Cost of Mat Mat multiply

In [1]:
import numpy as np
import numpy.linalg as la
import matplotlib.pyplot as pt
%matplotlib inline
from time import process_time

def get_solve_time(n):
    A = np.random.randn(n, n)
    B = np.random.randn(n, n)
    
    t_start = process_time()
    A @ B
    t_stop = process_time()
    
    return t_stop-t_start
In [2]:
n_values = np.array([100,1000,2000,3000,4000,5000,6000,8000,10000])
print(n_values)

[  100  1000  2000  3000  4000  5000  6000  8000 10000]
In [3]:
times = []
for n in n_values:
    newtime = get_solve_time(n)
    times.append(newtime)
In [4]:
pt.loglog(n_values, times)
pt.xlabel('n')
pt.ylabel('time')
pt.grid()
  • Can we predict individual values?
  • What does the overall behavior look like?
  • How could we determine the "underlying" function?
In [ ]: