2009 DrosophilaGeneExpressionPatternAnnotation Abstract

From GM-RKB
Jump to navigation Jump to search

See: Abstract Analysis Report, (Ji et al., 2009) ⇒ Shuiwang Ji, Lei Yuan, Ying-Xin Li, Zhi-Hua Zhou, Sudhir Kumar, and Jieping Ye. (2009). “Drosophila Gene Expression Pattern Annotation Using Sparse Features and Term-term Interactions.” In: Proceedings of ACM SIGKDD Conference (KDD-2009).


The The  
Drosophila Drosophila  
gene gene expression pattern Gene Expression Pattern
expression
pattern
images images  
document document  
the the  
spatial spatial  
and and  
temporal temporal  
dynamics dynamics  
of of  
gene gene expression Gene Expression
expression
and and  
they they  
are are  
valuable valuable  
tools tools  
for for  
explicating explicating  
the the  
gene gene  
functions functions  
, ,  
interaction interaction  
, ,  
and and  
networks networks  
during during  
Drosophila Drosophila  
embryogenesis embryogenesis  
. .  
     
To To  
provide provide  
text-based text-based  
pattern pattern  
searching searching  
, ,  
the the  
images images  
in in  
the the  
Berkeley Berkeley Drosophila Genome Project Berkeley Drosophila Genome Project
Drosophila
Genome
Project
( (  
BDGP BDGP  
) )  
study study  
are are  
annotated annotated  
with with  
ontology ontology  
terms terms  
manually manually  
by by  
human human curator  
curators  
. .  
     
We We  
present present  
a a  
systematic systematic  
approach approach  
for for  
automating automating  
this this  
task task  
, ,  
because because  
the the  
number number  
of of  
images images  
needing needing  
text text  
descriptions descriptions  
is is  
now now  
rapidly rapidly  
increasing increasing  
. .  
     
We We  
consider consider  
both both  
improved improved  
feature feature  
representation representation  
and and  
novel novel  
learning learning ?
formulation formulation  
to to  
boost boost  
the the  
annotation annotation Annotation Task
performance performance Performance
. .  
     
For For  
feature feature  
representation representation  
, ,  
we we  
adapt adapt  
the the  
bag-of-words bag-of-words  
scheme scheme  
commonly commonly  
used used  
in in  
visual visual recognition problem Image Recognition Task
recognition
problems
so so  
that that  
the the  
image image  
group group  
information information  
in in  
the the  
BDGP BDGP BDGP
study study  
is is  
retained retained  
. .  
     
Moreover Moreover  
, ,  
images images  
from from  
multiple multiple  
views views  
can can  
be be  
integrated integrated  
naturally naturally  
in in  
this this  
representation representation  
. .  
     
To To  
reduce reduce  
the the  
quantization quantization  
error error  
caused caused  
by by  
the the  
bag-of-words bag-of-words representation Bag-of-Words Representation
representation,
, ,  
     
we we  
propose propose  
an an  
improved improved  
feature feature representation Feature Representation
representation
scheme scheme  
based based  
on on  
the the  
sparse sparse learning technique Sparse Learning Technique
learning
technique.
. .  
     
In In  
the the  
design design  
of of  
learning learning  
formulation formulation  
, ,  
we we  
propose propose  
a a  
local local  
regularization regularization  
framework framework  
that that  
can can  
incorporate incorporate  
the the  
correlations correlations  
among among  
terms terms  
explicitly explicitly  
. .  
     
We We  
further further  
show show  
that that  
the the  
resulting resulting  
optimization optimization problem Optimization Task
problem
admits admits  
an an  
analytical analytical  
solution solution  
. .  
     
Experimental Experimental results [[Experimental Result]]
results
show show  
that that  
the the  
representation representation  
based based  
on on  
sparse sparse learning Sparse Learning
learning
outperforms outperforms  
the the  
bag-of-words bag-of-words Bag-of-Words Representation
representation representation
significantly significantly  
. .  
     
Results Results Experimental Result
also also  
show show  
that that  
incorporation incorporation  
of of  
the the  
term-term term-term Term-Term Relation
correlations correlations Correlation
improves improves  
the the  
annotation annotation Annotation Task
performance performance Performance
consistently consistently  
. .  


,