate with a central server to achieve optimal social cost over the network with no requirement for exchanging information among agents, has attracted significant interest from the research community. In this context, agents demand resources based on their local computation. Due to the exchange of optimization parameters such as states, constraints, or objective functions with a central server, an adversary may infer sensitive information of agents. We develop a differentially-private additive-increase and multiplicative-decrease algorithm to allocate multiple divisible shared heterogeneous resources to agents in a network. The developed algorithm provides a differential privacy guarantee to each agent in the network. The algorithm does not require inter-agent communication, and the agents do not need to share their cost function or their derivatives with other agents or a central server; however, they share their allocation states with a central server that keeps track of the aggregate consumption of resources. The algorithm incurs very little communication overhead; for $m$ heterogeneous resources in the system, the asymptotic upper bound on the communication complexity is $\mathcal {O}(m)$ bits at a time step. Furthermore, if the algorithm converges in $K$ time steps, then the upper bound communication complexity will be $\mathcal {O}(mK)$ bits. The algorithm can find applications in several areas, including smart cities, smart energy systems, resource management in the sixth generation (6G) wireless networks with privacy guarantees, etc. We present experimental results to check the efficacy of the algorithm. Furthermore, we present empirical analyses for the trade-off between privacy and algorithm efficiency.

As usage of artificial intelligence (AI) technologies across industries increases, there is a growing need for creating large marketplaces to host and transact good-quality data sets to train AI algorithms. Our study analyzes the characteristics of such an oligopsony crowdsourced AI Marketplace (AIM) that has a large number of producers and few consumers who transact data sets as per their expectations of price and quality. Using agent-based modeling (ABM), we incorporate heterogeneity in agent attributes and self-learning by the agents that are reflective of real-world marketplaces. Our research augments the existing studies on the effect of and reputation systems in such market places. Extensive simulations using ABM indicate that ratings of the data sets as a feedback mechanism plays an important role in improving the quality of said data sets, and hence the reputations of producers. While such marketplaces are evolving, regulators have started enacting varying rules to oversee the appropriate functioning of such marketplaces, to minimize market distortions. In one of the first such studies, we integrate regulatory interventions in a marketplace model to analyze the impacts of various types of regulations on the functioning of an AIM. Our results indicate that very stringent regulatory measures negatively affect the production of quality data sets in the marketplace. On the other hand, regulatory oversight along with a ratings-based feedback mechanism improves the functioning of an AIM, and hence is recommended for governments and policy makers to adopt.

In toy environments like video games, a reinforcement learning agent is deployed and operates within the same state space in which it was trained. However, in robotics applications such as industrial systems or autonomous vehicles, this cannot be guaranteed. A robot can be pushed out of its training space by some unforeseen perturbation, which may cause it to go into an unknown state from which it has not been trained to move towards its goal. While most prior work in the area of RL safety focuses on ensuring safety in the training phase, this paper focuses on ensuring the safe deployment of a robot that has already been trained to operate within a safe space. This work defines a condition on the state and action spaces, that if satisfied, guarantees the robot’s recovery to safety independently. We also propose a strategy and design that facilitate this recovery within a finite number of steps after perturbation. This is implemented and tested against a standard RL model, and the results indicate a significant improvement in performance.

Early electrophysiological brain oscillations recorded in preterm babies and newborn rodents are initially mostly driven by bottom-up sensorimotor activity and only later can detach from external inputs. This is a hallmark of most developing brain areas, including the hippocampus, which, in the adult brain, functions in integrating external inputs onto internal dynamics. Such developmental disengagement from external inputs is likely a fundamental step for the proper development of cognitive internal models. Despite its importance, the developmental timeline and circuit basis for this disengagement remain unknown. To address this issue, we have investigated the daily evolution of CA1 dynamics and underlying circuits during the first two postnatal weeks of mouse development using two-photon calcium imaging in non-anesthetized pups. We show that the first postnatal week ends with an abrupt shift in the representation of self-motion in CA1. Indeed, most CA1 pyramidal cells switch from activated to inhibited by self-generated movements at the end of the first postnatal week, whereas the majority of GABAergic neurons remain positively modulated throughout this period. This rapid switch occurs within 2 days and follows the rapid anatomical and functional surge of local somatic GABAergic innervation. The observed change in dynamics is consistent with a two-population model undergoing a strengthening of inhibition. We propose that this abrupt developmental transition inaugurates the emergence of internal hippocampal dynamics.

Charging station (CS) planning for electric vehicles (EVs) for a region has become an important concern for urban planners and the public alike to improve the adoption of EVs. Two major problems comprising this research area are: (i) the EV charging station placement (EVCSP) problem, and (ii) the CS need estimation problem for a region. In this work, different explainable solutions based on machine learning (ML) and simulation were investigated by incorporating quantitative and qualitative metrics. The solutions were compared with traditional approaches using a real CS area of Austin and a greenfield area of Bengaluru. For EVCSP, a different class of clustering solutions, i.e., mean-based, density-based, spectrum- or eigenvalues-based, and Gaussian distribution were evaluated. Different perspectives, such as the urban planner perspective, i.e., the clustering efficiency, and the EV owner perspective, i.e., an acceptable distance to the nearest CS, were considered. For the CS need estimation, ML solutions based on quadratic regression and simulations were evaluated. Using our CS planning methods urban planners can make better CS placement decisions and can estimate CS needs for the present and the future.

Data-centric models of COVID-19 have been attempted, but have certain limitations. In this work, we propose an agent-based model of the epidemic in a confined space of agents representing humans. An extension to the SEIR model allows us to consider the difference between the appearance (black-box view) of the spread of disease and the real situation (glass-box view). Our model allows for simulations of lockdowns, social distancing, personal hygiene, quarantine, and hospitalization, with further considerations of different parameters, such as the extent to which hygiene and social distancing are observed in a population. Our results provide qualitative indications of the effects of various policies and parameters, for instance, that lockdowns by themselves are extremely unlikely to bring an end to an epidemic and may indeed make things worse, that social distancing is more important than personal hygiene, and that the growth of infection is significantly reduced for moderately high levels of social distancing and hygiene, even in the absence of herd immunity.

We propose a new class of mathematical structures called (m,n)-semirings (which generalize the usual semirings) and describe their basic properties. We define partial ordering and generalize the concepts of congruence, homomorphism, and so forth, for (m,n)-semirings. Following earlier work by Rao (2008), we consider systems made up of several components whose failures may cause them to fail and represent the set of such systems algebraically as an (m,n)-semiring. Based on the characteristics of these components, we present a formalism to compare the fault-tolerance behavior of two systems using our framework of a partially ordered (m,n)-semiring.