[ThH S SK r S SKrS SKrS SKJr S SKJr S SKJr S SKJ r J
r
S/r S r / SQr
/ S Qr/ S
Qr/ S Qr\
\/r\\/rSS jr\R( R* S
5 r " S S\5 rg) N)Tensor) constraints) Distribution)
broadcast_all
lazy_propertyVonMisesc [ U5 nUR 5 nU( a UR 5 X- - nU( a M U$ N)listpop)ycoefresults U/var/www/auris/envauris/lib/python3.13/site-packages/torch/distributions/von_mises.py
_eval_polyr s7 :D
XXZF
aj( $ M )g ?g$ @g03@g,?N?g2t?gIx?gtHZr?) e3E?g -5?gՒ+Hub?gJNYgTPÂ?g 'gZ?gUL+ߐg;^p?) ?gY ?g(z?g*O?gZ9?g.h?gӰ٩=5?) r g .kg?VmgtZOZ?g< Q g' 8`?gP⥝gqJ:N?g;PJ4qc Z US:X d US:X d eU S- nX"- n[ U[ U 5 nUS:X a U R 5 U- nUR 5 nSU - nU SU R 5 - -
[ U[ U 5 R 5 - n[
R " U S: X45 nU$ )zL
Returns ``log(I_order(x))`` for ``x > 0``,
where `order` is either 0 or 1.
r g @r )r _COEF_SMALLabslog _COEF_LARGEtorchwhere)xorderr
smalllarger s r _log_modified_bessel_fnr! D s
A:!# #
DA A q+e, -E z% IIKE
qA
aeeg
1k%.@ A E E G GE
[[T5
0F Mr c D [ R " UR [ R U R S9nUR 5 ( Gd [ R " SUR - U R U R S9nUR 5 u pgn[ R " [ R U- 5 n SX)- - X)- - n
XU
-
- n U SU -
- U-
S: X- R 5 S- U -
S: - n U R 5 ( a= [ R " XS-
R 5 U
R! 5 - U5 nXL- nUR 5 ( d GM U[ R - U - S[ R - - [ R -
$ )Ndtypedevice) r r r )r zerosshapeboolr% allrandr$ unbindcosmathpir anyr signacos)
loc
concentration
proposal_rr doneuu1u2u3zfcaccepts
r _rejection_sampler@ [ s4 ;;qwwejj DDhhjj JJtagg~SYYszz JXXZ
IIdggl #
JN 3 !^ ,A;#q(af\\^a-?!-Cq-HI ::<< F#XOO$5$@!DA=D hhjj
K# !dgg+ . 88r c v ^ \ rS rSrSr\R \R S.r\R r Sr
SU 4S jjr S r \
S\4S j5 r\
S\4S j5 r\
S\4S
j5 r\R& " 5 \R( " 5 4S j5 rSU 4S jjr\S\4S
j5 r\S\4S j5 r\
S\4S j5 rSrU =r $ )r k a(
A circular von Mises distribution.
This implementation uses polar coordinates. The ``loc`` and ``value`` args
can be any real number (to facilitate unconstrained optimization), but are
interpreted as angles modulo 2 pi.
Example::
>>> # xdoctest: +IGNORE_WANT("non-deterministic")
>>> m = VonMises(torch.tensor([1.0]), torch.tensor([1.0]))
>>> m.sample() # von Mises distributed with loc=1 and concentration=1
tensor([1.9777])
:param torch.Tensor loc: an angle in radians.
:param torch.Tensor concentration: concentration parameter
)r4 r5 Fc > [ X5 u U l U l U R R n[ R
" 5 n[ TU ] XEU5 g r
)r r4 r5 r) r Sizesuper__init__)selfr4 r5
validate_args batch_shape event_shape __class__s r rF VonMises.__init__ s> '4S'H$$$hhnn jjl
=Ar c & U R ( a U R U5 U R [ R " XR
-
5 - nU[ R " S[ R - 5 -
[ U R SS9-
nU$ )Nr' r r )
_validate_args_validate_sampler5 r r. r4 r/ r r0 r! )rG valuelog_probs r rR VonMises.log_prob sw ! !% (%% %((2B(CC hhq477{#
$%d&8&8B
C
r returnc T U R R [ R 5 $ r
)r4 tor doublerG s r _loc
VonMises._loc s xx{{5<<((r c T U R R [ R 5 $ r
)r5 rV r rW rX s r _concentrationVonMises._concentration s !!$$U\\22r c U R nSSSUS- - - R 5 - nUSU- R 5 -
SU- - nSUS- - SU- - nSU- U- n[ R " US: XT5 $ )Nr r' gh㈵>)r\ sqrtr r )rG kappataurho _proposal_r_proposal_r_taylors r rd VonMises._proposal_r s ##1q5!8|#))++a#g^^%%!e)436za#g. Y .{{54<);IIr c < U R U5 n[ R " X R R U R
R S9n[ U R U R U R U5 R U R
R 5 $ )a
The sampling algorithm for the von Mises distribution is based on the
following paper: D.J. Best and N.I. Fisher, "Efficient simulation of the
von Mises distribution." Applied Statistics (1979): 152-157.
Sampling is always done in double precision internally to avoid a hang
in _rejection_sample() for small values of the concentration, which
starts to happen for single precision around 1e-4 (see issue #88443).
r# ) _extended_shaper emptyrY r$ r4 r% r@ r\ rd rV )rG sample_shaper) r s r sampleVonMises.sample sl $$\2 KKYY__TXX__ M IIt**D,<, [ TU ] U5 $ ! [ ad U R R S5 nU R
R U5 nU R R U5 n[ U 5 " XEUS9s $ f = f)NrO )rH )rE expandNotImplementedError__dict__getr4 r5 type)rG rI _instancerH r4 r5 rK s r rn VonMises.expand sw O7>+. ." O MM - -.> ?M((//+.C . . 5 5k BM:c
N N Os A+B ?B c U R $ )z(
The provided mean is the circular one.
r4 rX s r mean
VonMises.mean s
xxr c U R $ r
rv rX s r mode
VonMises.mode s xxr c v S[ U R SS9[ U R SS9-
R 5 -
$ )z,
The provided variance is the circular one.
r rN r )r! r5 exprX s r varianceVonMises.variance s>
'(:(:!D)$*<*r s + 9 B ,
~. ~. . 9 9 i
| i
r