Non-Bayesian Social Learning With Uncertain Models

Hare, James Z.; Uribe, César A.; Kaplan, Lance; Jadbabaie, Ali

doi:10.1109/tsp.2020.3006755

Cited by 26 publications

(34 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In practice we often have access to feature data only, or to some approximate models for the distributions. For example, uncertain likelihoods in social learning have been considered in [9], albeit only for multinomial distributions. In this work, we will allow for a fairly broad class of distributions.…”

Section: Introduction and Related Workmentioning

confidence: 99%

Network Classifiers Based on Social Learning

Bordignon

Vlaski

Matta

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

This work proposes a new way of combining independently trained classifiers over space and time. Combination over space means that the outputs of spatially distributed classifiers are aggregated. Combination over time means that the classifiers respond to streaming data during testing and continue to improve their performance even during this phase. By doing so, the proposed architecture is able to improve prediction performance over time with unlabeled data. Inspired by social learning algorithms, which require prior knowledge of the observations distribution, we propose a Social Machine Learning (SML) paradigm that is able to exploit the imperfect models generated during the learning phase. We show that this strategy results in consistent learning with high probability, and it yields a robust structure against poorly trained classifiers. Simulations with an ensemble of feedforward neural networks are provided to illustrate the theoretical results.

show abstract

Section: Introduction and Related Workmentioning

confidence: 99%

Network Classifiers Based on Social Learning

Bordignon

Vlaski

Matta

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…This kind of adversarial behavior was considered in [16,19]. In [16] the authors devised a detection scheme where each agent This work was supported in part by the Swiss National Science Foundation grant 205121-184999.…”

Section: Introduction and Related Workmentioning

confidence: 99%

“…The focus of these works is on detection of adversaries. Instead, our work is focused on investigating adversarial strategies, which is not addressed by [16,19].…”

Section: Introduction and Related Workmentioning

confidence: 99%

Social Learning Under Inferential Attacks

Ntemos

Bordignon

Vlaski

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

A common assumption in the social learning literature is that agents exchange information in an unselfish manner. In this work, we consider the scenario where a subset of agents aims at driving the network beliefs to the wrong hypothesis. The adversaries are unaware of the true hypothesis. However, they will "blend in" by behaving similarly to the other agents and will manipulate the likelihood functions used in the belief update process to launch inferential attacks. We will characterize the conditions under which the network is misled. Then, we will explain that it is possible for such attacks to succeed by showing that strategies exist that can be adopted by the malicious agents for this purpose. We examine both situations in which the agents have minimal or no information about the network model.

show abstract

“…In practice, these models are only approximate, oftentimes the result of a previous training stage, where, from limited data, a parameterized model is learned. This issue is recognized in the works [11], [12], where the authors propose a framework for incorporating uncertainty into non-Bayesian social learning. While [11] focuses only on sets of Gaussian distributions, their proposed strategy in [12] broadens the approach, but requires nonetheless prior knowledge about the structure of the family of likelihoods, i.e., the exact parameterization of the distributions.…”

mentioning

confidence: 99%

“…This issue is recognized in the works [11], [12], where the authors propose a framework for incorporating uncertainty into non-Bayesian social learning. While [11] focuses only on sets of Gaussian distributions, their proposed strategy in [12] broadens the approach, but requires nonetheless prior knowledge about the structure of the family of likelihoods, i.e., the exact parameterization of the distributions. While relevant for numerically generated data, in practical applications there is generally little a priori evidence regarding the structure of likelihoods, e.g., in the distributed classification of images or videos.…”

mentioning

confidence: 99%

Learning from Heterogeneous Data Based on Social Interactions over Graphs

Bordignon¹,

Vlaski²,

Matta³

et al. 2021

Preprint

View full text Add to dashboard Cite

This work proposes a decentralized architecture, where individual agents aim at solving a classification problem while observing streaming features of different dimensions and arising from possibly different distributions. In the context of social learning, several useful strategies have been developed, which solve decision making problems through local cooperation across distributed agents and allow them to learn from streaming data. However, traditional social learning strategies rely on the fundamental assumption that each agent has significant prior knowledge of the underlying distribution of the observations. In this work we overcome this issue by introducing a machine learning framework that exploits social interactions over a graph, leading to a fully data-driven solution to the distributed classification problem. In the proposed social machine learning (SML) strategy, two phases are present: in the training phase, classifiers are independently trained to generate a belief over a set of hypotheses using a finite number of training samples; in the prediction phase, classifiers evaluate streaming unlabeled observations and share their instantaneous beliefs with neighboring classifiers. We show that the SML strategy enables the agents to learn consistently under this highly-heterogeneous setting and allows the network to continue learning even during the prediction phase when it is deciding on unlabeled samples. The prediction decisions are used to continually improve performance thereafter in a manner that is markedly different from most existing static classification schemes where, following training, the decisions on unlabeled data are not re-used to improve future performance.

show abstract

Non-Bayesian Social Learning With Uncertain Models

Cited by 26 publications

References 50 publications

Network Classifiers Based on Social Learning

Network Classifiers Based on Social Learning

Social Learning Under Inferential Attacks

Learning from Heterogeneous Data Based on Social Interactions over Graphs

Contact Info

Product

Resources

About