From d6388ab74589063b6143f69cdbf65cbf70825e7e Mon Sep 17 00:00:00 2001
From: KyleAMoore <kylito1@gmail.com>
Date: Thu, 23 Jul 2020 08:12:38 -0500
Subject: [PATCH] Added VisPred descriptions and fixed incorrect image

---
 docs/examples/redshift.rst      |  19 +++++------
 docs/examples/rs-tutorial.rst   |  57 +++++++++++++++++++++++++++-----
 docs/examples/vis-pred-plot.png | Bin 0 -> 9087 bytes
 docs/examples/vis-pred.png      | Bin 0 -> 8324 bytes
 4 files changed, 56 insertions(+), 20 deletions(-)
 create mode 100644 docs/examples/vis-pred-plot.png
 create mode 100644 docs/examples/vis-pred.png

diff --git a/docs/examples/redshift.rst b/docs/examples/redshift.rst
index 41ed4ad54..922c52b8d 100644
--- a/docs/examples/redshift.rst
+++ b/docs/examples/redshift.rst
@@ -16,9 +16,7 @@ Pipeline Overview
 * `Test Pretrained`_
 * `Download SDSS`_
 * `Download Train Predict`_
-
-.. * `Visualize Predictions`_
-.. * `Train Visualize`_
+* `Visualize Predictions`_
 
 .. figure:: application-pipelines.png
     :align: center
@@ -31,10 +29,6 @@ Train Test Single
 ~~~~~~~~~~~~~~~~~
 Trains and evaluates a single CNN model.  Uses predefined artifacts that contain the training and testing data. For this and all training pipelines, the artifacts should each contain a single numpy array. Input arrays should be a 4D array of shape **(n, y, x, c)** where n=number of images, y=image height, x=image width, and c=number of color channels. Output (label) arrays should be of shape **(n,)** .
 
-.. Visualize Predictions
-.. ~~~~~~~~~~~~~~~~~~~~~
-
-
 Train Test Compare
 ~~~~~~~~~~~~~~~~~~
 Trains and evaluates two CNN models and compares effectiveness of the models.
@@ -55,10 +49,6 @@ Test Pretrained
 ~~~~~~~~~~~~~~~
 Evaluates the performance of a pre-existing model that is saved as an artifact.
 
-.. Train Visualize
-.. ~~~~~~~~~~~~~~~
-
-
 Download SDSS
 ~~~~~~~~~~~~~
 Download SDSS images and save them as artifacts. Can be used in conjunction with the other pipelines that rely on artifacts rather than images retrieved at execution time.
@@ -66,3 +56,10 @@ Download SDSS images and save them as artifacts. Can be used in conjunction with
 Download Train Predict
 ~~~~~~~~~~~~~~~~~~~~~~
 Download SDSS images and use some images to train a model before using the model to predict the redshift value of the remaining galaxies.
+
+Visualize Predictions
+~~~~~~~~~~~~~~~~~~~~~
+This pipeline produces a visualization that can be helpful for understanding the effectiveness of your redshift estimation model. It generates a set of graphs like the one below that show the output probability distribution function (pdf) for the redshift values of a set of random galaxies' images. A pair of vertical  lines in each subplot indicate the actual redshift value (green) and the predicted redshift value (red) for that galaxy. This allows users to see how far the model's predictions are from the correct answers and can help with identifying biases or weak-points the model may have (for example, consistently underestimation or inaccuracy with galaxies in a specific redshift range).
+
+.. figure:: vis-pred-plot.png
+    :align: center
\ No newline at end of file
diff --git a/docs/examples/rs-tutorial.rst b/docs/examples/rs-tutorial.rst
index 20fb93951..75ca0caee 100644
--- a/docs/examples/rs-tutorial.rst
+++ b/docs/examples/rs-tutorial.rst
@@ -10,9 +10,8 @@ Pipeline Overview
 4. `Train CIFAR-10`_
 5. `Train-Test`_
 6. `Train-Test-Compare`_
-7. `Download-Train-Evaluate`_
-
-.. 6. `Visualize Predictions`_
+7. `Train-PredVis`_
+8. `Download-Train-Evaluate`_
 
 Pipelines
 ---------
@@ -134,7 +133,7 @@ This pipeline gives a very basic example of how to create, train, and evaluate a
 
 Train-Test
 ~~~~~~~~~~
-.. figure:: train-basic.png
+.. figure:: train-single.png
     :align: center
 
 This pipeline provides an example of how one might train and evaluate a redshift estimation model. In particular, the procedure implemented here is a simplified version of work by `Pasquet et. al. (2018) <https://www.aanda.org/articles/aa/abs/2019/01/aa33617-18/aa33617-18.html>`_. For readers unfamiliar with cosmological redshift, `this article <https://earthsky.org/astronomy-essentials/what-is-a-redshift>`_ provides a simple and brief introduction to the topic. For the training process, there are two primary additions that should be noted.
@@ -146,10 +145,6 @@ Second, a class has been provided to give examples of how researchers may define
 The evaluation node has also been updated to provide metrics more in line with redshift estimation. Specifically, it calculates the fraction of outlier predictions, the model’s prediction bias, the deviation in the MAD scores of the model output, and the average Continuous Ranked Probability Score (CRPS) of the output.
 
 
-.. Visualize Predictions
-.. ~~~~~~~~~~~~~~~~~~~~~
-
-
 Train-Test-Compare
 ~~~~~~~~~~~~~~~~~~
 .. figure:: train-compare.png
@@ -158,9 +153,53 @@ Train-Test-Compare
 This pipeline gives a more complicated example of how to create visualizations that may be helpful for understanding the effectiveness of a model. The **EvalCompare** node provides a simple comparison visualization of two models.
 
 
+Train-PredVis
+~~~~~~~~~~~~~
+.. figure:: vis-pred.png
+    :align: center
+
+This pipeline shows another more complex and useful visualization example that can be helpful for understanding the effectiveness of your redshift estimation model. It generates a set of graphs like the one below that show the output probability distribution function (pdf) for the redshift values of a set of random galaxies' images. A pair of vertical  lines in each subplot indicate the actual redshift value (green) and the predicted redshift value (red) for that galaxy.
+
+As shown in this example, any visualization that can be created using the `matplotlib.pyplot <https://matplotlib.org/3.2.2/api/_as_gen/matplotlib.pyplot.html>`_ python library can be created and displayed by a pipeline. Displaying these visualizations can be accomplished by calling the **pyplot.show()** function after building the visualization. They can then be viewed from the `Executions view <../fundamentals/interface.rst#Executions>`_.
