<s>
Transfer	B-General_Concept
learning	I-General_Concept
(	O
TL	O
)	O
is	O
a	O
research	O
problem	O
in	O
machine	O
learning	O
(	O
ML	O
)	O
that	O
focuses	O
on	O
applying	O
knowledge	O
gained	O
while	O
solving	O
one	O
task	O
to	O
a	O
related	O
task	O
.	O
</s>
<s>
In	O
1976	O
,	O
Bozinovski	O
and	O
Fulgosi	O
published	O
a	O
paper	O
addressing	O
transfer	B-General_Concept
learning	I-General_Concept
in	O
neural	B-Architecture
network	I-Architecture
training	O
.	O
</s>
<s>
In	O
1981	O
,	O
a	O
report	O
considered	O
the	O
application	O
of	O
transfer	B-General_Concept
learning	I-General_Concept
to	O
a	O
dataset	O
of	O
images	O
representing	O
letters	O
of	O
computer	O
terminals	O
,	O
experimentally	O
demonstrating	O
positive	O
and	O
negative	O
transfer	B-General_Concept
learning	I-General_Concept
.	O
</s>
<s>
In	O
1997	O
,	O
Pratt	O
and	O
Thrun	O
guest-edited	O
a	O
special	O
issue	O
of	O
Machine	O
Learning	O
devoted	O
to	O
transfer	B-General_Concept
learning	I-General_Concept
,	O
and	O
by	O
1998	O
,	O
the	O
field	O
had	O
advanced	O
to	O
include	O
multi-task	B-General_Concept
learning	I-General_Concept
,	O
along	O
with	O
more	O
formal	O
theoretical	O
foundations	O
.	O
</s>
<s>
Transfer	B-General_Concept
learning	I-General_Concept
has	O
been	O
applied	O
in	O
cognitive	O
science	O
.	O
</s>
<s>
Pratt	O
guest-edited	O
an	O
issue	O
of	O
Connection	O
Science	O
on	O
reuse	O
of	O
neural	B-Architecture
networks	I-Architecture
through	O
transfer	O
in	O
1996	O
.	O
</s>
<s>
Ng	O
said	O
in	O
his	O
NIPS	O
2016	O
tutorial	O
that	O
TL	O
would	O
become	O
the	O
next	O
driver	O
of	O
machine	O
learning	O
commercial	O
success	O
after	O
supervised	B-General_Concept
learning	I-General_Concept
.	O
</s>
<s>
The	O
definition	O
of	O
transfer	B-General_Concept
learning	I-General_Concept
is	O
given	O
in	O
terms	O
of	O
domains	O
and	O
tasks	O
.	O
</s>
<s>
Given	O
a	O
source	O
domain	O
and	O
learning	O
task	O
,	O
a	O
target	O
domain	O
and	O
learning	O
task	O
,	O
where	O
,	O
or	O
,	O
transfer	B-General_Concept
learning	I-General_Concept
aims	O
to	O
help	O
improve	O
the	O
learning	O
of	O
the	O
target	O
predictive	O
function	O
in	O
using	O
the	O
knowledge	O
in	O
and	O
.	O
</s>
<s>
Algorithms	O
are	O
available	O
for	O
transfer	B-General_Concept
learning	I-General_Concept
in	O
Markov	O
logic	O
networks	O
and	O
Bayesian	O
networks	O
.	O
</s>
<s>
Transfer	B-General_Concept
learning	I-General_Concept
has	O
been	O
applied	O
to	O
cancer	O
subtype	O
discovery	O
,	O
building	O
utilization	O
,	O
general	B-Algorithm
game	I-Algorithm
playing	I-Algorithm
,	O
text	B-Algorithm
classification	I-Algorithm
,	O
digit	O
recognition	O
,	O
medical	O
imaging	O
and	O
spam	O
filtering	O
.	O
</s>
<s>
In	O
2020	O
it	O
was	O
discovered	O
that	O
,	O
due	O
to	O
their	O
similar	O
physical	O
natures	O
,	O
transfer	B-General_Concept
learning	I-General_Concept
is	O
possible	O
between	O
electromyographic	O
(	O
EMG	O
)	O
signals	O
from	O
the	O
muscles	O
and	O
classifying	O
the	O
behaviors	O
of	O
electroencephalographic	B-Application
(	O
EEG	B-Application
)	O
brainwaves	O
,	O
from	O
the	O
gesture	B-General_Concept
recognition	I-General_Concept
domain	O
to	O
the	O
mental	O
state	O
recognition	O
domain	O
.	O
</s>
<s>
It	O
was	O
noted	O
that	O
this	O
relationship	O
worked	O
in	O
both	O
directions	O
,	O
showing	O
that	O
electroencephalographic	B-Application
can	O
likewise	O
be	O
used	O
to	O
classify	O
EMG	O
.	O
</s>
<s>
The	O
experiments	O
noted	O
that	O
the	O
accuracy	O
of	O
neural	B-Architecture
networks	I-Architecture
and	O
convolutional	B-Architecture
neural	I-Architecture
networks	I-Architecture
were	O
improved	O
,	O
through	O
transfer	B-General_Concept
learning	I-General_Concept
prior	O
to	O
any	O
learning	O
(	O
first	O
epoch	O
)	O
,	O
ie	O
.	O
</s>
<s>
Several	O
compilations	O
of	O
transfer	B-General_Concept
learning	I-General_Concept
and	O
domain	B-General_Concept
adaptation	I-General_Concept
algorithms	O
have	O
been	O
implemented	O
:	O
</s>
