arviz-devs / preliz Goto Github PK

View Code? Open in Web Editor NEW

78.0 12.0 9.0 40.9 MB

A tool-box for prior elicitation.

Home Page: https://preliz.readthedocs.io

License: Apache License 2.0

Python 51.87% Jupyter Notebook 47.42% TeX 0.70%

bayesian-data-analysis bayesian-statistics prior-distribution prior-elicitation probability-distribution statistics

preliz's People

Contributors

Stargazers

Watchers

Forkers

oriolabril aleicazatti aloctavodia adhaka olexandr-konovalov rohanbabbar04 nishant42491 madhav-kanda panagiotisanastasakis

preliz's Issues

Add more plotting arguments to maxent or add plot_kwargs

Currently the only available option is to plot or not plot.

Add function to create sliders for predictive elicitation

We should explore two approaches. Create sliders from

A PyMC model (and equivalent object for other PPLs)
Combination of PreliZ distributions, something like a PreliZ model

Automatically render the output of ipywidgets

Currently, we need to

Click on "delete widget state".
Restart the notebook
Click on "save widget state".

Here there are other alternatives that we could implement https://ipywidgets.readthedocs.io/en/8.0.2/embedding.html

Using nbsphinx is not a solution as stated in #177 by @OriolAbril

Tkinter not working with Thebe/Binder

Maybe we can make it work by doing

from xvfbwrapper import Xvfb

with Xvfb() as xvfb:
    pz.roulette()

import pyvirtualdisplay

_display = pyvirtualdisplay.Display(visible=False,  # use False with Xvfb
                                    size=(1400, 900))
_ = _display.start()

Both approaches require to have xvfb and maybe also x11-utils, which are not installed by default. I try to install them by adding a apt.txt file to the root of the repo. But it seems binder is not installing it.

evalute if a betterr optimization routine is needed for quartile

It is often the case that the quartile function returns a distribution with a mass slightly different from 0.5 (inside the interquartile range). This could be partially caused by a fundamental limitation, but also by a suboptimal optimization routine.

DOCS: Add nbsphinx

https://ipywidgets.readthedocs.io/en/7.6.5/embedding.html#using-the-nbsphinx-project

Explore if preliz can benefit from probabilistic numerics routines

Maybe some of our functions could benefit from a Probabilistic Numerics perspective. https://www.probabilistic-numerics.org/

Functions like maxent or roulette could either return a single distribution (current behavior) or a few distributions representing the optimization uncertainty and input uncertainty. For example, for the roulette method, the discreteness of the bins/chips.

The main goal is to avoid users to be overconfident of their decisions.

PPA: Allow representations other that KDE

Other options could be empirical cdf, quantile dot plots, histograms. Maybe point-interval only?

Improve ipywidgets layout for ppa

Add interactivity to distributions

It would be nice to be able to explore distributions interactively. Ipywidgets already provides functions like interactive to do this. Thus, it may be not that difficult to wrap that into an interactive method for the PreliZ distributions.

Improve ipywidgets layout for roulette

Improve aesthetics
Separate the list of distributions into discrete and continuous. Or create a panel of checkboxes with general control like "all discrete", "all continuous", "deselect all"... This panel could be placed to the right of the matplotlib subplot to take advantage of the blank space (and wide screens).

PPA: Expand to more models/families

The logic for back_fitting the prior draws to the prior was done quickly and needs improvements. Additionally, we need more test.

Allow PPA to return priors from other families than the ones in the original model.

add test for quartile function

Create public roadmap

Interactive examples are not able to show the widgets

This is the error when trying to run pz.roulette() from https://preliz.readthedocs.io/en/latest/examples/param_space_1d_examples.html