+
+
+.. code-block:: python
+
+    import numpy as np
+    from matplotlib import pyplot as plt
+
+    class PredVis():
+        def __init__(self, num_bins=180, num_rows=1, num_cols=1, max_val=0.4):
+            self.num_rows = num_rows
+            self.num_cols = num_cols
+            self.xrange = np.arange(0, max_val, max_val / num_bins)
+            return
+
+        def execute(self, pt, gt, pdfs):
+            fig, splts = plt.subplots(self.num_rows, self.num_cols, sharex=True, sharey=True)
+            
+            num_samples = self.num_rows * self.num_cols
+            
+            random_indices = np.random.choice(list(range(len(gt))), num_samples, replace=False)
+            
+            s_pdfs = np.take(pdfs, random_indices, axis=0)
+            s_pt = np.take(pt, random_indices, axis=0)
+            s_gt = np.take(gt, random_indices, axis=0)
+            
+            for i in range(num_samples):
+                col = i % self.num_cols
+                row = i // self.num_cols
+                splts[row,col].plot(self.xrange, s_pdfs[i],'-')
+                splts[row,col].axvline(s_pt[i], color='red')
+                splts[row,col].axvline(s_gt[i], color='green')
+            plt.show()
+
+.. figure:: vis-pred-plot.png
+    :align: center
+
 Download-Train-Evaluate
 ~~~~~~~~~~~~~~~~~~~~~~~
 .. figure:: download.png
     :align: center
 
-This pipeline provides an example of how data can be retrieved and utilized in the same pipeline. The previous pipelines use manually uploaded artifacts.  In many real cases, users may desire to retrieve novel data or more specific data using SciServer’s CasJobs API. In such cases, the **DownloadSDSS** node here makes downloading data relatively simple for users. It should be noted that the data downloaded is not in a form easily usable by our models and first requires moderate preprocessing, which is performed in the **Preprocessing** node. This general structure of download-process-train is a common pattern, as data is rarely supplied in a clean, immediately usable format.
+This pipeline provides an example of how data can be retrieved and utilized in the same pipeline. The previous pipelines use manually uploaded artifacts.  In many real cases, users may desire to retrieve novel data or more specific data using SciServer’s CasJobs API. In such cases, the **DownloadSDSS** node here makes downloading data relatively simple for users. It should be noted that the data downloaded is not in a form easily usable by our models and first requires moderate preprocessing, which is performed in the **Preprocessing** node. This general structure of download-process-train is a common pattern, as data is rarely supplied in a clean, immediately usable format.
\ No newline at end of file
diff --git a/docs/examples/vis-pred-plot.png b/docs/examples/vis-pred-plot.png
new file mode 100644
index 0000000000000000000000000000000000000000..0aed10ce04d423ca90d4996771aea48658d55985
GIT binary patch
literal 9087
zcmdsdcQl+|+pZA3ccP3=L>JN9h>|FylVK1&5{78eBZyv;5MA^xBzhSpqC^)$^wEPb
z`e5)q-rxIv=lpTjS>O5htXWf@=h<`bYhU}`_rA-H*4NV{C%Hp{gM&l<Nb8{?4$iGW
z;CCAl0q_~{iv9=ib<5LG6O2<m%=!noxb38>tBQkDn@Ea%iVs{9yK6o1#KEEXcJsc~
z3oo_D!QrNR^ib9KmDS&DVq=r3=3cw(I@`$l7E`m#28tk2Z>{lZ=K|ds7gyTIQ>!>X
zSNcq*iy0osPbH~6smFsV#wvrriy)erjOx9rU6zwL$FQV~1{F1-v9Ym!fAA?{WFgX%
za4G*Z#}kDOc^<T#6ExgT*wnd~)q2{92ztj6PPI%T5C|snAf{8|3~#zkl1l})#EpOu
zaf*Z@`NOHwR8X|wC~fC)WKCHaH7_r3etteUOWrr3%SOzS*`33Zi2s-7I|ilT&P=zK
zXy7JO&THE3f=8qvGpS28f|f0DB&b)9jaFJc;A&bS)b(&$e57H+mm6<3bxMpWZuD)(
zS;Rl>l?8&e6lk1EE34_G+gGVO_&>WE{`q}^b+fn0q@sE7#+x$&;zk7E+xUcz7S_jv
z%TPu@UH~Aw{R9GSNdfr7YA9NZJT$QfJ4x;|FovkZa7)aiN00K2O7!wwXB+Z06WO`=
z`8C;ybsQawH-A-SI{z$=4;&mGHq~^wD<D8csf5YHoHQbYJI;pSF$?VhvCGT0-d~>z
z1Oj~$d*mdMC@U($zPimrdHDIYA0@KMn|^&hRjntxv9=bSDr^yh!MrjsHx~d|)VuH=
z93D<ehReJ{=V5kz5%+TZ)|qR)w@p1ZM>%V4x^O*~yNQqE-hM)7N8StpEQqyg_m4Ix
z&?rk)ovMbZdwG3z@ssr0)E9SOQkj%s60><76%%79%=wm_g^xYx$Y8QX>A!IiBU)OZ
z3so0B{c+5a8b^PAYSy<sTkL4p($O&%#``MowaL+FutW%xdB+c7xV$)JT)bE~T~&A3
zoohaL1v54==?7^==IG6hQa(Zw-VoOr%1Isk>{*$Dfty?Pj2#stBfp&QZY6Xf?}A@S
z%A}rbR3k)5;NHD=8*OW;7ON39vkk7cTN7pZ>ah&ALm5&BvY}DY(aAN3<Rpu-fPQ81
zuOc-OZ6Btm&Buzg87B`2kPvNcBjK?3@9AWhNRMPADM2E$cgU8YtwDh(w%}D&6z#Ht
zyNy;jC_#iRi#Dvus8}c2bqKS+@;&Huz6456yYnBc;W`+c%k2WqL{nkBLg+7PXL;vQ
zj!B75j_Tvx%Y}fwW0tRM682Gxu#g~q%#DsLD{#_x=nZ7adKT*puu8jS=?n-73a0dG
z#xbFILyk^Yj8wtkH#!5|1#l>zlK4`Vq~jnJAq6WHy!|*eaHz$5dqVQgTcj_bW>g!*
zW$0UcR^B_)1{7_!%ab$p&nKxl&rkl|(@0<~gAy<^GuI`83^hrdOifee3`s>e8s~j7
zo6goV)lYZlzqk1vP9!ho{YMEzuv+2&mis?fa(zrU!wtoN9yI`~by<N<sTTkLZJqr8
zAmWV)hSzDhw4EK`;zbz1g$@1HfOccWKR39AkdViZX|m<KX#h(=ujqe7B`YhtrDY=_
zDXHG*Hvcg5f$OTBk57Z#^2j@g(p>~<$~px0_Zbt(qF*I4<Uv5e%=V3~c>or{9vyj&
zwRk^I&dL%ZAtC9WuC-Hd_4)hI!=u*!t+bz?U%taYDwmRy6&(Y^k6zO%XfzfZcpL{^
zAI`ql;=3mjjQIvK#ORV$pmTG+Sp7W)=FFCblQFo)s=d3^pul(_53ng*a&pl6HfA}m
zQYIPqJQrp%CNY(V>gr{wU1J5B-xGt6xN8GW-NtI|G)WNAPwB5VCoJoy@PMW38g7i4
zoSZyxn{U*19Q-Jr56M+3U(>xf`KxYgoA1J$H~&eEaFQm!xLD(8dkUhHs|0Icp$RrI
z`G5`edjI~tK8C`M{>oaZ&TIckxV2vkSV7r0%{-*kl8q-koJK~>9w{m;{DFU9sD7@?
