<s>
Proactive	B-General_Concept
learning	I-General_Concept
is	O
a	O
generalization	O
of	O
active	B-General_Concept
learning	I-General_Concept
designed	O
to	O
relax	O
unrealistic	O
assumptions	O
and	O
thereby	O
reach	O
practical	O
applications	O
.	O
</s>
<s>
"	O
Active	B-General_Concept
learning	I-General_Concept
seeks	O
to	O
select	O
the	O
most	O
informative	O
unlabeled	O
instances	O
and	O
ask	O
an	O
omniscient	O
oracle	O
for	O
their	O
labels	O
,	O
so	O
as	O
to	O
retrain	O
a	O
learning	O
algorithm	O
maximizing	O
accuracy	O
.	O
</s>
<s>
Active	B-General_Concept
learning	I-General_Concept
also	O
assumes	O
that	O
the	O
single	O
oracle	O
is	O
perfect	O
,	O
always	O
providing	O
a	O
correct	O
answer	O
when	O
requested	O
.	O
</s>
<s>
Finally	O
,	O
active	B-General_Concept
learning	I-General_Concept
presumes	O
the	O
oracle	O
is	O
either	O
free	O
or	O
charges	O
uniform	O
cost	O
in	O
label	O
elicitation	O
.	O
</s>
<s>
Proactive	B-General_Concept
learning	I-General_Concept
relaxes	O
all	O
four	O
of	O
these	O
assumptions	O
,	O
relying	O
on	O
a	O
decision-theoretic	O
approach	O
to	O
jointly	O
select	O
the	O
optimal	O
oracle	O
and	O
instance	O
,	O
by	O
casting	O
the	O
problem	O
as	O
a	O
utility	O
optimization	O
problem	O
subject	O
to	O
a	O
budget	O
constraint	O
.	O
</s>