[Open Browser Console for more detailed log - Double click to close this message]
Failed to load model class 'MPLCanvasModel' from module 'jupyter-matplotlib'
makeError@https://hub.gke2.mybinder.org/user/arviz-devs-arviz_sandbox-pyp8w7j3/static/nbclassic/components/requirejs/require.js?v=d37b48bb2137faa0ab98157e240c084dd5b1b5e74911723aa1d1f04c928c2a03dedf922d049e4815f7e5a369faa2e6b6a1000aae958b7953b5cc60411154f593:168:17
checkLoaded@https://hub.gke2.mybinder.org/user/arviz-devs-arviz_sandbox-pyp8w7j3/static/nbclassic/components/requirejs/require.js?v=d37b48bb2137faa0ab98157e240c084dd5b1b5e74911723aa1d1f04c928c2a03dedf922d049e4815f7e5a369faa2e6b6a1000aae958b7953b5cc60411154f593:696:23
newContext/checkLoaded/checkLoadedTimeoutId<@https://hub.gke2.mybinder.org/user/arviz-devs-arviz_sandbox-pyp8w7j3/static/nbclassic/components/requirejs/require.js?v=d37b48bb2137faa0ab98157e240c084dd5b1b5e74911723aa1d1f04c928c2a03dedf922d049e4815f7e5a369faa2e6b6a1000aae958b7953b5cc60411154f593:717:25

add mle function for discrete distributions

scipy does not provide a .fit method for discrete distributions. We should add a function for internal use.

We will implement method on a distribution basis

Binomial
NegativeBinomial
Poisson

ppa not showing in docs

https://preliz.readthedocs.io/en/latest/api_reference.html#module-preliz

@OriolAbril do you know what am I missing?

Add method to elicit a Dirichlet prior

There are a few methods as described for example here.

This method or this one may be good candidates

inherit docstrings

For example the plot() method

Quantile distributions

This paper https://osf.io/paby6/ and this library https://github.com/dmi3kno/qpd are very interesting.

predictive sliders: allow users to change min and max values

While the function tries to provide some useful starting values, It will be useful to allow users to define the values themself
This could be done by passing a dictionary of variable name tuple min/max and/or by linking text boxes to the sliders

Explore ipywidgets and other jupyter/html5 friendly GUIs

We may want to keep Tkinter interface and offer more than one option.