zFDf>cHe24;R2V*b`7R@)97^C^plluHmP5$?N+s~}c&ctb32kR<yLyUZq3iH8wv>aD
zC=2?<&un*bVZU!v#w5@(L@D!Iv;-1yaspHIe=!jc;0r<>*Oz+(p3h%JE*@dZ+ve!5
zzU@^S2Pm3{D+lO)qPtRqhm=v!fWgG9OcJQb0&PND_*i~~LJdXWw?#~kbRi~;a~hNq
zo~WzdtAnES_e}p6jKtL46eI|J4)so=qInD!41BFR3vgVKddpCF8!#RBos(RcZK%vB
zXHo-p$$bCLtkL(4r;9y(Gu`T26R4*K*px5!BTtPmmsYUK6aUXd*MHAui_jPCb<i8i
zQj$omOB{x?WQ~d)5?G}5iX9l}>HAl>HR8}I%I6mzQ#naV96u%Y&CQ_2Z=cBNCQguF
z=%B7&ZsFEXg{fm|>|Ill2zYtXKqM871Xv{Sb+lQM1c8<KzRTa4nZ(VYixYZo%>-lN
zu(&v~-@S>X_=YlH>gpzvJM}TTiPC0&??+r8_8(4KQ8J(qM^8HFf=BMAih#gceMGTn
zZooSG*{;JML@fUO?&nfN7>ni)%;Jv->5bK%Pz7rV+!!G&x9|vEaEa3w`xw`rb(-31
z{q(z7#7JK<65+ZGJ{J6Uvxh_Gt|W-daGhj8SAHl8dNc#AI@+PTYL*q^>y}htFtD+_
zKwV9x`4fkn^l%&QATqyp(p`xqV(ZbOCe}tZ09ho!#*LgeS`Q$;{%b;5IK3C0<3w`Z
zRi>?j$}Ouj{dbe_?2L>W2SmFGMs#RV`1``~9~~o1)9e&dw4$lfN|%V3o=%BaevPrb
zaJ33~tvO?=jJiH(9??ZHelROHqXmQUBOu7(j+KqeaM1CUzcOm^Sf}A`VB1?N8mK@Z
zjn4zd?0wp+J!^{<{=gT{QF{Wi_o^)`sMBiP(Wy!Ls;GCL57$DCf{}Yl&z`j-HFdEa
zqBS&#$L4oi!-+VhDQ<)xDE3~4*L=51LvOgsySQEdUC~{K5``yi14jC()78ni2I+Ny
zC_<u<`iaHkiQhz=Q=~V4%hJ`nBaEF&-IZ#j4!LwLkYnLo_5Xx^G@bszw`mL+;<l^j
z$s+!-!iM_gH+>AxP8;IHrXb?l_y>b0%l`JW%4A4iJopA*Pt0XJL&xyXJ#oMEz_v^L
z0l_Cg=Wl?{`JZ|`Er>|;cHNJvW^GTDJqw>>x}i%vchBuBdM{c^&`D{(pYI6>G^;sc
zW$45cZb^CboL0j3J`@8zfRw;#+n34C_Jqpx|NEhx^6ZQ=YFnjYIPXu-WAloWod5L<
zGCV`4k7<fMrN4sYw?}{wpH&pjLu;G-)*bB*O#0Jfmk~~`46O+V3Asr<Dgby%+x54v
zk(%X2HE3_sgoHl>rvnh8X$^n=(RX$3oE+h|0lyD9oVzeYDiF8>NWCnNDNb0;QnMY$
zD!!BtHV9Q%lw_WDxUkNJ2?&Z<C0`J6lB?a2vShK1OL7ed!H=NJRLJX}0e;0h)MaPi
z*I!0mXMKU??1_)?69fhB{wjEB9b^n^=M$``E@$A}jX?5q;`y#10=(;Gf`<eAz;pG4
zG)gTlm9ZbyFq@fE5}J7~>rN~&YbPOP{&I4fAn~oALx2dftQ_uO>eehjZ1*uRTlQop
zbNy<_(u8ODQh)J?LCi9gx)xi!*iE6mJ965-Y5koI)g^X+H}+1D%l_&KcIVBSde+7P
z#2)+r|I97dY`&|e{(4Us^f5+^^fA=%$|IVqd)jRk_O|<Ul^-Kq$qr1dtHcxM&z5P<
zn)cDpjz3_#O2{v<mzEEbn^;M5!{aWBtKZwIPbGYiSFokv!8V%iGD3#>_GzvNx7)Ut
z;KEAW6uIT^XPYv6Us$~0E~2%=o3=7=w+!M9IZ4fy9*J<ECH9cK0pnXy%}{9Q^4V|_
zq0lOphpz4GUp*qXLN-==MReM(Z-`in&o&c6`ZX;r-;RipRHCH30kbd{DBSRG0k9lv
zp6_=WN40Wqw~+EBp(P)rMIEe^=%RW@e)BZ2$v!8!WAJi4zs(uzmG7~ezD3)aR4D3@
z&Gw>Xm9ea5eS~K6I?;S6y`_5%AeP6n;%tU)yL%_wrG|<=Ew1%Y1!>(@p%eAvu@Pok
zL`dE#htQ+A4w_TqJ4f>S3L&dOp;YIm??gE34>sj+ge%zQJm0Uja}nj-4M+0dYJORz
zk%eHz78aA5nCaGX%^P}xDKj*tNUp8<9Q%W|PjU}vuguC3R)jA%yAbV9Cw{8N@UmNH
zHEQ2%m2<(azm?9iRF-V>Mn1HE*j}db1(ff#LQ0nC#VY5Hjf_Xf1D16_2@L}x1O%mT
zFC0fS7o*`J9yxQ7yDZkIv~txaw;jjx<q0<rNaVzrNuMZ9_vp;Ed~BG0MS=USiRfkq
z_SH_(m7R{#N1)ce*G=@oKDSLP#y3SWcHb#UM_ez>4yZegoIwg&do(fAB_oGlKJsn}
zA>m?-qy^i|O)HW<rA}vIyS$fo>~W}<9R&iHBkKb+M%-MZ-q_U#?P;Y=Y^qK#?2`gU
zx5ZsAm@5fG1mdMS@*9QMs9Fc}Yx54(U+Dc10=-cF(4ww50Al@F(}n+rN*Zvjm171(
zGg7*<F-2Q%wP`l{4Wfc;s#7n-2Bh}#k{Qls%@{M0HI*k?Rx8}tiYXU<(?5sK?0V&}
zU8=L=rh(p*K1%6QCrnuN-OD>VTU^~kfMZIzIMMeZ!W;5p+y=)~8FbGWCDn(}486a@
zg6XNt3#UZOwXG{JR9+V1ELheqV(&DqSbshIICu3;zIluJ3VV^+!y#&W!03H?OuI|~
zjFr&7iZ}Xm@ssm6@-M@yeZx8jp^ACsYl^Gc^s1!SCyWdOQ)EG+kOnDp^z91|))g|@
z3&@WLF=dnZTbLzTqKQA8|G=6z$>*1!<TYMhxW1~FI*|E6_q;a01~yeWMJ;9LJ&K%?
