Scaling down a plot when using matplotlib - python

I've been trying to plot a graph of Epoch vs Accuracy and val_accuracy from a train log I have generated. Whenever I try to plot it, the y-axis starts from 0.93 rather than it being in 0, 0.1 ,0.2... intervals. I'm new at using matplotlib or any plot function.
Here's the code for it:
import pandas as pd
import matplotlib.pyplot as plt
acc = pd.read_csv("train_log", sep = ',')
acc.plot("epoch", ["accuracy","val_accuracy"])
plt.savefig('acc' , dpi = 300)
I'm open to suggestion in complete different ways to do this.
Picture of plot :
[1]: https://i.stack.imgur.com/lgg0W.png

This has already been discussed here. There are a couple of different ways you can do this (using plt.ylim() or making a new variable like axes and then axes.set_ylim()), but the easiest is to use the set_ylim function as it gives you heaps of other handles to manipulate the plot. You can also handle the x axis values using the set_xlim function.
You can use the set_ylim([ymin, ymax]) as follows:
import numpy as np
import matplotlib.pyplot as plt
x = np.arange(0,5)
y = np.arange(5,10)
axes = plt.gca()
axes.plot(x,y)
axes.set_ylim([0,10])
You can use the plt.ylim() like this:
import numpy as np
import matplotlib.pyplot as plt
x = np.arange(0,5)
y = np.arange(5,10)
plt.plot(x,y)
plt.ylim([0,10])
This will produce the same plot.

You need to set the lower/bottom limit using ylim().
For details please refer:
https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.ylim.html

Related

Finding certain plotting style

I am looking for a plotting function in matplotlib that plots the y-values as bars just like in an autocorrelogram but for a general function. Is there a method to do this in matplotlib or do I have to write my own function?
You could use stem
import numpy as np; np.random.seed(21)
import matplotlib.pyplot as plt
x = np.linspace(5,75)
y = np.random.randn(len(x))
plt.stem(x,y, linefmt="k", markerfmt="none", basefmt="C0", use_line_collection=True)
plt.show()

Custom scale from simple list or dict?

I need to make a custom scale for an axis. Before diving into http://matplotlib.org/examples/api/custom_scale_example.html, I'm wondering if there is an easier way for my special case.
A picture is worth a thousand words, so here we go:
See the value in each row next to the filename ? I would like the row height to be relative to the difference between it and the previous one. I'd start from 0 and would have to define a top limit so I see the last row.
Try matplotlib's pcolormesh with which you can create irregularly shaped grids.
from matplotlib import pyplot as plt
import numpy as np
y1D = np.hstack([0, np.random.random(9)])
y1D = np.sort(y1D)/np.max(y1D)
x, y = np.meshgrid(np.arange(0,1.1,0.1),y1D)
plt.pcolormesh(x,y, np.random.random((10,10)))
plt.show()
You can use this recipe and adapt to your needs:
import numpy as np
import matplotlib.pyplot as plt
grid = np.zeros((20,20))
for i in range(grid.shape[0]):
r = np.random.randint(1,19)
grid[i,:r] = np.random.randint(10,30,size=(r,))
plt.imshow(grid,origin='lower',cmap='Reds',interpolation='nearest')
plt.yticks(list(range(20)),['File '+str(i) for i in range(20)])
plt.colorbar()
plt.show()
, the result is this:

Python: Vertical 1D DotPlot

I'm trying to create a 1D-Dotplot with python, similar to this:
https://owncloud.tu-berlin.de/public.php?service=files&t=9ead31dfc988757321c7ac391920c48a
I tried using the plot.scatter method from matplotlib, but it nees data for the x-axis. I tried setting all x-values to '1', but it turns out as kind of a 2d-diagram, anyway:
https://owncloud.tu-berlin.de/public.php?service=files&t=ab9f0f521f57526e871259f3a520d94a
How can I draw a real 1d-dotplot? I found nothing in the matplotlib-docs...
I would like to use matplotlib but am also open to other suggestions.
Thanks in advance!
Cheers, Jakob
As far as I can see, also the 1D-Dotplot you're showing is two dimensional but only strongly limited in x direction.
I don't know whether there already exists something like that but the following code is doing what you ask for.
import numpy as np
import matplotlib.pyplot as mpl
# your data
data = 3. + 0.7 * np.random.randn(N)
# a small spreading of the data in x direction
x = 0.2 * np.random.randn(data.size)
# the plotting
fig,ax = mpl.subplots(1,figsize=(0.5,5))
ax.set_axis_bgcolor('#FFD7B1')
ax.scatter(x,data,alpha=0.2,c='k')
ax.plot([-1,1],[np.mean(data),np.mean(data)],'r',linewidth=2)
ax.set_xlim((-1,1))
ax.set_ylim((1,6))
ax.set_xticks([])
ax.grid(True,axis='y')
ax.set_ylabel('Note')
Your own solution is close! Just play with the aspect ratio to "squash" down the size along the x-axis:
import numpy as np
import matplotlib.pyplot as plt
x = np.random.random(100)
y = np.random.randn(100)
fh, ax = plt.subplots(1,1)
ax.scatter(x,y)
ax.set_xlim(-.5, 1.5)
ax.axes.get_xaxis().set_visible(False) # remove the x-axis and its ticks
ax.set_aspect(5, adjustable='box') # adjustable='box' is important here
plt.show()

