Human vision is able to immediately recognize novel visual categories after
seeing just one or a few training examples. We describe how to add a similar
capability to ConvNet classifiers by directly setting the final layer weights
from novel training examples during low-shot learning. We call this process
weight imprinting as it directly sets weights for a new category based on an
appropriately scaled copy of the embedding layer activations for that training
example. The imprinting process provides a valuable complement to training with
stochastic gradient descent, as it provides immediate good classification
performance and an initialization for any further fine-tuning in the future. We
show how this imprinting process is related to proxy-based embeddings. However,
it differs in that only a single imprinted weight vector is learned for each
novel category, rather than relying on a nearest-neighbor distance to training
instances as typically used with embedding methods. Our experiments show that
using averaging of imprinted weights provides better generalization than using
nearest-neighbor instance embeddings.
%0 Generic
%1 citeulike:14514525
%A xxx,
%D 2017
%K classification oneshot
%T Learning with Imprinted Weights
%U http://arxiv.org/abs/1712.07136
%X Human vision is able to immediately recognize novel visual categories after
seeing just one or a few training examples. We describe how to add a similar
capability to ConvNet classifiers by directly setting the final layer weights
from novel training examples during low-shot learning. We call this process
weight imprinting as it directly sets weights for a new category based on an
appropriately scaled copy of the embedding layer activations for that training
example. The imprinting process provides a valuable complement to training with
stochastic gradient descent, as it provides immediate good classification
performance and an initialization for any further fine-tuning in the future. We
show how this imprinting process is related to proxy-based embeddings. However,
it differs in that only a single imprinted weight vector is learned for each
novel category, rather than relying on a nearest-neighbor distance to training
instances as typically used with embedding methods. Our experiments show that
using averaging of imprinted weights provides better generalization than using
nearest-neighbor instance embeddings.
@misc{citeulike:14514525,
abstract = {{Human vision is able to immediately recognize novel visual categories after
seeing just one or a few training examples. We describe how to add a similar
capability to ConvNet classifiers by directly setting the final layer weights
from novel training examples during low-shot learning. We call this process
weight imprinting as it directly sets weights for a new category based on an
appropriately scaled copy of the embedding layer activations for that training
example. The imprinting process provides a valuable complement to training with
stochastic gradient descent, as it provides immediate good classification
performance and an initialization for any further fine-tuning in the future. We
show how this imprinting process is related to proxy-based embeddings. However,
it differs in that only a single imprinted weight vector is learned for each
novel category, rather than relying on a nearest-neighbor distance to training
instances as typically used with embedding methods. Our experiments show that
using averaging of imprinted weights provides better generalization than using
nearest-neighbor instance embeddings.}},
added-at = {2019-02-27T22:23:29.000+0100},
archiveprefix = {arXiv},
author = {xxx},
biburl = {https://www.bibsonomy.org/bibtex/2d0d258a083faf342dda43055fabb62b2/nmatsuk},
citeulike-article-id = {14514525},
citeulike-linkout-0 = {http://arxiv.org/abs/1712.07136},
citeulike-linkout-1 = {http://arxiv.org/pdf/1712.07136},
day = 19,
eprint = {1712.07136},
interhash = {029ed3f8d4be48362f741599f8219468},
intrahash = {d0d258a083faf342dda43055fabb62b2},
keywords = {classification oneshot},
month = dec,
posted-at = {2018-01-10 14:20:51},
priority = {3},
timestamp = {2019-02-27T22:23:29.000+0100},
title = {{Learning with Imprinted Weights}},
url = {http://arxiv.org/abs/1712.07136},
year = 2017
}