Search for: All records

We consider a novel application of inverse reinforcement learning with behavioral economics constraints to model, learn and predict the commenting behavior of YouTube viewers. Each group of users is modeled as a rationally inattentive Bayesian agent which solves a contextual bandit problem. Our methodology integrates three key components. First, to identify distinct commenting patterns, we use deep embedded clustering to estimate framing information (essential extrinsic features) that clusters users into distinct groups. Second, we present an inverse reinforcement learning algorithm that uses Bayesian revealed preferences to test for rationality: does there exist a utility function that rationalizes the given data, and if yes, can it be used to predict commenting behavior? Finally, we impose behavioral economics constraints stemming from rational inattention to characterize the attention span of groups of users. The test imposes a Rényi mutual information cost constraint which impacts how the agent can select attention strategies to maximize their expected utility. After a careful analysis of a massive YouTube dataset, our surprising result is that in most YouTube user groups, the commenting behavior is consistent with optimizing a Bayesian utility with rationally inattentive constraints. The paper also highlights how the rational inattention model can accurately predict commenting behavior. The massive YouTube dataset and analysis used in this paper are available on GitHub and completely reproducible 

The paper presents risk-neutral and risk-averse caching policies that can be deployed in a femtocell network with limited storage capacity to reduce the time delay of servicing content requests. The caching policies use a forecasting algorithm to estimate the cumulative distribution function of content requests based on the content features. Given the cumulative distribution function, a mixed-integer linear program is used to compute where to cache content in the femtocell network. The caching policies account for the uncertainty associated with estimating the content requests using the coherent Conditional Value-at-Risk (CVaR) measure. For a large number of content, a risk-neutral caching policy is constructed that accounts for both the content features and routing protocol that only requires the evaluation of a unimodular linear program. Using data from YouTube (comprising 25,000 videos) and the NS-3 simulator, the caching policies reduce the delay of retrieving content in femtocell networks compared with industry standard caching policies. Specifically, a 6\% reduction in delay is achieved by accounting for the uncertainty, and a 60\% reduction in delay is achieved if both the uncertainty and femtocell routing protocol are accounted for compared to the risk-neutral caching policy that neglects the routing protocol.