Filling region between curve and x-axis in Python using Matplotlib

I am trying to simply fill the area under the curve of a plot in Python using MatPlotLib.
Here is my SSCCE:
import json
import pprint
import numpy as np
import matplotlib.pyplot as plt
y = [0,0,0,0,0,0,0,0,0,0,0,863,969,978,957,764,767,1009,1895,980,791]
x = np.arange(len(y))
fig2, ax2 = plt.subplots()
ax2.fill(x, y)
plt.savefig('picForWeb.png')
plt.show()
The attached picture shows the output produced.
Does anyone know why Python is not filling the entire area in between the x-axis and the curve?
I've done Google and StackOverflow searches, but could not find a similar example. Intuitively it seems that it should fill the entire area under the curve.
I usually use the fill_between function for these kinds of plots. Try something like this instead:
import numpy as np
import matplotlib.pyplot as plt
y = [0,0,0,0,0,0,0,0,0,0,0,863,969,978,957,764,767,1009,1895,980,791]
x = np.arange(len(y))
fig, (ax1) = plt.subplots(1,1);
ax1.fill_between(x, 0, y)
plt.show()
See more examples here.
If you want to use this on a pd.DataFrame use this:
df.abs().interpolate().plot.area(grid=1, linewidth=0.5)
interpolate() is optional.
plt.fill assumes that you have a closed shape to fill - interestingly if you add a final 0 to your data you get a much more sensible looking plot.
import numpy as np
import matplotlib.pyplot as plt
y = [0,0,0,0,0,0,0,0,0,0,0,863,969,978,957,764,767,1009,1895,980,791,0]
x = np.arange(len(y))
fig2, ax2 = plt.subplots()
ax2.fill(x, y)
plt.savefig('picForWeb.png')
plt.show()
Results in:
Hope this helps to explain your odd plot.

PyLab: Plotting axes to log scale, but labelling specific points on the axes

Basically, I'm doing scalability analysis, so I'm working with numbers like 2,4,8,16,32... etc and the only way graphs look rational is using a log scale.
But instead of the usual 10^1, 10^2, etc labelling, I want to have these datapoints (2,4,8...) indicated on the axes
Any ideas?
There's more than one way to do it, depending on how flexible/fancy you want to be.
The simplest way is just to do something like this:
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
x = np.exp2(np.arange(10))
plt.semilogy(x)
plt.yticks(x, x)
# Turn y-axis minor ticks off
plt.gca().yaxis.set_minor_locator(mpl.ticker.NullLocator())
plt.show()
If you want to do it in a more flexible manner, then perhaps you might use something like this:
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
x = np.exp2(np.arange(10))
fig = plt.figure()
ax = fig.add_subplot(111)
ax.semilogy(x)
ax.yaxis.get_major_locator().base(2)
ax.yaxis.get_minor_locator().base(2)
# This will place 1 minor tick halfway (in linear space) between major ticks
# (in general, use np.linspace(1, 2.0001, numticks-2))
ax.yaxis.get_minor_locator().subs([1.5])
ax.yaxis.get_major_formatter().base(2)
plt.show()
Or something like this:
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
x = np.exp2(np.arange(10))
fig = plt.figure()
ax = fig.add_subplot(111)
ax.semilogy(x)
ax.yaxis.get_major_locator().base(2)
ax.yaxis.get_minor_locator().base(2)
ax.yaxis.get_minor_locator().subs([1.5])
# This is the only difference from the last snippet, uses "regular" numbers.
ax.yaxis.set_major_formatter(mpl.ticker.ScalarFormatter())
plt.show()

Categories