z^udb<ar|`r&Avk0o4P!6N8bC>_Fs;kBDm4bqJBypY-pIWIZ^h`dA6ZybF3h;ufN|w
z6Zc(u`duKz;G6mSj2raj%NHP=sJ3k8>PZBN?zSE3c?<|QMp?kyP2!lu!S7T3&Fm{F
zNe~`wjXiy^y<F$_oE`i=&2r1}^<uwUx1{7MgEpcNf$wR*qC=SBI(6p(iea@l4vc68
zOAg%7=G(vaZoR7qR0?oE{ry{<=UeKv^z<0;@$qeUrfc(Itv>lpGVT;!R#rLAzw3&k
z5)&CCA|eiIO)5;elo!oAs)1y{T*&pM?a9t>+r5Q$+spIgeBZr=e5a9|gI8kun0UN<
z+NwzfMsYJo0>;IMs48>PcTc(AFKA#LX9R0bHYeutA{I{_R_c$7EO233<hc)V4T3-Y
z*beAK-VJX`Pa;K_)mYh1lp2&fjMhIlqh@2Pf-aD;$r+Qe%IMdV5s)!)00{}>f!uf&
z=^yZx>ewhzy5JQ|iJ&#Q94gNcYOHJh%Yx*cV)fR{W@q8u=K*s&$Eu{(qV5Od*!emb
zW6KLXT&aLS*;fZ|{8&RLzur(mAMEX2-)7P1)BjrBt|v+<=)5#l`tNKbV{~+Mxx;Ip
zND4Nf!?l65SO#I}#F?%c5I)<<J%0RH0IPwXN{k!K_)7lWQT6`zGG~(VF2;aVbAXOn
z%3oezA@gsu7UA8%_N2#Z&))x{(}Q{T#z>jaJ<SHGM6rb8sqqCKRjk-IH!nPU>EYp#
z@3}EtzDC{>VrXP!GJxZHv_ZIhur^R9zmcrTZV-S4;3O|b!u2s-`=hfO9rA*@TUTB%
zZT57Drhvq=z{chy4&2zR&A%$*e`1AINkOrhD^guL?@7E=$1#p9PeU2Nb2{eXiNSvN
zYHMrrwbMj^&cbt3Q;VSlckkYP**V2Xwxo14s<yn_x>tS5vd(rD|LJ0}Hf4gtq*5o1
zLC>-Q(qJ$+=$qHj@v*@^U=$g0glE^uDEW*9Z0#)eb1JP;nD^l0@Puj%AEz4u1z9@!
z+wUU0-;~S}59hq6VEJur9;8ohU#QXG32PmNUd#SAw|{7sF(IL`p1?jw&bMIREa{BW
zbo!fAL+pr5V_%x5vq@Kj$y<xIN5$fCy73hv37NjKDf)kPVdj{u>rkCl4I?N^y+z<r
z71jVf@tyOCl<_+#oMBO*3r6R$imyh=1k^%{HEdZ9D$CAh+L&y>AK=wNLE_v_kWpkW
z(=8>~(zaB%NG7YZ2km>^9`T@LWVkTEQC~#OPYI{tth+xYdgEF6e@ev0#FfYPeQNvx
zy1c{~C;j8UJL?X#`5%`vDZKiO*-s1+YJe|_{P^)>u#-Q=CJ!A}Qi<%ViX7#%gn6<v
z%m2Es(+yf2I(Ff%{4J|oODBI}f4Kmx1g0dvu)~nrZy{E1_2?%V`2v+g$F+$)NI4mm
zI=d?p*yMkNn(4GbTN`|e9Tz)-zV{__n+$MpbE}=~uYg18hAS(DydOq=i=-TYi|JQd
z{e@XphzMYpe#2bei*Cw(|M;S4FC{m#)oA$T_lAU4E9lmw#^#4IDoUO^u)9&RnUvw`
zn2+(Xhb8bM$k^l|(dVA(E2X11g5gccpNW*mfnjJes)ki|MYi1b09;PLu5)BYPP8F`
z|HUC&B$5kz{Ya8QgV=c6%J<afnc8KLu_akBy(l<?1=z8`qW2|sg1CP^Zj2l;7d;1}
zBe(54@}bM3AUCl_0~PXIy3g&Ec>+KTsCqvO<hyS=>Rr&waspC@`$&Bnh@ol(qi97d
zeUEI0lnS_t>bH0v(`~vgex_KpReW+(P+)nGEa?P5n8puSor<wA&pTn-{P0D;kKFsL
z9$V`u!s!NhaaNlzNti0}Ke|UWBm}IOeM}*z6BR9ksXeg!lKdmO)m2N06&VuLEZ&2z
zgb+Ky9?{bFc8c1@L+%7TK9g3N16=gBPTW;0Zb3)fK^t`oLWcWY*{$BWQb%=RM9<lS
zn!<ClJhg(W#B*Iu0?53Y9brb+5I(hLeDEAe1Vl$aQ^PlCN2o06qH=qd%jz%?x06R`
zei!i2r=VX*eI_W9+>YlQo+Eh!VGFm8Ln=21S-r$>!M4r@@Jj46siXk{Xy?h_cxel{
z;f}J|u@$QK#CprB4cgWhJ3;R`GklmLQLsejBHE3Eg%>Alb8=kmiT?O2gqZ~8t-0*e
zf#-+-5Jwrc1xm+O0=FILK7&H>he-#9(_M8s#^-qj?28XE)ElBx#VT*LF%=|`zM6>o
z=T+DZ`JO^CODdeYgi6<cWPGJ_s*-k8f9@6J?{#W(Qsw=WdZJ77_tO`4Yh@{t8o41G
z8w`F!F63!w=|FF2u%egnr&fZnp{CZU?d2`OTsfH6NilbwZWPNWG>oJ>ZbE&iXzRJ-
zR6Kkuzk@38h4|~E<K@*&?x}j0-OLul$L#5Hx#yPm0bW)or=&$yAYaG7xGDF`3(bZ%
zQIjl2<Ox8fN7%&97Yr8JD`%d2#XCQ<7^{1z*G|xUJRONWEin$2Qk0jbLeVIx#OLxm
zxzfOLt%Wo&gVdv&N<rB{5B~+E4=R;TG8Qy$_?y2bwovz|--*!tj6`Cn-o7bsgmnZ0
z4N~bKm}O$BighPhOdA~HmGC6;_PKu^@Nl7nfboIXnT;(2r`v0!FK5$01lB8WnP+#W
z7G_57pIu)t_I!Hyx7R}*ix)Z#G5#U#a$AYnfbp{IFo(5L?7bI_>$fdM)r?8`9Pxai
z$M5~V&UM{#9)6!uAdoRXHUm|C+7Mm87MJWmyE89h`Vm<<_oKXG9lu+eaUo_Nd8$6b
z(cq)4Ysz?wOb(<N@F*41)9ICmKSunFLLVdY&}n`t!6)!-ZHuuQvYK*rK`pfdf3Q+U
zIB<1k#VL?S)|x`Vy4lo@;=v;Gm+t467K54`g|c8&E7@H8R;&NqB)gnnSJvI_%qKsY
zhM&CL#Y25XdmG5GmFB{$=4)Dus92|Yhv-Qb9rErG7+$1rRGT`j<h8D#0C)XFugM*X
zX}KLn(e55Yaal$f>aVJ$_`L)NoIy+1G0-<-JJHppQk4y0vEOz6#9cWcdQLEI!Wras
zoFy}IepYy<0>D@uQgiv$r7aJL`p6ddl60?H)>-ckDPQWYh>;UV;U6`iO{FD%GH2t0
zUKT?KJ&Xi5l!nTm<*DM^PPvYMSPX5o4DyZbg33qppdcC6t*>LNuM<_0WwM6~zR4NH
zCKXde?BRa?K}+!5X^kzF<{PYX{$UEbDmJ6wW5z;HxA+7d5~kSF_IbpYR~VcjK^H7Z
z_$Q?_Gfchh_yx3*#z8un2WylAk$a^ad_Ko@R;F1}1RZ!>xXRccgt@Jv<r#{EuiLz&
z?=Hdv0@UHF!FO7KF4oFS(BYbwFzyWzZg-@(etwa--$CxP^plDc^#*9ac}JpU?`ovq
zd#aU~KbbG^-nG`O;35l59`L(ug%+wARrw2LRZ!BNUMAXJ1~vqLp5{VNeW?49H7)+S
zBEQ_$%)j%Vb$r?;eD|5?MiYjZgswhalt5%g<L!7;B0xlgb>A=EswgXS4&5R_FuzG5
z&V`G9xEn!biZC;);2s{E2&epP&%K~*FEdHFu=l<72P-oRXQ4PCiR$?bCl`$&X=Xm!
