Plotting & Visualization Maryam Tavakol Machine Learning Group - PowerPoint PPT Presentation

Plotting & Visualization Maryam Tavakol Machine Learning Group Winter semester 2016/17 1

Motivation • Making plots and static or interactive visualizations is one of the most important tasks in data analysis • Helping identify outliers • Needed data transformations • Coming up with ideas for models • etc. 2

matplotlib • Desktop plotting package designed for creating publication-quality plots • Does have interactive features like zooming and panning • Supports many different GUI backends on all operating systems • Can export graphics to all of the common vector and raster graphics formats 3

Figures • Plots in matplotlib reside within a Figure object • Creates a new empty window 4

Subplots 5

Subplot Options 6

Plotting 7

Formatting • Line style, color, and marker 8

Formatting 9

Titles & Labels 10

Adding Legends 11

Example 12

Annotations 13

Annotations • Annotations and text can be added using the text, arrow, and annotate functions 14

Adding Shapes 15

Saving Options 16

Histogram 17

Histogram 18

Box Plot 19

Box Plot 20

Profiling & Optimization Maryam Tavakol Machine Learning Group Winter semester 2016/17 21

Introduction • Python is generally known (a bit unfairly) as a slow language, but it is possible to achieve very good performance with the right methods • Profile : To evaluate what makes a program slow • Optimize : To make the code more efficient 22

%timeit • The %timeit allow you to quickly evaluate the time taken by one or several Python statements in IPython 23

%timeit • The %timeit command accepts several optional parameters, e.g.: • Using -r or - n to set number of statement evaluations • Inside your script: timeit.timeit() 24

cProfile • Creates a detailed report of the execution time of our code, function by function • The total number of calls, total time, etc. 25

line_profiler • Analyzing of code performance by a line-by-line report • Need the package kernprof • We decorate the functions we wish to profile with @profile 
 26

Memory Usage • Memory is also a critical factor • Using memory_profiler • We decorate the functions we wish to profile with @profile 27

Optimization Tips • Choosing the right data structure • Using NumPy over native Python code • Avoiding unnecessary array copying • Using in-place operations instead of implicit copy • Parallel computing and much more… 28