<s>
Data	B-General_Concept
augmentation	I-General_Concept
is	O
a	O
technique	O
in	O
machine	O
learning	O
used	O
to	O
reduce	O
overfitting	B-Error_Name
when	O
training	O
a	O
machine	O
learning	O
model	O
,	O
by	O
training	O
models	O
on	O
several	O
slightly-modified	O
copies	O
of	O
existing	O
data	O
.	O
</s>
<s>
Residual	B-Application
or	I-Application
block	I-Application
bootstrap	I-Application
can	O
be	O
used	O
for	O
time	O
series	O
augmentation	O
.	O
</s>
<s>
Synthetic	O
data	B-General_Concept
augmentation	I-General_Concept
is	O
of	O
paramount	O
importance	O
for	O
machine	O
learning	O
classification	O
,	O
particularly	O
for	O
biological	O
data	O
,	O
which	O
tend	O
to	O
be	O
high	O
dimensional	O
and	O
scarce	O
.	O
</s>
<s>
noted	O
that	O
it	O
is	O
possible	O
to	O
use	O
a	O
Generative	B-Algorithm
adversarial	I-Algorithm
network	I-Algorithm
(	O
in	O
particular	O
,	O
a	O
DCGAN	O
)	O
to	O
perform	O
style	O
transfer	O
in	O
order	O
to	O
generate	O
synthetic	O
electromyographic	O
signals	O
that	O
corresponded	O
to	O
those	O
exhibited	O
by	O
sufferers	O
of	O
Parkinson	O
's	O
Disease	O
.	O
</s>
<s>
The	O
approaches	O
are	O
also	O
important	O
in	O
electroencephalography	B-Application
(	O
brainwaves	O
)	O
.	O
</s>
<s>
explored	O
the	O
idea	O
of	O
using	O
Deep	B-Architecture
Convolutional	I-Architecture
Neural	I-Architecture
Networks	I-Architecture
for	O
EEG-Based	O
Emotion	O
Recognition	O
,	O
results	O
show	O
that	O
emotion	O
recognition	O
was	O
improved	O
when	O
data	B-General_Concept
augmentation	I-General_Concept
was	O
used	O
.	O
</s>
<s>
More	O
recently	O
,	O
data	B-General_Concept
augmentation	I-General_Concept
studies	O
have	O
begun	O
to	O
focus	O
on	O
the	O
field	O
of	O
deep	O
learning	O
,	O
more	O
specifically	O
on	O
the	O
ability	O
of	O
generative	O
models	O
to	O
create	O
artificial	O
data	O
which	O
is	O
then	O
introduced	O
during	O
the	O
classification	O
model	O
training	O
process	O
.	O
</s>
<s>
observed	O
that	O
useful	O
EEG	B-Application
signal	O
data	O
could	O
be	O
generated	O
by	O
Conditional	O
Wasserstein	O
Generative	B-Algorithm
Adversarial	I-Algorithm
Networks	I-Algorithm
(	O
GANs	O
)	O
which	O
was	O
then	O
introduced	O
to	O
the	O
training	O
set	O
in	O
a	O
classical	O
train-test	O
learning	O
framework	O
.	O
</s>
<s>
The	O
prediction	O
of	O
mechanical	O
signals	O
based	O
on	O
data	B-General_Concept
augmentation	I-General_Concept
brings	O
a	O
new	O
generation	O
of	O
technological	O
innovations	O
,	O
such	O
as	O
new	O
energy	O
dispatch	O
,	O
5G	O
communication	O
field	O
,	O
and	O
robotics	O
control	O
engineering	O
.	O
</s>
<s>
integrate	O
constraints	O
,	O
optimization	O
and	O
control	O
into	O
a	O
deep	O
network	O
framework	O
based	O
on	O
data	B-General_Concept
augmentation	I-General_Concept
and	O
data	O
pruning	O
with	O
spatio-temporal	O
data	O
correlation	O
,	O
and	O
improve	O
the	O
interpretability	O
,	O
safety	O
and	O
controllability	O
of	O
deep	O
learning	O
in	O
real	O
industrial	O
projects	O
through	O
explicit	O
mathematical	O
programming	O
equations	O
and	O
analytical	O
solutions	O
.	O
</s>
