1986

Ronald J. Williams

b. 1945

Backpropagation & Policy Learning

Co-authored the landmark backpropagation paper with David Rumelhart and Geoffrey Hinton, and later developed REINFORCE.

Williams was one of the three authors of the 1986 paper Learning Representations by Back-propagating Errors. Collaborative credit is essential to understanding this milestone.

His 1992 REINFORCE work described gradient-based learning for stochastic decisions, linking neural learning with reinforcement learning.

Dated contributions

  1. 1986

    Backpropagation & Policy Learning

    David Rumelhart, Geoffrey Hinton & Ronald J. Williams

    Co-authored the landmark backpropagation paper with David Rumelhart and Geoffrey Hinton, and later developed REINFORCE.

    Paper · 1986 — Learning representations by back-propagating errors ↗
  2. 1992

    REINFORCE

    Ronald J. Williams

    Gradient-based learning for stochastic decisions connects neural models with reinforcement learning.

    Paper · 1992 — Simple statistical gradient-following algorithms for connectionist reinforcement learning ↗

Reference

Original work & historical records

Wikipedia