<s>
XGBoost	B-Language
(	O
eXtreme	O
Gradient	B-Algorithm
Boosting	I-Algorithm
)	O
is	O
an	O
open-source	B-Application
software	I-Application
library	O
which	O
provides	O
a	O
regularizing	O
gradient	B-Algorithm
boosting	I-Algorithm
framework	O
for	O
C++	B-Language
,	O
Java	B-Language
,	O
Python	B-Language
,	O
R	B-Language
,	O
Julia	B-Application
,	O
Perl	B-Language
,	O
and	O
Scala	B-Application
.	O
</s>
<s>
It	O
works	O
on	O
Linux	B-Application
,	O
Windows	O
,	O
and	O
macOS	B-Application
.	O
</s>
<s>
From	O
the	O
project	O
description	O
,	O
it	O
aims	O
to	O
provide	O
a	O
"	O
Scalable	O
,	O
Portable	O
and	O
Distributed	B-Architecture
Gradient	B-Algorithm
Boosting	I-Algorithm
(	O
GBM	O
,	O
GBRT	O
,	O
GBDT	O
)	O
Library	O
"	O
.	O
</s>
<s>
It	O
runs	O
on	O
a	O
single	O
machine	O
,	O
as	O
well	O
as	O
the	O
distributed	B-Architecture
processing	I-Architecture
frameworks	O
Apache	B-Application
Hadoop	I-Application
,	O
Apache	B-Language
Spark	I-Language
,	O
Apache	B-Application
Flink	I-Application
,	O
and	O
Dask	B-Operating_System
.	O
</s>
<s>
XGBoost	B-Language
initially	O
started	O
as	O
a	O
research	O
project	O
by	O
Tianqi	O
Chen	O
as	O
part	O
of	O
the	O
Distributed	B-Architecture
(	O
Deep	O
)	O
Machine	O
Learning	O
Community	O
(	O
DMLC	O
)	O
group	O
.	O
</s>
<s>
Initially	O
,	O
it	O
began	O
as	O
a	O
terminal	O
application	O
which	O
could	O
be	O
configured	O
using	O
a	O
libsvm	B-Language
configuration	O
file	O
.	O
</s>
<s>
Soon	O
after	O
,	O
the	O
Python	B-Language
and	O
R	B-Language
packages	O
were	O
built	O
,	O
and	O
XGBoost	B-Language
now	O
has	O
package	O
implementations	O
for	O
Java	B-Language
,	O
Scala	B-Application
,	O
Julia	B-Application
,	O
Perl	B-Language
,	O
and	O
other	O
languages	O
.	O
</s>
<s>
This	O
brought	O
the	O
library	O
to	O
more	O
developers	O
and	O
contributed	O
to	O
its	O
popularity	O
among	O
the	O
Kaggle	B-Application
community	O
,	O
where	O
it	O
has	O
been	O
used	O
for	O
a	O
large	O
number	O
of	O
competitions	O
.	O
</s>
<s>
It	O
has	O
now	O
been	O
integrated	O
with	O
scikit-learn	B-Application
for	O
Python	B-Language
users	O
and	O
with	O
the	O
package	O
for	O
R	B-Language
users	O
.	O
</s>
<s>
It	O
can	O
also	O
be	O
integrated	O
into	O
Data	O
Flow	O
frameworks	O
like	O
Apache	B-Language
Spark	I-Language
,	O
Apache	B-Application
Hadoop	I-Application
,	O
and	O
Apache	B-Application
Flink	I-Application
using	O
the	O
abstracted	O
Rabit	O
and	O
XGBoost4J	O
.	O
</s>
<s>
XGBoost	B-Language
is	O
also	O
available	O
on	O
OpenCL	B-Application
for	O
FPGAs	O
.	O
</s>
<s>
An	O
efficient	O
,	O
scalable	O
implementation	O
of	O
XGBoost	B-Language
has	O
been	O
published	O
by	O
Tianqi	O
Chen	O
and	O
Carlos	O
Guestrin	O
.	O
</s>
<s>
While	O
the	O
XGBoost	B-Language
model	O
often	O
achieves	O
higher	O
accuracy	O
than	O
a	O
single	O
decision	O
tree	O
,	O
it	O
sacrifices	O
the	O
intrinsic	O
interpretability	O
of	O
decision	O
trees	O
.	O
</s>
<s>
To	O
achieve	O
both	O
performance	O
and	O
interpretability	O
,	O
some	O
model	O
compression	O
techniques	O
allow	O
transforming	O
an	O
XGBoost	B-Language
into	O
a	O
single	O
"	O
born-again	O
"	O
decision	O
tree	O
that	O
approximates	O
the	O
same	O
decision	O
function	O
.	O
</s>
<s>
Salient	O
features	O
of	O
XGBoost	B-Language
which	O
make	O
it	O
different	O
from	O
other	O
gradient	B-Algorithm
boosting	I-Algorithm
algorithms	O
include	O
:	O
</s>
<s>
XGBoost	B-Language
works	O
as	O
Newton-Raphson	O
in	O
function	O
space	O
unlike	O
gradient	B-Algorithm
boosting	I-Algorithm
that	O
works	O
as	O
gradient	O
descent	O
in	O
function	O
space	O
,	O
a	O
second	O
order	O
Taylor	O
approximation	O
is	O
used	O
in	O
the	O
loss	O
function	O
to	O
make	O
the	O
connection	O
to	O
Newton	O
Raphson	O
method	O
.	O
</s>
<s>
A	O
generic	O
unregularized	O
XGBoost	B-Language
algorithm	O
is	O
:	O
</s>