zJ1CpCb)Xx_ceHP9Ta4kYly*PT;!#6MtL0=quiP!Po4=a*P@a{4m&7^^^6QBLBE&U*
zIt1G$2aT=%8hv#|Y7#0+68;buptLX~-@_@Ry}lbRHhwoFAARk6;&Y}1UAb||qhFX-
zevsaO^g2l^0tt7t$uk`wq>bZ>C>Frp){YOXBsl5y^;|otp)rm5`Juk3FuA2Hm+|n-
ztv)TOFANDs+GIaj<R$hFd5JK~0+rT}(?!5x_t*NLWUhYyIx<jU&$9OJ)-S5j5ml@R
zE<qo087YueeKSZk5!@AC{&a7n|0S{d+6UV$u9a#!ZAJ|mT?5RG!4576Alng^hwItS
z`m9#=$(xHb2IxI;{57pfq0E^~6ahlL&uQn$$!XW!5vkG-AoF3X5_7O8pNQ;Q0Wv)k
zlH6ec{`g49z$r~wEnps(PD3x72Bvax*{|*2F=0>AUb5VaqNa(~$50Q4>)0jlFLu;H
zhlwWJ?uRtaDIa_zH<T;&gu|xsd@qt88qq27St|k&nMj81+EPF1`g~h;?sbOd!C_<C
z4_DUN)n;=lnrDD0G6vS(mjTdac-5Tf4?&TE2?V=~;b=LMcI!7-r1vyg^XjmH%_M*;
z5&(D3bdhclqS=eg?p?j|_|r45Ecouh*U$2tK58gUfzri%LYuK7n-8&Wr&)EDmabf+
zuF}$|+|xeR-0Gvf)q5rdL>D27Ljr1wJ)Q;}JSll-ZZfi-d=|p91%YTB4~|I)1b4jN
zgK9sPWTJvAcbzZrwa{I!)3#b$#EjkN=FPa@?GHTX%?`MdnqK6-;<IbbDx@!MC2zj0
zIQ<n2<>p}pvembq<nhV#X=E!{^S0uOF-2wUKsV(Y7;=+GG_4fbf&Hx?D%){4Jz2+-
zE;wk3D=#F(x2>sr`_!+bBtG`lUK9OtG~_`);?^ZAtROeZor(d)&&wOFmn|3F;5r+_
zuIL|?mBs4KBbJ_?o^R3Umam)n;6Pkn-hxfM69(7OiEBFkW$}GTdQs@6=;O0T5&TA*
zhX@1{6@XwPeP6CA%QO7yRF)yY?8fag;Nto%3M@z+wJV$*N_)@o1$y<PcCCF8mq0Y+
zN6nOCZ+~F^O-}DnW%*9JZS3!Hrns6WFLg)zr+?JJZts&w|F`mE?3*{Z;yyd(W7RNu
z154Id$JcFb3S)H++SDMBC<!U)nq7ry)x+RmSI+YA@Z0-U@E~n_2Zv}Nq>Fy~?d_wp
zT;)CMDIO{mmD1HGP7nLxCA|wBhx$HcV-Z5rWbwn#8z2_H1E#i%v%;bSx8ZY0Lg$Wd
zdpsD1SHe=i&557IlHbb~<k{(poXV?M|9PzLh>{}fF1SY#!UDKd+5NFzGr9Op=xr56
z5E+@BC7~xSkZ0j~&ckU(MB2EN|GJuM;{5jp@!;(qg<A@62b4$puh<Rto82p*y0S$O
z!HCVST|fj7HKPNLw|F(;?-W4^0POT*V`JrOL!+P6zITS=nGD!;zP`mJDEQb-&0iVQ
zMOk<xWnyBoEWdb`@M^F9bb=l3&Z?Z0kdl(JO>@OHqRezY-l~K!zfxOx;PELv!B;U<
z{;Y74aJ5iiZmmE=@k(`{GA@?j>ekrMPgYX*W)lp#cu@1$aac#Nz4=p?5WBG_^6I`L
z@whX;Si;;1Iz*I`v0lquW4{ki6Kb|ura=PeYXj>g*>hG*2-v2|0UIJ-E$VJJZNAi*
z^gk)lQnJbMAiAn7+9lZuDA|SB19rrLO6k1vt9bZ*Q;hYF*g<(1`P@*!J;&6I1d8La
z4L3!>!`iWpoG;gf6^wv?zO7LLU;a4<<RZVPE=$(MLlT>edP~9{9iH|fL@RCH-oo}e
zx+Seij-Dn$l}@`w(r~}4gMF!=vS>q>-*HQ{xPQBqesMJ~k>QwUmRAc@l8TCo4uF8P
z*(Fic(lTpvuKA0LU)XN$#DwvrgbK^$8O{b$*2C={L}5RCa}j}X@uMx551f0pVkbc`
zi8f%8MK|5+UAlWkEq+0BU5R-$_ntsM`!%wW(Nbt@tGp{Xj@Tv;r~Wm%F%I!;aHVrm
z#?dz|t6AWB7Y{!Hp9GtDn}qdHQPbm0{R|PI*>j8^42j35AZtdCiNbjwJTUe8Gp=6W
z7K;8kK5jg~!ON>ofx#8V+V#do3up9#LM(kQZ>xP)<{<{%91fs0L}@-wOAw{&IJ`ML
z0MUw|6mypMKXL{}Qtl8G)(}QOZYYdLwtP-8qnG(Wx8atE&@uVUIT$4mB%J?smOxI}
zbA2#fLfCVz$wNZ;OKGVVJ4oC%JPkYM5C*~TL#jgg*zt-Sz-Y|s>#u{TZ;|pjsgLgb
ztDp{?M8Kh~k@MLRApCdjX3SnE<BsghLKkLxd;316U`v4{WOQJ_J^y9;-J`L_#9huP
zW%K`3_;13m($dl-Wu7{3beP0WqFqTsbY{LH^7MAYkX^6BEC%nP&FsM)f-vftQRY~8
z+o1c;P}`uCn*$j=L-3ocsfmd`5S3ZIidp+P*jfn9v?=qAr~}f4zgY4%Y|p^mLvfFm
zHEQEq^zm3>L<skiE;2Jl(@;7v_f6QI(KoarNrHO^gih%ivlX#t<c<z<Vd~KAq8>ht
zMs9$t+tEQJ6de%qZ>It>$TOv8Q%vbV*n542lor$l9cy6jowq;_;fFyltI))p-LAj{
mP%YE^pU<6Ct3+KBM!CM-$R%6&2ppQhd8DrQuo`Uh=Dz^gN1q7*

literal 0
HcmV?d00001

diff --git a/docs/examples/vis-pred.png b/docs/examples/vis-pred.png
new file mode 100644
index 0000000000000000000000000000000000000000..9dae7ba222001477b2d9b139bbda121b0f1e588e
GIT binary patch