Binder Thebe not working with python >= 3.8

Waiting for build to start...
Picked Git content provider.
Cloning into '/tmp/repo2dockerrm4_86m6'...
HEAD is now at 6e39d29 Merge pull request #1 from arviz-devs/display
Building conda environment for python=3.7Using PythonBuildPack builder
Building conda environment for python=3.7Building conda environment for python=3.7Step 1/54 : FROM buildpack-deps:bionic
 ---> 38c3a7e0952a
Step 2/54 : ENV DEBIAN_FRONTEND=noninteractive
 ---> Using cache
 ---> 107534e99c42
Step 3/54 : RUN apt-get -qq update &&     apt-get -qq install --yes --no-install-recommends locales > /dev/null &&     apt-get -qq purge &&     apt-get -qq clean &&     rm -rf /var/lib/apt/lists/*
 ---> Using cache
 ---> aa90368103ab
Step 4/54 : RUN echo "en_US.UTF-8 UTF-8" > /etc/locale.gen &&     locale-gen
 ---> Using cache
 ---> 891957450365
Step 5/54 : ENV LC_ALL en_US.UTF-8
 ---> Using cache
 ---> fc51f9229e37
Step 6/54 : ENV LANG en_US.UTF-8
 ---> Using cache
 ---> aab53d369b48
Step 7/54 : ENV LANGUAGE en_US.UTF-8
 ---> Using cache
 ---> 8ae6cbeac1b0
Step 8/54 : ENV SHELL /bin/bash
 ---> Using cache
 ---> 30e8f62ce4d0
Step 9/54 : ARG NB_USER
 ---> Using cache
 ---> a09cc57ebe4e
Step 10/54 : ARG NB_UID
 ---> Using cache
 ---> 81921c230f5a
Step 11/54 : ENV USER ${NB_USER}
 ---> Using cache
 ---> bbddf314ed72
Step 12/54 : ENV HOME /home/${NB_USER}
 ---> Using cache
 ---> 835d094c6f39
Step 13/54 : RUN groupadd         --gid ${NB_UID}         ${NB_USER} &&     useradd         --comment "Default user"         --create-home         --gid ${NB_UID}         --no-log-init         --shell /bin/bash         --uid ${NB_UID}         ${NB_USER}
 ---> Using cache
 ---> 0653cf63f3e3
Step 14/54 : RUN apt-get -qq update &&     apt-get -qq install --yes --no-install-recommends        less        unzip        > /dev/null &&     apt-get -qq purge &&     apt-get -qq clean &&     rm -rf /var/lib/apt/lists/*
 ---> Using cache
 ---> 7808c66cd89b
Step 15/54 : EXPOSE 8888
 ---> Using cache
 ---> 1b4a643f0375
Step 16/54 : ENV APP_BASE /srv
 ---> Using cache
 ---> db10df214b44
Step 17/54 : ENV CONDA_DIR ${APP_BASE}/conda
 ---> Using cache
 ---> 95ed60f307ac
Step 18/54 : ENV NB_PYTHON_PREFIX ${CONDA_DIR}/envs/notebook
 ---> Using cache
 ---> 554833aa351d
Step 19/54 : ENV NPM_DIR ${APP_BASE}/npm
 ---> Using cache
 ---> 2caa15484789
Step 20/54 : ENV NPM_CONFIG_GLOBALCONFIG ${NPM_DIR}/npmrc
 ---> Using cache
 ---> 3c266a026bb4
Step 21/54 : ENV NB_ENVIRONMENT_FILE /tmp/env/environment.lock
 ---> Using cache
 ---> 711cc3b384ce
Step 22/54 : ENV MAMBA_ROOT_PREFIX ${CONDA_DIR}
 ---> Using cache
 ---> e2fd2586d4d5
Step 23/54 : ENV MAMBA_EXE ${CONDA_DIR}/bin/mamba
 ---> Using cache
 ---> a8b1f732adb1
Step 24/54 : ENV KERNEL_PYTHON_PREFIX ${NB_PYTHON_PREFIX}
 ---> Using cache
 ---> 41b3c469b791
Step 25/54 : ENV PATH ${NB_PYTHON_PREFIX}/bin:${CONDA_DIR}/bin:${NPM_DIR}/bin:${PATH}
 ---> Using cache
 ---> d60104446ce4
Step 26/54 : COPY --chown=1000:1000 build_script_files/-2fusr-2flib-2fpython3-2e9-2fsite-2dpackages-2frepo2docker-2fbuildpacks-2fconda-2factivate-2dconda-2esh-e9bee0 /etc/profile.d/activate-conda.sh
 ---> Using cache
 ---> fda4fa6e91e9
Step 27/54 : COPY --chown=1000:1000 build_script_files/-2fusr-2flib-2fpython3-2e9-2fsite-2dpackages-2frepo2docker-2fbuildpacks-2fconda-2fenvironment-2epy-2d3-2e7-2elock-d12193 /tmp/env/environment.lock
 ---> Using cache
 ---> a9654070d37a
Step 28/54 : COPY --chown=1000:1000 build_script_files/-2fusr-2flib-2fpython3-2e9-2fsite-2dpackages-2frepo2docker-2fbuildpacks-2fconda-2finstall-2dbase-2denv-2ebash-41d468 /tmp/install-base-env.bash
 ---> Using cache
 ---> 65bd65962e4b
Step 29/54 : RUN TIMEFORMAT='time: %3R' bash -c 'time /tmp/install-base-env.bash' && rm -rf /tmp/install-base-env.bash /tmp/env
 ---> Using cache
 ---> 8304472795aa
Step 30/54 : RUN mkdir -p ${NPM_DIR} && chown -R ${NB_USER}:${NB_USER} ${NPM_DIR}
 ---> Using cache
 ---> 7a1b4348b810
Step 31/54 : USER root
 ---> Using cache
 ---> 5d60aea9f45a
Step 32/54 : ARG REPO_DIR=${HOME}
 ---> Using cache
 ---> cc31b856c17d
Step 33/54 : ENV REPO_DIR ${REPO_DIR}
 ---> Using cache
 ---> 6b26acd6d870
Step 34/54 : WORKDIR ${REPO_DIR}
 ---> Using cache
 ---> eae88cc05961
Step 35/54 : RUN chown ${NB_USER}:${NB_USER} ${REPO_DIR}
 ---> Using cache
 ---> b44497fad57a
Step 36/54 : ENV PATH ${HOME}/.local/bin:${REPO_DIR}/.local/bin:${PATH}
 ---> Using cache
 ---> 7da3e318c390
Step 37/54 : ENV CONDA_DEFAULT_ENV ${KERNEL_PYTHON_PREFIX}
 ---> Using cache
 ---> 0b893f8c0680
Step 38/54 : COPY --chown=1000:1000 src/requirements.txt ${REPO_DIR}/requirements.txt
 ---> Using cache
 ---> 3c4b992dd281
