The `np.dot` resultant computation was automatically rounded

By the code below, in the highlighted part as ======= Weird Behaviour Part =======, I was expecting a vector containing not rounded numbers. However, as I tried three different ways in the code followed by the output on my terminal. I was getting the rounded number as the outcome of np.dot API.

Reproducing code example:

import numpy as np

def get_samples():
  """
  this prepares the dataset
  """
  M = np.array([[0.8, 0.2],[0.3,0.7]])
  B = np.array([[0.3,0.4,0.1,0.2],[0.2,0.2,0.3,0.3]])
  p = np.array([0.4,0.6])
  v = np.array([4,1,2])
  return M, B, p, v

def dummy(M,p,B,v):
  # ALPHA DYNAMIC calculate alpha
  result = []
  T = np.size(v)
  for t in range(T):
    if t == 0:
      result.append(B[:, int(v[t])-1]*p)
    else:

      """
      ======= Weird Behaviour Part =======
      """

      # manually compute dot product
      # output: 0.11800000000000002 0.14200000000000002
      print(M[:,0]@result[-1], M[:,1]@result[-1])

      # np.dot
      # output: [0.1  0.15]
      print(np.dot(M, result[-1]))

      # pythonic way of np.dot
      # output: [0.118 0.142]
      print(sum([i*j for (i, j) in zip(M, result[-1])]))

      result.append(sum([i*j for (i, j) in zip(M, result[-1])])*B[:, int(v[t])-1])
  return np.array(result).T

if __name__ == '__main__':
  M, B, p, v = get_samples()
  print(dummy(M,p,B,v))

Error message:

No ERROR

Numpy/Python version information:

Python: 3.6.6 Numpy: ‘1.14.2’

[charris]

You are computing the wrong things. print(np.dot(M, result[-1])) should be print(np.dot(M.T, result[-1])) in order to agree with the other computations. The “manual” computations are not matrix multiplication.

[charris]

Try

>>> np.set_printoptions(precision=30)