Hi, I'm Gabi.

I'm a medical student based in Barcelona and a self-made data scientist. I love Medicine, coding and data mining.

Right now I work in the crosspoint of these amazing fields exploring the deep secrets of the brain thanks to MRI Technology.

Public PGP

3D plotting made easy

Lately, I’ve been working a lot on data with multiple dimensions. I’m used to work a lot with Bokeh but one of it’s limitations is 3D. Matplotlib always offers a way of plotting almost anything I can imagine, but this time I didn’t want to sacrifice the interactivity that I’m used to have in Bokeh.

The last couple of months I’ve been doing a nice Python postgrade in the UB by Ramon Crehuet and Fermin Huarte and one of the most exciting things we have been working with is a service called plotly.

I’m not a huge fan of using a proprietary service for doing research, but in this case it looked like a great way to go. It’s easy to install and easy to integrate inside the Jupyter Notebooks.

There is just one observation I would like to do, WHY is this the easiest plotting library ever ? It’s super easy and very very intuitive to use..

Set the credentials:

from plotly.tools import set_credentials_file
set_credentials_file(username='xx', api_key='xx', stream_ids=['xx', 'xx'])

Then look for the plot type you want to create, in this case a 3D Scatter made from a 3D random walker with 4 walkers:

def adimensional_random_walkers(walkers, steps, dimensions):
    workers = np.zeros(walkers*dimensions)
    for i in range(steps):
        random_step = np.random.randint(-1, 2, size=walkers*dimensions)
        workers = np.c_[workers, random_step]
    return np.cumsum(workers, axis=1).reshape(walkers,dimensions,steps+1) # +1 for starting point

four_walkers = adimensional_random_walkers(4, 100, 3)
from plotly.plotly import iplot
from plotly.graph_objs import Scatter3d, Data, Marker

walkers = []

for walker in four_walkers:
    trace = Scatter3d(
        x=walker[0],
        y=walker[1],
        z=walker[2],
    )
    walkers.append(trace)

data = Data(walkers)

iplot(data, filename = 'four-3d-random-walkers')
trace0_y, trace1_y, trace2_y, trace3_y

Very useful to if doing a PCA and you want to check the result in a moment:

from plotly.plotly import iplot
from plotly.graph_objs import Scatter3d, Data, Marker

# First three dimensions from reduced X VS the Y
trace0 = Scatter3d(
    x=Reduced_with_PCA[:, 0],
    y=Reduced_with_PCA[:, 1],
    z=Reduced_with_PCA[:, 2],
    marker=Marker(color=Y, colorscale='Portland'),
    mode='markers'
)
data = Data([trace0])

iplot(data, filename = 'pca-cloud')