literal 8324
zcmdUUdpMM9+xN&S6&V!6GPc#Kl+DeuXKYI)6hdK!Awr0W3}e$Odr_;6DVtd>X)4U5
z?8dejMhq&Ok!=h!82iaEw%=_%t>^u|@A$qy-s5<W_j%qw=D6p+cISCso8S4nes}VD
zTdRG0f8Gm$K=z$EZD|jI2ynrl>F!-%$;%6m?tu>hjJ=gPgwzh71YdsgHM22;K#DVj
z*FATF?|ZJFzJ!55L>sn$0<_?l-VlhQ%o$5F#|Y$1j!lvr?N%HrCJDOtsK@SFnRA<>
zfHF~k_u=2fvfw%xmxkN}e8n{A9-pG0$Yw55zeFU8Wj8Ck6n;I^dmE!+SoXGcH15Qc
zL%mA9Et_J;(w*(AQ*8?M^43q>>b(#gcC$O`Vs|%r_`74+&5O|rxTBqe<SJC%>?V?N
zy1fu|F7Blg3w+$SpbJ1C%87VE2*fg7oEy+zcU%n_BM0mqUY!ZUvJysSICVFJV1Sg8
zLK;3(VN<JPtD@H+RWGCwyNar-R_Pefs9r7DQq3KYe1BSXZu}m!9e>ak8S?JuLo7DD
z|5Tb~vk*Ex<Sbp>FdnDKk|t8&#SZ_{q9hfTtd&pWwXI}yC(uEZwGusGMF{V39n@<4
zyzve_(<?)(&WWhh`!!24(U!uZ><i1*$&fqh^DZiNAFaB@VOfnZ=YVcs`j&0-VyU*Y
z&I(2I2vbr>q6A7FQqWaLNnnUnqfU-ziv&g*iVm`-yB=nB{*Je^L5ng0B%Oe`P?)?w
zR;OI%)|5{^D>bX<!Wyf)G+nannJdk|7#IEd5$t8f6S?ET{^Q>p@3!^U*)yB`-78dX
zbQ4k;H>~=^F#&SCHmAX{0%rc>$%;wmIrkIqJmG@e2yHLqTtDt-U{0N3F>Xsv^v|-g
zb?qk&X96}Syhy|$LL)B3sFfhwdaOIVTOSpWK*=TAIndtZo}w$knN+tS$M3@BmGdR3
zb+uwJ-7Nmyld7HDj*XO5QYJziVPq^bfMl(vTPITygk%LwFHblPT8?$x-3TBNHz^eL
z)}bio%HEl!r$FYMGm2PgRzR?LYEP%gzxLxr-3UNJmjX(bT9zlDmL<KcmY?S2;oPn0
ztyfsIH^oDdGE47P)1+tv2D=1z^EGVga%P%01|ohoo4RbpSl4j{4eq=;bp9c0Ut1XW
zEsw@W1j^~EuO)aD^<y2Td!3@0#Tl6#mLrL=dg5Vs%#Fb`<0tm?s`iT{AsgC~M>d9p
zaj<n@*pvn?FMXE51k`ebgO=nA0xL|kxV`a`i=NLt!?B((_YddJRrcOtwLSBa%i+3S
zqhL#iJ*4_yxkYkAhfTvnxnGKcy;~X)&YiNl`Qy63PrdLSfCMasz4}Ieq<f0S?&MK+
z0<S!I>6k|xv(AB9tVoGp6F<0VVCm<mCofdzs``Axa8p-$c$ti3j-GnmE(<W&QB>Rm
zSccx@qbYNsZbt3eE@HG?LEFuW(YAt0rI$-B(VD>$J@K4L)<v>apzAELw4*KS(&VD-
z$0<O4IB$0(pIEbx5hZdZG*%#;qx1-$N&7>zGG^dxk(Wl?J?OzEH|Ig`?|Rz7$l28M
zenoj+n?~o(mJVGELaOTK3=URH>F&r!)NBqf8tWab>}9js3QLe3?dYJOj-<1tHXP{^
zT#(s0HCAO0WgygzB0oU;1G|~UE?%p6()w#y^_)F(WrBfsBNWSVq(v0R8Z6Ps*X}3U
zvu<-n-vm;=IpwOmjkCt9DLt)Ab=cBZ!NWCM-EC+3ec5fVy<C^t++v6GS-doS6#~i*
z+2j2MQ14DRX54;3*+2lOvKbP-pWCMXVM&t})is<ECbCOx(Ni;4pj{+>Z`UZB|KN7z
z^~#>&dYx<?B)bCpQ5aQsl(W&jtWRxg4G|>wxCZxEe0bE!(v_!iPRP=nuMQuTrF9@g
zNz2kL)U6%a(S;QGhb-O<<_04061_jBZ0NRM&0y#g{-!ps9QgtZznuO7Vva~#_C~tX
zml`hw3|&17)BWjLg@^1GsVqS5__e&r(Rz}4IIXRq3t8?`R;&Ly&;>>!W@W}WyprUU
z1ihH@@qyRz?-rExIa|HVaEfhVxagi?wABqI;5xS5=*p!#>p4^8&$i7t)2Ie^-Tb<U
zyErJxFEo~so};r&9<OvY46m=)q(@(Dy;}Tu?$uD)Kjy)_VOsA6>AYRvSmM_Z7uVY7
zF5E<sVkn43`c4wuU#LblZQ_E_eQ&u0d_##6%jFC`ON@I%JLbxTuVJ(MF?by+ocR!c
zFnllHln&S?>!XSHp#NS)-_o#HL5u9S<5&YrIw9nmEU+sXi<AcH$~LN0CX#@ao?vpT
zmaa%JYO4lQl~lZ&??Hvv&krrvmgC!e8ms1Nj}sy9b8Ze@a))&oFK1`(fJ7^?1WoAz
zPsK2hSTU}EA`4=Zq#tLJ#D|Ds;(npR-z!jYkR*N6{b<o|P&5KUh5xY!5SkYP6d-8P
zPj~QwE4T4b2q1LToG!3#Mi+|13;qhG(oMzx+F}2w_I!l+{>h?ofk_iTGSbwPU}QnN
z^$+WDDf*)itNE;~aTx>AA0FFg>y6+v&EXB?n56<LToC|6@NugOzl#?<v<nbw7D0>N
zlLyPS!172iQaTv;q#0d6^EO`axzPXCV1EyTQ4mpSgj;#zrg9=HE317bb9S)GS#2l8
zFdi@HUH|r8P+;K4{Q9>wH8TS-ljl)KA@<<3c$>SPPK;3wl2wwP65eT=0@_Hs9)Dxv
z2%525z+xIw1sd|cbNso5iFnni-O^~GUOCX6gVXoYmn^Skh-pwo-xzNDjN9ZA2v#==
zkD#OPcy#0%hpIrnZ`*D9_}~azNsg23qM)=gz(yP)?l47dcc|v|1zfZo=H@OpA7jps
zKw~TVG8VTUOG3i`3Tp`!S2g(}YwTgj%GeQ;OEW@SJN4oQwNBGxH+bP83yYkj2jxc2