Step 39/54 : RUN apt-get -qq update && apt-get install --yes --no-install-recommends x11-utils xvfb && apt-get -qq purge && apt-get -qq clean && rm -rf /var/lib/apt/lists/*
 ---> Using cache
 ---> 20654a22bef8
Step 40/54 : USER ${NB_USER}
 ---> Using cache
 ---> 6fa8d1d05ac8
Step 41/54 : RUN ${KERNEL_PYTHON_PREFIX}/bin/pip install --no-cache-dir -r "requirements.txt"
 ---> Running in b6fa2354922f
Collecting preliz@ git+https://github.com/arviz-devs/preliz.git
  Cloning https://github.com/arviz-devs/preliz.git to /tmp/pip-install-k3vraino/preliz_42788d36f102427896796c25f163cb13
  Running command git clone --filter=blob:none --quiet https://github.com/arviz-devs/preliz.git /tmp/pip-install-k3vraino/preliz_42788d36f102427896796c25f163cb13
  Resolved https://github.com/arviz-devs/preliz.git to commit bbde3f8f6bca53492a881264e72398eadc991691
  Installing build dependencies: started
  Installing build dependencies: finished with status 'done'
  Getting requirements to build wheel: started
  Getting requirements to build wheel: finished with status 'done'
  Preparing metadata (pyproject.toml): started
  Preparing metadata (pyproject.toml): finished with status 'done'
Collecting nbgitpuller
  Downloading nbgitpuller-1.1.0-py2.py3-none-any.whl (456 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 456.9/456.9 KB 21.7 MB/s eta 0:00:00
Collecting arviz
  Downloading arviz-0.12.1-py3-none-any.whl (1.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.6/1.6 MB 11.3 MB/s eta 0:00:00
Collecting xarray-einstats[einops]
  Downloading xarray_einstats-0.2.2-py3-none-any.whl (33 kB)
Collecting pyvirtualdisplay
  Downloading PyVirtualDisplay-3.0-py3-none-any.whl (15 kB)
Collecting xvfbwrapper
  Downloading xvfbwrapper-0.2.9.tar.gz (5.6 kB)
  Preparing metadata (setup.py): started
  Preparing metadata (setup.py): finished with status 'done'
Requirement already satisfied: jupyter-server>=1.10.1 in /srv/conda/envs/notebook/lib/python3.7/site-packages (from nbgitpuller->-r requirements.txt (line 1)) (1.16.0)
Requirement already satisfied: notebook>=5.5.0 in /srv/conda/envs/notebook/lib/python3.7/site-packages (from nbgitpuller->-r requirements.txt (line 1)) (6.4.10)
Requirement already satisfied: tornado in /srv/conda/envs/notebook/lib/python3.7/site-packages (from nbgitpuller->-r requirements.txt (line 1)) (6.1)
Collecting pandas>=0.23
  Downloading pandas-1.3.5-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (11.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 11.3/11.3 MB 30.4 MB/s eta 0:00:00
Collecting scipy>=0.19
  Downloading scipy-1.7.3-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (38.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 38.1/38.1 MB 30.5 MB/s eta 0:00:00
Collecting matplotlib>=3.0
  Downloading matplotlib-3.5.2-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (11.2 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 11.2/11.2 MB 30.4 MB/s eta 0:00:00
Requirement already satisfied: typing-extensions>=3.7.4.3 in /srv/conda/envs/notebook/lib/python3.7/site-packages (from arviz->-r requirements.txt (line 4)) (4.1.1)
Requirement already satisfied: setuptools>=38.4 in /srv/conda/envs/notebook/lib/python3.7/site-packages (from arviz->-r requirements.txt (line 4)) (62.0.0)
Collecting xarray>=0.16.1
  Downloading xarray-0.20.2-py3-none-any.whl (845 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 845.2/845.2 KB 48.4 MB/s eta 0:00:00
Collecting numpy>=1.12
  Downloading numpy-1.21.6-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (15.7 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 15.7/15.7 MB 30.2 MB/s eta 0:00:00
Collecting netcdf4
  Downloading netCDF4-1.6.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.1/5.1 MB 33.6 MB/s eta 0:00:00
Requirement already satisfied: packaging in /srv/conda/envs/notebook/lib/python3.7/site-packages (from arviz->-r requirements.txt (line 4)) (21.3)
Requirement already satisfied: nbclient<0.6,>=0.2 in /srv/conda/envs/notebook/lib/python3.7/site-packages (from preliz@ git+https://github.com/arviz-devs/preliz.git->-r requirements.txt (line 5)) (0.5.13)
INFO: pip is looking at multiple versions of <Python from Requires-Python> to determine which version is compatible with other requirements. This could take a while.
INFO: pip is looking at multiple versions of arviz to determine which version is compatible with other requirements. This could take a while.
Collecting arviz
  Downloading arviz-0.12.0-py3-none-any.whl (1.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.6/1.6 MB 42.1 MB/s eta 0:00:00
  Downloading arviz-0.11.4-py3-none-any.whl (1.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.6/1.6 MB 39.1 MB/s eta 0:00:00
