Transferrable Representations for Visual Recognition by Jeffrey Donahue Doctor of Philosophy in Computer Science University of California, Berkeley Professor Trevor Darrell, Chair The rapid progress in visual recognition capabilities over the past several years can be attributed largely to improvements in generic and transferrable feature representations, particularly learned representations ba...

We complete the Wilf classification of signed patterns of length 5 for both signed permutations and signed involutions. New general equivalences of patterns are given which prove Jaggard’s conjectures concerning involutions in the symmetric group avoiding certain patterns of length 5 and 6. In this way, we also complete the Wilf classification of S5, S6, and S7 for involutions.

This dissertation deals with firms’ visual and pre-visual self-representations in accounting narratives. Self-representations are those descriptions about the company that firms include in accounting narratives to convey the current standings and their identity. External stakeholders increasingly expect non-numerical information about firms to be disclosed, and accounting narratives are a key m...

Representations in sensory cortices are organized topographically: auditory cortex is organized tonotopically, somatosensory cortex is organized somatotopically, and visual cortex is organized retinotopically. Substantial progress has been made in understanding how topography develops at a neurocomputational level, particularly in the early and middle stages of processing in the visual system. ...

Entity images provide significant visual information that helps the construction of knowledge representations. Most conventional methods learn knowledge representations solely from structured triples, ignoring rich visual information extracted from entity images. In this paper, we propose a novel Image-embodied Knowledge Representation Learning model, where knowledge representations are learned...

Traditional semantic models are disembodied from the human perception and action. In this work, we attempt to address this problem by grounding semantic representations of words to the acoustic and visual modalities. Specifically we estimate multimodal word representations via the fusion of auditory and visual modalities with the text modality. We employ middle and late fusion of representation...

Although the problems on Raven’s Progressive Matrices intelligence tests resemble geometric analogies, studies of human behavior suggest the existence of two qualitatively distinct types of strategies: verbal strategies that use propositional representations and visual strategies that use iconic representations. However, all prior computational models implemented to solve these tests have model...