zw-;H~Cl`K?kdXT-Y*z?_)p2#(u@-zH7ka=ow!A|>_&kcKK5VuwBEcoNug;A=VaK{%
z%Q=}%8rKKk-1fD^2TL}7T$XGsr{*Uz*K9`WHcQv9Lq<<#e;m00SY?;^#{PPaZZ|k4
zDx7{0Nc>66H#8zi|CAEY;eZz1qu4$+FNR4oN9Tg{NWv8Cqr{SqryAKM>mQ2Kd(dD(
z2lggob0sjRAtApcCgQygNN|k~v+hAg&xt|-+R;NS>{+^!0QCu69-V&SG@bhpZ?Pju
zOW%kJ&%1|bAN%h!=#as0x_=4ikNvav2Aw<EbYHOXGjWi+T8W)D4VictmHXDsCux~M
z^B?wEmXb}!J|fP#BZi3->UwY^AHkXDU0Qg-DJ?k4*!nYI=)gZiggISUgmZ~@2LeW4
ze3j%Lv!>IQ+ZygSGRprsU*_o9hxiP2;0R011e}(RKY+|zSx}gP`T0iIHki1HKdpLi
z5+zN1mBfD>_C~6eEN7@i@SRRQ$YbGVN_hyleqgmN1MXwu4wq<EFu1toS38v5v+}#g
zHHT?V@gke2e$>B8wrNe*B*TZ^jcE3x8dU@W<QH4rg%MNX{IAJdMXWaBu~IBgBdS!s
zcs(`Lu!Si98L9Q+hl?^NSm{SDuDz;CXDW+wKfS*4I%49MfdhO{omJox-WL3Bp^{bm
z(&hC=`Et#|rV~!{W~1A1CLg$Zu9L?YilvZwQC|Y#uFuw13JBx1lXH|lQ2;vm1)n;}
z8MTX!lngZ<Qf}d_hLZ4^go2GLroC&+i-;dH*x>=({HgPkjERhwjF^_1O%<!$f6mo9
z2_}FVv91q7lgQ2_D+NG<;XaoNpj;9u$637f$TffWRHk2yD0e&=d%X%BbS`tGjPUyI
z`*P}lONN_|gdO|u(1P&p#$h@Q4qqj9G55HfDs)TK`rVz|UxFT5%{T3*MLr32RR#J_
z(%YC0c3M4o<f=eJ9_)ZMrb@VgsakigYi_}`*eRi!86n+iahjghnQ3g?`i8-NdHoW7
zW(h=U^7dwK*xaedze61KcD&QKqc5#})Y`0QZSrk&t%WarDQv=j&85aRi#X@239_Qh
zZB}F~VMf1Q)rps_Q_mEb!)YF2LCnO#gO_lZLfTs;Vrwm(c7&UtG7xA6uo)H|PgQK&
z@pBsXD%1WsPTsO8v9YM9PV_dEf;7#n*2K$Dm2gQc8H-s>5&wk+VJd@XHVwwLS9NSZ
zTDnehDvxxJDluSZ_KgJWd1e@=%I#MUpm6Z5Ufx?%e{PO`5$87HFs`TQ?u4BHbk-I<
zEAsB=Akd>crR{<Euy1simk`w$5<AsTQ<1mGWfbb6Yyu{Fb)!&j0@{f=MPg7X%rkCl
zQt8%XUa>0iEA4fN(Res}%?gb10lr|jiOQy9C#zxB{}NhML+`I6V8`M%AI-8$53|fy
zKEbQgB3NA|$vhzOgRW$^j+dIst||Z8P*`)ykLn&(Q?^xxL?iv-t_!&{pPa~-;K7np
zd8-R=dGEwAo*TJikxAI=HxFR6qhsW|whqmlTyR97K!BBogNWoi;2AEG&p-U^%I0|`
zp63go1Nk?qaeIph1DT?H!$U05B>wk9e}k5tss1!6!O1!^zWO2FP?6R7;9mt<|Lqw0
ze;Rb86O0q}+*>g>?%@eIf}x6D_3(ioNY7>j2g6|Mb?JK8R*d%1;9AE6>T<MW{JzKA
zx#h||{Owrz$cblYoh(VH$@-h@9U5?=9f8zd7d1<<vK5E}lvwx8>7M73^xu~#{yYB-
zU0+$4@`;@P_F*d4D+XB;c`<h%nDqo+>;y6%F8Uyc&H*S?#K=ypjtCSG!b?$&?Avn;
z84>mkM%7C_uF3ikX^?L;uY4z(xW|8A=W|0_z=Rs-2NE1*kaxuG1B8-51_S{aucBs>
zKBRZMHOPQGKpX#T5J!`bDoi1^E)rhviw{_9DJoK>9Ds;y_l-O5Owtp)9621l3^Jd0
zwY9SYi8WEnEr%f<A|MAjG~Hj`A@fBTrq+D8{vCxfIgr?p2yOklwj!Y~5sp^>Iy4kC
zZE+0bLQIf5Du2-KLX3%3)yw<)Cdql=z+o(VEoLl4i;qt~3s2}3XaYqU%R7j!m{!&D
z=2s19XsH%eR09D<p#1$AuWGi^+Ez|p+kE~O6leO6wT9d;_U*`3d2WZ`KhM@Q8vhhY
zfhw%^(#8%zq(M3>CR08b?kf2O*b=_(U$K>LID5|l;<3FsUJpbNkCv(5o#j2sOo@$}
z$^yAhWhsBTee!nX#zp?fosp)yvW<<c<b$^Dx<jjXjjRK<dsVLy|EyTu__Jd1gzbld
zdelbh`t|zxS=6SzQPeE0vF8$Xa)r({Q9-U|5L(}rB;W-L&&qH$<()geIl_;!I*c;(
zRFm~V7PpNH_*Wp)R;S`3B`|bHkZHrv2;*dZ9~D4oLKq<AQsG*U@PZ%{w6jB}KRHVm
zST&~;j{vlDhgsW{9)u6C7ZRAbS~2eA89LzbUp=gd43`j-??PxxdD5varCy0-T2Kb9
z%BKj!I9w5aaN*IYO%%nc4rx@GH-6J^7KmN7EdB*Swjjc}xXVH8X20d?p<Ff58Iudw
z4A{pSH4SmGvQ|{>FO;F2D(rMz1#ebW2hLTo8?3_B`ffXaDv6TvdNvjMlgR4tGbD6r
zFcqgKl+1rJIuWx#iA?AFe?H6#+z$BMAy#+ioZ4?yj_4p$x^*)47$9^yKDiOMu`uUk
z;%qSj#pOp#iH7Xo-{}j*4Y~v&jOJ1b2huOIKGWJ}DigjI`1J)$ZLG|q*B%8lI+(au
z|AJ3$tloP+{2F-u1Tyb0IB}tJ0Zr8bayk7Pg<h2|$ka!r>UVqlxoq+ylqjnc3hHM)
zNZ*MhtXEC?vN}4(F_z#SLhd=nV?Lw=Mc1_93%m{z#W8WG@n`8pSQV1WY^mlL_l;o!