Collecting typing-extensions<4,>=3.7.4.3
  Downloading typing_extensions-3.10.0.2-py3-none-any.whl (26 kB)
Collecting arviz
  Downloading arviz-0.11.3-py3-none-any.whl (1.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.6/1.6 MB 41.9 MB/s eta 0:00:00
  Downloading arviz-0.11.2-py3-none-any.whl (1.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.6/1.6 MB 40.1 MB/s eta 0:00:00
  Downloading arviz-0.11.1-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 40.5 MB/s eta 0:00:00
  Downloading arviz-0.11.0-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 38.0 MB/s eta 0:00:00
  Downloading arviz-0.10.0-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 40.3 MB/s eta 0:00:00
INFO: pip is looking at multiple versions of arviz to determine which version is compatible with other requirements. This could take a while.
  Downloading arviz-0.9.0-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 37.9 MB/s eta 0:00:00
  Downloading arviz-0.8.3-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 39.5 MB/s eta 0:00:00
  Downloading arviz-0.8.2-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 39.8 MB/s eta 0:00:00
  Downloading arviz-0.8.1-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 38.3 MB/s eta 0:00:00
  Downloading arviz-0.8.0-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 39.4 MB/s eta 0:00:00
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. See https://pip.pypa.io/warnings/backtracking for guidance. If you want to abort this run, press Ctrl + C.
  Downloading arviz-0.7.0-py3-none-any.whl (1.5 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.5/1.5 MB 40.2 MB/s eta 0:00:00
  Downloading arviz-0.6.1-py3-none-any.whl (1.4 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.4/1.4 MB 39.7 MB/s eta 0:00:00
  Downloading arviz-0.6.0-py3-none-any.whl (1.4 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.4/1.4 MB 41.5 MB/s eta 0:00:00
  Downloading arviz-0.5.1-py3-none-any.whl (1.4 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.4/1.4 MB 42.2 MB/s eta 0:00:00
  Downloading arviz-0.4.1-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 41.1 MB/s eta 0:00:00
  Downloading arviz-0.4.0-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 43.2 MB/s eta 0:00:00
  Downloading arviz-0.3.3-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 40.4 MB/s eta 0:00:00
  Downloading arviz-0.3.2-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 41.6 MB/s eta 0:00:00
Collecting numpy==1.15
  Downloading numpy-1.15.0-cp37-cp37m-manylinux1_x86_64.whl (13.8 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 13.8/13.8 MB 30.5 MB/s eta 0:00:00
INFO: pip is looking at multiple versions of numpy to determine which version is compatible with other requirements. This could take a while.
Collecting arviz
  Downloading arviz-0.3.1-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 41.7 MB/s eta 0:00:00
  Downloading arviz-0.3.0-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 41.3 MB/s eta 0:00:00
  Downloading arviz-0.2.1-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 41.3 MB/s eta 0:00:00
  Downloading arviz-0.2.0-py3-none-any.whl (1.3 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 42.1 MB/s eta 0:00:00
  Downloading arviz-0.1.0-py3-none-any.whl (1.0 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.0/1.0 MB 48.0 MB/s eta 0:00:00
  Downloading arviz-0.0.1-py3-none-any.whl (1.2 kB)
INFO: pip is looking at multiple versions of nbgitpuller to determine which version is compatible with other requirements. This could take a while.
Collecting nbgitpuller
  Downloading nbgitpuller-1.0.2-py2.py3-none-any.whl (430 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 430.7/430.7 KB 17.6 MB/s eta 0:00:00
ERROR: Could not find a version that satisfies the requirement scipy>=1.9.0rc1 (from preliz) (from versions: 0.8.0, 0.9.0, 0.10.0, 0.10.1, 0.11.0, 0.12.0, 0.12.1, 0.13.0, 0.13.1, 0.13.2, 0.13.3, 0.14.0, 0.14.1, 0.15.0, 0.15.1, 0.16.0, 0.16.1, 0.17.0, 0.17.1, 0.18.0, 0.18.1, 0.19.0, 0.19.1, 1.0.0, 1.0.1, 1.1.0, 1.2.0, 1.2.1, 1.2.2, 1.2.3, 1.3.0rc1, 1.3.0rc2, 1.3.0, 1.3.1, 1.3.2, 1.3.3, 1.4.0rc1, 1.4.0rc2, 1.4.0, 1.4.1, 1.5.0rc1, 1.5.0rc2, 1.5.0, 1.5.1, 1.5.2, 1.5.3, 1.5.4, 1.6.0rc1, 1.6.0rc2, 1.6.0, 1.6.1, 1.6.2, 1.6.3, 1.7.0rc1, 1.7.0rc2, 1.7.0, 1.7.1, 1.7.2, 1.7.3)
ERROR: No matching distribution found for scipy>=1.9.0rc1
Removing intermediate container b6fa2354922f
The command '/bin/sh -c ${KERNEL_PYTHON_PREFIX}/bin/pip install --no-cache-dir -r "requirements.txt"' returned a non-zero code: 1Built image, launching...
Failed to connect to event stream

