Styjun

How to run a command 1 out of N times in Bash

Moscow SVO airport, how to avoid scam taxis without pre-booking?

Where should I draw the line on follow up questions from previous employer

Can you use Apple Care+ without any checks (bringing just MacBook)?

Why haven't the British protested Brexit as ardently as the Hong Kong protesters?

What is the following VRP?

An idiom for “Until you punish the offender, they will not give up offenses”

How to investigate an unknown 1.5GB file named "sudo" in my Linux home directory?

How would a disabled person earn their living in a medieval-type town?

How to get frequency counts using column breaks by row?

Could a simple hospital oxygen mask protect from aerosol poison?

Turn off Google Chrome's Notification for "Flash Player will no longer be supported after December 2020."

How to load files as a quickfix window at start-up

What is the chance of getting a Red Cabbage in year 1?

How smart contract transactions work?

German equivalent to "going down the rabbit hole"

How many possible file types in the output `ls -l` command?

Don't look at what I did there

Can I leave a large suitcase at TPE during a 4-hour layover, and pick it up 4.5 days later when I come back to TPE on my way to Taipei downtown?

What is the practical impact of using System.Random which is not cryptographically random?

Calculate Landau's function

Should a TA point out a professor's mistake while attending their lecture?

Break down the phrase "shitsurei shinakereba naranaindesu"

Why does the U.S. military maintain their own weather satellites?

Taking the dotproduct a 2D array

Should I put #! (shebang) in Python scripts, and what form should it take?How do I get indices of N maximum values in a NumPy array?average numpy array but retain shapeValue error in optimize.curve_fit not solved by inputing numpy arraysTaking dot products of high dimensional numpy arraysFail to convert List of ndarrays to numpy arraypassing relu function to all element of a numpy arrayPython Numpy Logistic RegressionHow to fix NumPy dotting errors due to 0 dimensional matrices?Numpy matrix multiplication with 2D elements

.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty margin-bottom:0;

2

I want to take the dotproduct for a sigmoid function.

import numpy as np


def sigmoid(x):
 sigm = 1 / (1 + np.exp(-x))
 return sigm


def p(D, w, b):
 prob=sigmoid(np.dot(D[:,7],w)+b)
 return prob

Works

 x = np.array([D[0,7], D[0,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x[:],w)+b)

also works, however

 x = np.array([D[:,7], D[:,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x,w)+b)

does NOT work.
It throws a dimensional alignment error.

ValueError: shapes (2,426) and (2,) not aligned: 426 (dim 1) != 2 (dim 0)

I guess I could use a loop to iterate the first array position, but maybe Python3 has a way to do this more elegantly.
How can I format the array correctly so it gets accepted as input for the numpy.dot() function ?

python python-3.x numpy numpy-ndarray dot-product

share|improve this question

edited Mar 28 at 0:05

Valentin Metz

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges

• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  What are the shapes of x,w,and b? What are those of D, w1, and w2?
  
  – norio
  Mar 27 at 23:22
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  [:] does nothing.
  
  – hpaulj
  Mar 28 at 0:03
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  The error indicates that x has (2, 426) shape, and w (2,). x.T would pair those correctly, so would np.dot(w, x), resulting in a (426,) array.
  
  – hpaulj
  Mar 28 at 0:15
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  In the second example, x.shape is (2,). It is not any more a matrix but a vector. In the first example, you need to review the shape of D that you passed to the function p. Maybe it happened to be (2,something).
  
  – norio
  Mar 28 at 0:49
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  If x and y are both vectors of real-valued elements, then x dot y = y dot x. However, in your case, one of them is a matrix and the other is a vector. You can still use np.dot to do such a matrix-vector multiplication, but now you need to be careful about the order -- one order is valid, but the operation is not defined for the other order -- unless the matrix is square. Please take a look at matrix-matrix multiplication, and reduce it to matrix-vector multiplication, and then to vector-vector multiplication. The last one is the dot product you looked up in wikipedia and wolfram.
  
