skip to main content

Title: One for One, or All for All: Equilibria and Optimality of Collaboration in Federated Learning
In recent years, federated learning has been embraced as an approach for bringing about collaboration across large populations of learning agents. However, little is known about how collaboration protocols should take agents’ incentives into account when allocating individual resources for communal learning in order to maintain such collaborations. Inspired by game theoretic notions, this paper introduces a framework for incentive-aware learning and data sharing in federated learning. Our stable and envy-free equilibria capture notions of collaboration in the presence of agents interested in meeting their learning objectives while keeping their own sample collection burden low. For example, in an envy-free equilibrium, no agent would wish to swap their sampling burden with any other agent and in a stable equilibrium, no agent would wish to unilaterally reduce their sampling burden. In addition to formalizing this framework, our contributions include characterizing the structural properties of such equilibria, proving when they exist, and showing how they can be computed. Furthermore, we compare the sample complexity of incentive-aware collaboration with that of optimal collaboration when one ignores agents’ incentives.
Authors:
; ; ;
Award ID(s):
1815011 1733556
Publication Date:
NSF-PAR ID:
10283402
Journal Name:
International Conference on Machine Learning
Volume:
38
Page Range or eLocation-ID:
1005-1014
Sponsoring Org:
National Science Foundation
More Like this
  1. Networked public goods games model scenarios in which self-interested agents decide whether or how much to invest in an action that benefits not only themselves, but also their network neighbors. Examples include vaccination, security investment, and crime reporting. While every agent's utility is increasing in their neighbors' joint investment, the specific form can vary widely depending on the scenario. A principal, such as a policymaker, may wish to induce large investment from the agents. Besides direct incentives, an important lever here is the network structure itself: by adding and removing edges, for example, through community meetings, the principal can changemore »the nature of the utility functions, resulting in different, and perhaps socially preferable, equilibrium outcomes. We initiate an algorithmic study of targeted network modifications with the goal of inducing equilibria of a particular form. We study this question for a variety of equilibrium forms (induce all agents to invest, at least a given set S, exactly a given set S, at least k agents), and for a variety of utility functions. While we show that the problem is NP-complete for a number of these scenarios, we exhibit a broad array of scenarios in which the problem can be solved in polynomial time by non-trivial reductions to (minimum-cost) matching problems.« less
  2. We initiate the study of information elicitation mechanisms for a crowd containing both self-interested agents, who respond to incentives, and adversarial agents, who may collude to disrupt the system. Our mechanisms work in the peer prediction setting where ground truth need not be accessible to the mechanism or even exist. We provide a meta-mechanism that reduces the design of peer prediction mechanisms to a related robust learning problem. The resulting mechanisms are ϵ-informed truthful, which means truth-telling is the highest paid ϵ-Bayesian Nash equilibrium (up to ϵ-error) and pays strictly more than uninformative equilibria. The value of ϵ depends onmore »the properties of robust learning algorithm, and typically limits to 0 as the number of tasks and agents increase. We show how to use our meta-mechanism to design mechanisms with provable guarantees in two important crowdsourcing settings even when some agents are self-interested and others are adversarial.« less
  3. It takes great effort to manually or semi-automatically convert free-text phenotype narratives (e.g., morphological descriptions in taxonomic works) to a computable format before they can be used in large-scale analyses. We argue that neither a manual curation approach nor an information extraction approach based on machine learning is a sustainable solution to produce computable phenotypic data that are FAIR (Findable, Accessible, Interoperable, Reusable) (Wilkinson et al. 2016). This is because these approaches do not scale to all biodiversity, and they do not stop the publication of free-text phenotypes that would need post-publication curation. In addition, both manual and machine learningmore »approaches face great challenges: the problem of inter-curator variation (curators interpret/convert a phenotype differently from each other) in manual curation, and keywords to ontology concept translation in automated information extraction, make it difficult for either approach to produce data that are truly FAIR. Our empirical studies show that inter-curator variation in translating phenotype characters to Entity-Quality statements (Mabee et al. 2007) is as high as 40% even within a single project. With this level of variation, curated data integrated from multiple curation projects may still not be FAIR. The key causes of this variation have been identified as semantic vagueness in original phenotype descriptions and difficulties in using standardized vocabularies (ontologies). We argue that the authors describing characters are the key to the solution. Given the right tools and appropriate attribution, the authors should be in charge of developing a project's semantics and ontology. This will speed up ontology development and improve the semantic clarity of the descriptions from the moment of publication. In this presentation, we will introduce the Platform for Author-Driven Computable Data and Ontology Production for Taxonomists, which consists of three components: a web-based, ontology-aware software application called 'Character Recorder,' which features a spreadsheet as the data entry platform and provides authors with the flexibility of using their preferred terminology in recording characters for a set of specimens (this application also facilitates semantic clarity and consistency across species descriptions); a set of services that produce RDF graph data, collects terms added by authors, detects potential conflicts between terms, dispatches conflicts to the third component and updates the ontology with resolutions; and an Android mobile application, 'Conflict Resolver,' which displays ontological conflicts and accepts