Integrating binder/thebe with documentation?

I there isn't much to win from static docs or notebooks in preliz, but we can write some "ready to run" notebooks that can be executed with binder or thebe (which would be even better I think).

With thebe for example, the docs themseves have a button to run the code without even leaving the website (and it runs binder behind the scenes and interfaces with the docs website). We used that for example in the pymc sprint, however, binder doesn't offer a lot of computing power, so it's use for pymc it's limited due to slowness and lack of available RAM, but hopefully with preliz things would work fast enough to show interactiveness.

Add PolyaGamma distribution

Tell us about it

The Polya-Gamma distribution is a continuous distribution. Here there's some information about it.

Thoughts on implementation

It is implemented in PyMC but not in SciPy.

Add missing docstrings

Most of the distributions need docstrings

Add maximum entropy restriction

Distributions returned by the constraint function are not uniquely defined, one solution could be to return the maximum entropy distribution with the requested mass inside the requested interval.

Refactor plotting code

refactor ploting code in constraints and plot and make some general reusable functions

Implement quartile method

Similar to maxent but the user provides the median, lower and upper quartiles. of X, we can then use least squares to fit the distribution.

[F(Xq1, theta) - 0.25]^2 + [F(Xq2, theta) - 0.5]^2 + [F(Xq3, theta) - 0.75]^2

where F(X, theta) is the CDF of X

PPA: add option to specify reference values

Currently, the default reference value is 0. And it is fixed.

