Fathom

Abstract: Deep learning has been popularized by its recent successes on challenging artificial intelligence problems. One of the reasons for its dominance is also an ongoing challenge: the need for immense amounts of computational power. Hardware architects have responded by proposing a wide array of promising ideas, but to date, the majority of the work has focused on specific algorithms in somewhat narrow application domains. While their specificity does not diminish these approaches, there is a clear need for more flexible solutions. We believe the first step is to examine the characteristics of cutting edge models from across the deep learning community. Consequently, we have assembled Fathom: a collection of eight archetypal deep learning workloads for study. Each of these models comes from a seminal work in the deep learning community, ranging from the familiar deep convolutional neural network of Krizhevsky et al., to the more exotic memory networks from Facebook's AI research group. Fathom has been released online, and this paper focuses on understanding the fundamental performance characteristics of each model. We use a set of application-level modeling tools built around the TensorFlow deep learning framework in order to analyze the behavior of the Fathom workloads. We present a breakdown of where time is spent, the similarities between the performance profiles of our models, an analysis of behavior in inference and training, and the effects of parallelism on scaling.

Abstract: The sorrows and joys of this rare condition are well understood, but there have been relatively few convincing attempts to fathom its origins. In this major contribution to a neglected area, Dr. Markowitz hypothesizes that creativity develops as a product of the necessities facing an individual whose early environment permitted occasional depersonification of thought.

Abstract: The hero of "Ferdinand Count Fathom" (1753) is a monster of treachery and fraud. Fate and coincidence play a large part in his picaresque progress through England and Europe, and much of the narrative is written in a mock-heroic style.

Abstract: While technology has become an integral part of introductory statistics courses, the programs typically employed are professional packages designed primarily for data analysis rather than for learning Findings from several studies suggest that use of such software in the introductory statistics classroom may not be very effective in helping students to build intuitions about the fundamental statistical ideas of sampling distribution and inferential statistics The paper describes an instructional experiment which explored the capabilities of Fathom, one of several recently-developed packages explicitly designed to enhance learning Findings from the study indicate that use of Fathom led students to the construction of a fairly coherent mental model of sampling distributions and other key concepts related to statistical inference The insights gained point to a number of critical ingredients that statistics educators should consider when choosing statistical software They also provide suggestions about how to approach the particularly challenging topic of statistical inference