Cat Chenal CatChenal

Importing SOMPY from github in Colab

Need: SOMPY is not packaged on pipy, and conda is not installed on the colab platform (as of 9/21):

Run this in a colab cell:

!pip install git+https://github.com/sevamoo/SOMPY.git

Then import and test:

Code that I ususally include in the first cell of a jupyter nootebook:

from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = "all"

This avoids using print() to obtain multiple outputs in one cell. For example, given this code in one cell:

a = 1

Output:

Code:

import pandas as pd
import matplotlib.pyplot as plt

df1 = pd.DataFrame({'sizes':[75, 30, 45, 10]},
 index=['Frogs', 'Hogs', 'Dogs', 'Logs'])

def get_min_pca(scaled_data, min_var_explained=0.95, verbose=False):
  """
  Decompose `scaled_data` into principal components.
  Return the number of components above `min_var_explained` threshold,
  the threshold value and the transformed data.
  Args
  ----
  scaled_data: numpy array: scaled data
 min_var_explained: float, default 0.95: variance explained threshold

Output (if show == True):

Motivation:

Scikit-learn decomposition module includes many decomposition algorithms that can be used for dimensionality reduction. PCA and SVD are two algorithms that differ in their functional implementation in regards to their parameters. PCA uses its n_components parameter in different ways depending on its type:

n_components int, float or ‘mle’, default=None

Notably:

If 0 < n_components < 1 and svd_solver == 'full', select the number of components such that the amount of variance that needs to be explained is greater than the percentage specified by n_components.

I used to command line version given in this SO post to upgrade my current git version:

Check current version:

(base) C:\Users\me>git --version
git version 2.34.1.windows.1

Update if necessary, which it is for any version below 2.35 as per this offical announcement from 2022-04-12:

(base) C:\Users\me&gt;git update-git-for-windows

Example 1 - Simplest implementation: using `interact`:

=> No need for the function to have a print/display statement => This should also be the behavior of interactive & interactive_output, but it's not

import ipywidgets as ipw


@ipw.interact(x=(20.,250.,5),
          a=(-60.,60.,10),
          b=(-6_000.,6_000,100))

Data points and sample figure to replicate in E. Downey's plotting

import matplotlib.pyplot as plt

sandwich_list = [
    'Lobster roll',
    'Chicken caesar',

Tip: when populating a grid of mpl axes with unknown actual number of datasets, some axes may be empty: clear them AFTER the main loop.

Code source: DETR_panoptic notebook on colab

In the enumerated loop below, we don't know where i will end (because out["pred_masks"][keep] was not saved into a variable, hence we can't access its length).
We then use the main loop index into a second one to clear any empty axis.

#[...]

	idea = """
	GitHub Gist can store code snippets, which can be called using their gist url...
	This is going to be very handy for coding presentations especially when used in combination with
	Steamlit (https://streamlit.io/).
	Imagine: you are presenting at a major conference & to reduce your anxiety, you buy yourself
	this free insurance by following these steps:
	For each steps in your presentation, create a gist. Then, create another gist, i.e. confX_presentation.py that list each steps.
	"""

	print(idea)