# Using ADVI's question

**URL:** <https://discourse.pymc.io/t/using-advis-question/2310>\
**Category:** Questions\
**Created:** [December 3, 2018, 10:12am UTC](https://discourse.pymc.io/t/using-advis-question/2310 "2018-12-03T10:12:57Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![Abraxas2071](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/abraxas2071/32/1287_2.png) [@Abraxas2071](https://discourse.pymc.io/u/Abraxas2071)\
**Post date:** [December 3, 2018, 10:12am UTC](https://discourse.pymc.io/t/using-advis-question/2310/1 "2018-12-03T10:12:57Z")

</div>

![IMG_20181203_180117](https://canada1.discourse-cdn.com/flex036/uploads/pymc3/original/2X/6/675504cfc19ca917f6bcea2d5f1de4bee595a166.jpeg)

ferrine,Our Dear Maxim told me look this link “[https://docs.pymc.io/notebooks/variational\_api\_quickstart.html](https://docs.pymc.io/notebooks/variational_api_quickstart.html)”。 \* But I can’t understand how to use ADVI method in example of the book  
Book’s title《Bayesian Analysis with Python》

Help me~~o(╥﹏╥)oo(╥﹏╥)oo(╥﹏╥)o

---

<div class="post-metadata">

**Author:** ![junpenglao](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/junpenglao/32/8_2.png) [@junpenglao](https://discourse.pymc.io/u/junpenglao)\
**Post date:** [December 3, 2018, 10:39am UTC](https://discourse.pymc.io/t/using-advis-question/2310/2 "2018-12-03T10:39:59Z")

</div>

I am glad that you are reading @aloctavodia’s book (I helped edited the Chinese translation).  
I think the latest code is in [https://github.com/aloctavodia/BAP/blob/master/first\_edition/code/Chp4/04\_Understanding\_and\_predicting\_data\_with\_linear\_regression\_models.ipynb](https://github.com/aloctavodia/BAP/blob/master/first_edition/code/Chp4/04_Understanding_and_predicting_data_with_linear_regression_models.ipynb)  
But seems ADVI parts are removed, you should have a look at [https://docs.pymc.io/notebooks/variational\_api\_quickstart.html](https://docs.pymc.io/notebooks/variational_api_quickstart.html) instead

---

<div class="post-metadata">

**Author:** ![Abraxas2071](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/abraxas2071/32/1287_2.png) [@Abraxas2071](https://discourse.pymc.io/u/Abraxas2071)\
**Post date:** [December 3, 2018, 11:41am UTC](https://discourse.pymc.io/t/using-advis-question/2310/3 "2018-12-03T11:41:55Z")

</div>

你就是知乎上的那个！原来你在这里！😃😃😃

---

<div class="post-metadata">

**Author:** ![aloctavodia](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/aloctavodia/32/8642_2.png) [@aloctavodia](https://discourse.pymc.io/u/aloctavodia)\
**Post date:** [December 3, 2018, 12:35pm UTC](https://discourse.pymc.io/t/using-advis-question/2310/4 "2018-12-03T12:35:33Z")

</div>

Just to clarify, in that example ADVI was used to initialize NUTS, but now we recommend just using `pm.sample()` and let PyMC3 do the initialization for us. Although I should fix all that divergences!

---

<div class="post-metadata">

**Author:** ![Abraxas2071](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/abraxas2071/32/1287_2.png) [@Abraxas2071](https://discourse.pymc.io/u/Abraxas2071)\
**Post date:** [December 3, 2018, 1:15pm UTC](https://discourse.pymc.io/t/using-advis-question/2310/5 "2018-12-03T13:15:34Z")

</div>

- Look who this is!Our Dear Osvaldo!( • ̀ω•́ )✧  
You mean now don’t use NUTS()、Metropolis()、find\_MAP()，just use pm.sample???

---

<div class="post-metadata">

**Author:** ![junpenglao](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/junpenglao/32/8_2.png) [@junpenglao](https://discourse.pymc.io/u/junpenglao)\
**Post date:** [December 3, 2018, 1:16pm UTC](https://discourse.pymc.io/t/using-advis-question/2310/6 "2018-12-03T13:16:57Z")

</div>

Yes, always go for `pm.sample(...)` first 😉