  – norio
  Mar 28 at 1:03
  
  

  
  
  
  

 | 
show 12 more comments

2

I want to take the dotproduct for a sigmoid function.

import numpy as np


def sigmoid(x):
 sigm = 1 / (1 + np.exp(-x))
 return sigm


def p(D, w, b):
 prob=sigmoid(np.dot(D[:,7],w)+b)
 return prob

Works

 x = np.array([D[0,7], D[0,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x[:],w)+b)

also works, however

 x = np.array([D[:,7], D[:,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x,w)+b)

does NOT work.
It throws a dimensional alignment error.

ValueError: shapes (2,426) and (2,) not aligned: 426 (dim 1) != 2 (dim 0)

I guess I could use a loop to iterate the first array position, but maybe Python3 has a way to do this more elegantly.
How can I format the array correctly so it gets accepted as input for the numpy.dot() function ?

python python-3.x numpy numpy-ndarray dot-product

share|improve this question

edited Mar 28 at 0:05

Valentin Metz

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges

• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  What are the shapes of x,w,and b? What are those of D, w1, and w2?
  
  – norio
  Mar 27 at 23:22
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  [:] does nothing.
  
  – hpaulj
  Mar 28 at 0:03
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  The error indicates that x has (2, 426) shape, and w (2,). x.T would pair those correctly, so would np.dot(w, x), resulting in a (426,) array.
  
  – hpaulj
  Mar 28 at 0:15
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  In the second example, x.shape is (2,). It is not any more a matrix but a vector. In the first example, you need to review the shape of D that you passed to the function p. Maybe it happened to be (2,something).
  
  – norio
  Mar 28 at 0:49
  

  
  
  
  


• 
  
  
  

  
  1
  

  
  
  
  
  
  

  
  
  If x and y are both vectors of real-valued elements, then x dot y = y dot x. However, in your case, one of them is a matrix and the other is a vector. You can still use np.dot to do such a matrix-vector multiplication, but now you need to be careful about the order -- one order is valid, but the operation is not defined for the other order -- unless the matrix is square. Please take a look at matrix-matrix multiplication, and reduce it to matrix-vector multiplication, and then to vector-vector multiplication. The last one is the dot product you looked up in wikipedia and wolfram.
  
  – norio
  Mar 28 at 1:03
  
  

  
  
  
  

 | 
show 12 more comments

I want to take the dotproduct for a sigmoid function.

import numpy as np


def sigmoid(x):
 sigm = 1 / (1 + np.exp(-x))
 return sigm


def p(D, w, b):
 prob=sigmoid(np.dot(D[:,7],w)+b)
 return prob

Works

 x = np.array([D[0,7], D[0,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x[:],w)+b)

also works, however

 x = np.array([D[:,7], D[:,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x,w)+b)

does NOT work.
It throws a dimensional alignment error.

ValueError: shapes (2,426) and (2,) not aligned: 426 (dim 1) != 2 (dim 0)

I guess I could use a loop to iterate the first array position, but maybe Python3 has a way to do this more elegantly.
How can I format the array correctly so it gets accepted as input for the numpy.dot() function ?

python python-3.x numpy numpy-ndarray dot-product

share|improve this question

edited Mar 28 at 0:05

Valentin Metz

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges

import numpy as np


def sigmoid(x):
 sigm = 1 / (1 + np.exp(-x))
 return sigm


def p(D, w, b):
 prob=sigmoid(np.dot(D[:,7],w)+b)
 return prob

 x = np.array([D[0,7], D[0,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x[:],w)+b)

 x = np.array([D[:,7], D[:,9]])
 w = np.array([w1, w2])
 prob=sigmoid(np.dot(x,w)+b)

ValueError: shapes (2,426) and (2,) not aligned: 426 (dim 1) != 2 (dim 0)

share|improve this question

edited Mar 28 at 0:05

Valentin Metz

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges

share|improve this question

edited Mar 28 at 0:05

Valentin Metz

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges

asked Mar 27 at 23:16

Valentin MetzValentin Metz

3266 bronze badges