zdpuH#>`JXuWwj~Aa8q~Mu{HvYF0@>2yR96Hq}~+2TzZaSU{F5NwctI}3!|i6F88Y8
zmC3G+cBN3fYQnr>sNJuZQm#m?1<YhwQ;^KmlalPlGCG)a$>ay%>8U@2CgsM#%!<5%
zHbn=u1M}yDqAygFR9Q=$9F)d$YBhP#;%obg%!kR#R?9jAIw<x4%($dGMm1`57RFV1
z*;;lyhU67=vU}n1`}6HL)6vZ{s_3D`aq!M(S-%?~=W1Lla_O_{;C>6R8*fo5>;4`o
z!}UW>L{2$#Nal68m2CWSdHPAdOx17ag`NY4f=>GL4Jz4HrCvy(GJPZR924F32gNOr
zfaj<USr>s60rvPYlUSIZ$81fMnuv@NN-14*sYq7<A?VJ>_4a`hZC2mi_BSaM{=?c&
zxYeMhftWts!>lStgkMJvjimZVdT%MotC7EN5YX*;M(MTU8wBRo5^B(Db=A_GceS7H
z+{IEllqbdsJe_UaqDo<C5BTPw0^vE2=e(Y9`g}F{zU144CWA_r?m>4lLyp^{VQ|s2
zJf=J_`jTHze;S1={QfL3*UNs4X9zI7)Vx$PvO)&H9ec=fKez1VQ-Pam@1}3Ej!*9g
zmIEZkWYm3|-7XY&F+TgoXt8)5XSJ$roOfd^HneBee+~0C?CD?1w`wnE6vYA>Xj1yE
zS~(D>_J}vS?aSoYvAaXH0`YD^UJEzWP<eq2<W0$dXt|!F5MZw0_`RM^?}K_v>nEH4
zq5l{l*+~TP&nnkg4@fc%2q2UE`S!8pJfmIysHrP=xEo=27~!?@frO%EkGWXpcDgtX
zD#7`wdJKb@Vb$`y$CKPS`rQ#+V06Ls8p&qrquVJfld@A0@t3R5W^J{D+yD5rhJ3hy
z2HJPV`ru<O<7koRwV9<NJGv7+YYjzmh<SB?&cfKHBahUU<r-MUM|Lf^6-F&QJw=aS
z(isZ{aa$rGO`qgr570h40>K;h0Elltz;D6=yq=5y1%BU^z{~jVe=G3&PhhIG47Xqy
zCh*~mo(El=I3@)M_3qRiI+ikHt#_DQr(%+=F;EebR}sMvqkiN!xz^qH_1n&`X}5)-
zQt!SN>cwtWgPDt#p+M$6Ti@>?iHzZoPi6DFh={|)I6H95dW&I_#L@Cn+v$u7N8iCG
zg5rSXE+A1_iRB6EF!#;qN~e;rF8{iwfFqskMzJ?qpKq}3My#apLE)kx%(HR`(_u_C
zQru=B2`PLEhW&cx#JlP<Yv1w|HeO;j@VQhYk-sX_`9<YJeGx78k6Twv8`rH-tG{Cm
z@#%I(3wPGLJ{<Yc#uver(LW4VVP&eh{|@HyBmXZ-^Z(^M-->DS1!(r$b)pyIYt>CM
z#xNh0dO}|fVnB+cUw)x7co+eapTZs6<Y%$2iIsx%lQ~I%BXGTxoG&voft0RU%Jq@o
zFtYjMnJ|dUU&nU5^YwAul>^alOH0QG+E-UA-Rc#_U;Y;0`X5b5&%%c-ruQ<=S9K_b
z#|OO3Q3{XC*(PBi8=8$uav%BGAJlG1*Y|Ccoi~^tq^LHrSCV%=Dhcdj+|@2-_`V%n
zY4K;!(&5ln)lh$F-t<?kVW;hCMAv4r9nRc#XIu^_@+q4u_}6hrRbyQm16XzrH!xlr
z?L1*I14XR-k-ZmU2<mB;O+qFsA5?IEaVCibD|zhmzV<fV;S<C=%#FF&*wUhRz8x*c
zAaC|>?<m~ZL4{ZKx?QmXvrL7pifoB=X!L|dciD>idx7sD(Tw=k5sM>6Fpn^HEwDQ_
z-mcL0Z;fTM`-4qZppE0OO`3iOk!-X^Xf%oCthHlnYhw7&MF+vtLZH}qYin(ZkRm+1
zd2(Tq2@9#O4PCSk85as}w%eyD2TGJeE%>eJ?*)y=Gf?%()^w$l+hs{uBk-;u4t5tn
z%V$vG+e%zS9G(VhOb5U@l2>Fcoc@<KmHw5rJdC3IXqA`8=B?b~Xd?>C*1nE@CU3ln
z_|6NBT80gWs2<s9hmC1=a~`W3KK~Je#|l34LMi)WbHd)rT6cc@=(j<W8|rJN!3Ts|
z)nTZW+^Ef@ACcekqq@1AEV(iiq+PaFGumo>YDdiVW#9hg5btFD=et4i_I(K&al956
z%g1ay-yE+RcUZ)&qAC3OYR;a^h-gO2%wS+pxp8mLZnIMPK>Xkk@4dzR-+pKUx~x~j
zfv&ABc(sIPxdyK0Sldbf$G?!4jfgFh$3*46SsjdQ9$s?&4|fLSS0CRm%dN~=_*Iz`
zlfyYPP)dr_pYe0ez;bG#M}Dl0G=io4r`lZz<AM$$oAXU~YxHUDwVJfe?{1TOiw80r
z-!2Nr*D6HPyK!uU$lnW!r)yMsM$zx8L!NfOO@B7u;$K<g&N)`Ay|OyjI37?kqf;h%
z9xZwdykQuLxzaODB<H?;bbARcFif161sE4SQR5pc_wg2>J_~_T_hUF`B1!Vsmo;Z%
zqugu4zh=gj;WNvs)$aWWUojUiJg!lRs9CUd53~?2#M$(R7ciW;a-Vy=W_y>LP<Y`J
zKf>A0rd&PinP&_3GT$#38*_%A$Qslt(!0G7TbKHIXXhFTf6wVBJd*?}0m`6uEl4$z
z|FKFCAj9g1=caV#-#&_QBmT!Ubp*C}>M6n*%!kqAhk(TG?*j5GT!rCdLn|M()82D3
pw-imZ<8FW#&;RSoyUm4l6YhNcWBsVnF|ZDD=A^A9$=u_&{{`hD7wrH5

literal 0
HcmV?d00001