Geometric Generalization Based Zero-Shot Learning Dataset Infinite World: Simple Yet Powerful

  • 2018-07-11 09:09:10
  • Rajesh Chidambaram, Michael Kampffmeyer, Willie Neiswanger, Xiaodan Liang, Thomas Lachmann, Eric Xing
  • 0

Abstract

Raven's Progressive Matrices are one of the widely used tests in evaluatingthe human test taker's fluid intelligence. Analogously, this paper introducesgeometric generalization based zero-shot learning tests to measure the rapidlearning ability and the internal consistency of deep generative models. Ourempirical research analysis on state-of-the-art generative models discern theirability to generalize concepts across classes. In the process, we introduceInfinite World, an evaluable, scalable, multi-modal, light-weight dataset andZero-Shot Intelligence Metric ZSI. The proposed tests condenses human-levelspatial and numerical reasoning tasks to its simplistic geometric forms. Thedataset is scalable to a theoretical limit of infinity, in numerical featuresof the generated geometric figures, image size and in quantity. Wesystematically analyze state-of-the-art model's internal consistency, identifytheir bottlenecks and propose a pro-active optimization method for few-shot andzero-shot learning.

 

Introduction (beta)

None

 

Conclusion (beta)

None