solutions proposed by multiple experts. a web-based, ontology-aware software application called 'Character Recorder,' which features a spreadsheet as the data entry platform and provides authors with the flexibility of using their preferred terminology in recording characters for a set of specimens (this application also facilitates semantic clarity and consistency across species descriptions); a set of services that produce RDF graph data, collects terms added by authors, detects potential conflicts between terms, dispatches conflicts to the third component and updates the ontology with resolutions; and an Android mobile application, 'Conflict Resolver,' which displays ontological conflicts and accepts solutions proposed by multiple experts. Fig. 1 shows the system diagram of the platform. The presentation will consist of: a report on the findings from a recent survey of 90+ participants on the need for a tool like Character Recorder; a methods section that describes how we provide semantics to an existing vocabulary of quantitative characters through a set of properties that explain where and how a measurement (e.g., length of perigynium beak) is taken. We also report on how a custom color palette of RGB values obtained from real specimens or high-quality specimen images, can be used to help authors choose standardized color descriptions for plant specimens; and a software demonstration, where we show how Character Recorder and Conflict Resolver can work together to construct both human-readable descriptions and RDF graphs using morphological data derived from species in the plant genus Carex (sedges). The key difference of this system from other ontology-aware systems is that authors can directly add needed terms to the ontology as they wish and can update their data according to ontology updates. a report on the findings from a recent survey of 90+ participants on the need for a tool like Character Recorder; a methods section that describes how we provide semantics to an existing vocabulary of quantitative characters through a set of properties that explain where and how a measurement (e.g., length of perigynium beak) is taken. We also report on how a custom color palette of RGB values obtained from real specimens or high-quality specimen images, can be used to help authors choose standardized color descriptions for plant specimens; and a software demonstration, where we show how Character Recorder and Conflict Resolver can work together to construct both human-readable descriptions and RDF graphs using morphological data derived from species in the plant genus Carex (sedges). The key difference of this system from other ontology-aware systems is that authors can directly add needed terms to the ontology as they wish and can update their data according to ontology updates. The software modules currently incorporated in Character Recorder and Conflict Resolver have undergone formal usability studies. We are actively recruiting Carex experts to participate in a 3-day usability study of the entire system of the Platform for Author-Driven Computable Data and Ontology Production for Taxonomists. Participants will use the platform to record 100 characters about one Carex species. In addition to usability data, we will collect the terms that participants submit to the underlying ontology and the data related to conflict resolution. Such data allow us to examine the types and the quantities of logical conflicts that may result from the terms added by the users and to use Discrete Event Simulation models to understand if and how term additions and conflict resolutions converge. We look forward to a discussion on how the tools (Character Recorder is online at http://shark.sbs.arizona.edu/chrecorder/public) described in our presentation can contribute to producing and publishing FAIR data in taxonomic studies.« less
  4. It is clear that one of the primary tools we can use to mitigate the potential risk from a misbehaving AI system is the ability to turn the system off. As the capabilities of AI systems improve, it is important to ensure that such systems do not adopt subgoals that prevent a human from switching the system off. This is a challenge because many formulations of rational agents create strong incentives for self-preservation. This is not caused by a built-in instinct, but because a rational agent will maximize expected utility and cannot achieve whatever objective it has been given ifmore »it is dead. Our goal is to study the incentives an agent has to allow itself to be switched off. We analyze a simple game between a human H and a robot R, where H can press R’s off switch but R can disable the off switch. A traditional agent takes its reward function for granted: we show that such agents have an incentive to disable the off switch, except in the special case where H is perfectly rational. Our key insight is that for R to want to preserve its off switch, it needs to be uncertain about the utility associated with the outcome, and to treat H’s actions as important observations about that utility. (R also has no incentive to switch itself off in this setting.) We conclude that giving machines an appropriate level of uncertainty about their objectives leads to safer designs, and we argue that this setting is a useful generalization of the classical AI paradigm of rational agents.

    « less
  5. We consider the problem of detecting norm violations in open multi-agent systems (MAS). We show how, using ideas from scrip systems, we can design mechanisms where the agents comprising the MAS are incentivised to monitor the actions of other agents for norm violations. The cost of providing the incentives is not borne by the MAS and does not come from fines charged for norm violations (fines may be impossible to levy in a system where agents are free to leave and rejoin again under a different identity). Instead, monitoring incentives come from (scrip) fees for accessing the services provided bymore »the MAS. In some cases, perfect monitoring (and hence enforcement) can be achieved: no norms will be violated in equilibrium. In other cases, we show that, while it is impossible to achieve perfect enforcement, we can get arbitrarily close; we can make the probability of a norm violation in equilibrium arbitrarily small. We show using simulations that our theoretical results, which apply to systems with a large number of agents, hold for multi-agent systems with as few as 1000 agents–the system rapidly converges to the steady-state distribution of scrip tokens necessary to ensure monitoring and then remains close to the steady state.« less