Machine learning powers diverse services in industry including search, translation, recommendation systems, and security. The scale and importance of these models require that they be efficient, expressive, and portable across an array of heterogeneous hardware devices. These constraints are often at odds; in order to better accommodate them we propose a new high-level intermediate representation (IR) called Relay. Relay is being designed as a purely-functional, statically-typed language with the goal of balancing efficient compilation, expressiveness, and portability. We discuss the goals of Relay and highlight its important design constraints. Our prototype is part of the open source NNVM compiler framework, which powers Amazon's deep learning framework MxNet.

A fundamental task for sociology is to uncover the mechanisms that produce and reproduce social inequalities. While status characteristics theory is the favored account of how social status contributes independently to the maintenance of inequality, it hinges on an unobserved construct, expectation states, in the middle of the causal chain between status and behavior. Efforts to test the mediation mechanism have been complicated by the implicit, often unconscious, nature of status expectations. To solve this “black box” problem, we offer a new conceptualization and research approach that capitalizes on the accuracy and precision of neurological measurement to shed new light on the biasing role of expectations in the status–behavior relationship. Results from an experimental study provide a unique illustration of ways in which social status is inscribed in the brain and how, in turn, these inscriptions are related to behavioral inequalities that emerge during interaction.

No abstract

No abstract