Update documentation

add quartile example.
update roulette gif
move gif below cell calling roulette()
fix typos

Test_quartile for ZeroInflatedBinomial randomly fails

It seems the result can be (ZeroInflatedBinomial(), 2, 3, 4, (0.993, 9.934, 0.394) passing or [ (ZeroInflatedBinomial(), 2, 3, 4, (0.993, 9.033, 0.394)) failing

PPA: Use "preliz functions/models"

Related to #92. We should be able to use the same syntax used for predictive_sliders. This has at least two advantages. Making PreliZ agnostic of any PPLs, simplifies iterative fitting during a PPA session. One disadvantage on not using any PPL is that users will need to write their model in their PPL and in PreliZ. One possible solution will be to provide an option to return or export the solution to a given PPL (and maybe also importing).

fixing parameters for maxent and quartile does not work for alternative parameters

Missing distributions

All new distributions should include an alternative parametrization if useful/common. We should at least cover the parametrization in PyMC, but we can consider having others too (for example the beta has kappa parameter that is useful but not present in PyMC).

entropy logitnormal

In #206 we introduced the logitnormal, the entropy is only approximated. We need to use the correct expression

Move plot method from Distribution to Discrete and Continuous

Given the differences between discrete and continuous distributions it makes sense to have plot function for each of the these instead of a single method. This should lead to a simpler docstring and list of arguments

moments logitnormal

In #206 we introduced the logitnormal, we approximate the first two moments by sampling. We may want to approximate them in a different way. We may also want to add skewness and kurtosis.

Add test for ppa function

ppa uses ipywidgets and matplotlib interactively. So ideally test should include simulated interactions

Define _from_tau and _to_tau only once

A few of the alternative parametrization involve converting from tau to sigma and vice-versa. We could define these functions only once in the library.

Fix docs

All functions from the unidimensional and predictive modules are not show in the API section. And each one of the distributions is instead of just the two main groups continuous and discrete. The changes in the last PR #164 are at least partially responsible for this error.

Add plot_interactive method to Dirichlet and MvNormal

Fix dependecy issue for building docs

Related to #60

Diagnostic and recomendantion by @OriolAbril

Pip is messing up big time. RTD first installs requirements-docs, then installs PreliZ.

On the requirements docs step:

Collecting nbclient<0.6,>=0.2

Downloading nbclient-0.5.13-py3-none-any.whl (70 kB)

 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 70.6/70.6 kB 214.0 MB/s eta 0:00:00

But then, when installing PreliZ, now nbclient is also an indirect dependency via ipywidgets, and somehow this happens:

Requirement already satisfied: nbclient>=0.5.0 in /home/docs/checkouts/readthedocs.org/user_builds/preliz/envs/60/lib/python3.9/site-packages (from nbconvert>=5->notebook>=4.4.1->widgetsnbextension~=3.6.0->ipywidgets->preliz==0.0.1.dev0) (0.5.13)

Collecting nbclient>=0.5.0

Downloading nbclient-0.6.6-py3-none-any.whl (71 kB)

 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 71.8/71.8 kB 211.0 MB/s eta 0:00:00

And at the end of the PreliZ install it even prints this:

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.

jupyter-cache 0.5.0 requires nbclient<0.6,>=0.2, but you have nbclient 0.6.6 which is incompatible.

I think adding nbclient as an explicit dependency for PreliZ will fix this. But I think the preferred solution should be making ipywidgets and scikit-learn optional dependencies, then install doing the following:

Install requirements-optional
Install requirements-docs
Install PreliZ
Extra requirements can be added in the readthedocs.yml. It would look like

install:

requirements: requirements-optional.txt
requirements: requirements-docs.txt

Add AIC/BIC criterion to mle function

PPA: Allow non-random initilization

A few options, 2 to 4 first moments, a fitted preliz distribution (from maxent, roulette or whatever), or a hand-drawn distribution. Limits, (upper or lower, or both) or x% mass between limits, etc...

add pre-commit hooks

It would be nice to run pylint and black

Add test for roulette

Roulette uses ipywidgets and matplotlib interactively. So ideally test should include simulated interactions

Add a parameter to the roulette function

To let the user specify the collection of distributions to use and set the current collection as the default value.

predictive explorer: report unknow boundaries

In #173 sliders (now textboxes) become aware of a few functions that can be applied to the parameters. Providing a truly general solution will probably be too hard, and it may require a different approach to parse the function. Something easier may be to detect 3 categories, no function has been applied, a function/operation is applied and we can solve it, and a function/operation is applied but we don't have a clue what is going on, for this later case we could return something like "(unk, unk)", signaling that the function is not able to guess the boundaries of the parameter.

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.