No abstract

Goals for reinforcement learning problems are typically defined through handspecified rewards. To design such problems, developers of learning algorithms must inherently be aware of what the task goals are, yet we often require agents to discover them on their own without any supervision beyond these sparse rewards. While much of the power of reinforcement learning derives from the concept that agents can learn with little guidance, this requirement greatly burdens the training process. If we relax this one restriction and endow the agent with knowledge of the reward function, and in particular of the goal, we can leverage backwards induction to accelerate training. To achieve this, we propose training a model to learn to take imagined reversal steps from known goal states. Rather than training an agent exclusively to determine how to reach a goal while moving forwards in time, our approach travels backwards to jointly predict how we got there. We evaluate our work in Gridworld and Towers of Hanoi and empirically demonstrate that it yields better performance than standard DDQN.

No abstract

This essay is a set of extended reflections on a book that has both attracted much commentary and exercised a considerable influence in current thinking about gender. This book, Joan Scott's Gender and the Politics of History, l has had an especially powerful impact upon historians, for it is through her eyes that many of us have become acquainted with post-structuralist thought, particularly the ideas of Jacques Derrida. At the same time, Scott has become the spokesperson for gender history, the individual whose thoughts on women and gender are perhaps most widely known among those who otherwise know very little of feminist theory and the decades of scholarship and struggle on which it rests. There is much to admire in Scott's book, for as anyone who looks at the history of women must agree, discursively constructed conceptions of masculine and feminine have exerted enormous force in shaping both lives and opportunities. But lurking at the heart of Scott's call to recognize gender as a "useful category" of analysis lies a radical attack on the naive suppositions of all previous feminist scholarship: a rejection of that ingenuous chain of reasoning which links subjectivity and experience to the hope that oppressed persons, too, might find some agency in history. This rejection seems to rest upon the proposition that social identities are established solely through a process of differentiation analogous to that which Saussure identified as the structural ground of linguistic systems of meaning. The positive identity of one term (in this case, the socially powerful) emerges I would like to thank Kathleen Canning, Geoff Eley, Debbie Field, Pieter Judson, Tom Harrison, and the women and men from my graduate women's history seminar at the University of Michigan for many long and stimulating hours of discussion and debate over the issues aired in this review. I owe particular debts of gratitude to Mary Dearborn, Alice Echols, Sabine MacCormack, and Bill Sewell, who bravely endured several drafts of this piece and offered sage counsels on how to reshape each successive incarnation.

No abstract