CODEWITHSUNDEEP
pythonnumpyscipy
www.pieronigro.de
www.pieronigro.de

Reputation: 880

What is wrong with "numpy concatenate"?

I am trying to see an optimization progress when using scipy.optimize minimze.

I want to create a class, where i use some variables that exist outside of the actual optimization-function - x_it is one of these and after each iteration the new x vector should be concatenated with the previous one. I do this cause i would like to evaluate this iterations with matplotlib (not in the following code) and because scipy does not allow callback function for some optimization-methods:

class iter_progress:  
    x_it=[]
    def __init__(self):
        pass
    def build_iter():
        import numpy as np
        iter_progress.y_it=np.zeros((1,1), dtype=float)
        iter_progress.x_it=np.zeros((1,2), dtype=float)
    def obj(x):
        import numpy as np

        out=x[0]**2+x[1]**2

        out=np.array([[out]])
        x_copy=x.copy()[None]

        #iter_progress.x_it=np.concatenate(iter_progress.x_it.copy(), x_copy) 
        #the above line is commented because it does not work
        return out

    def mine():
        import numpy as np
        from scipy.optimize import minimize

        x0=np.array([[4,6]])

        res=minimize(iter_progress.obj,x0=x0, method='SLSQP')
        print(res.x)

in the console i do:

>>>from iter_progress import iter_progress

>>>iter_progress.build_iter()

>>>iter_progress.mine()

The Code works, but when i uncomment the line where i made a note i get: 

iter_progress.x_it=np.concatenate(iter_progress.x_it.copy(), x_copy)
TypeError: only length-1 arrays can be converted to Python scalars

Upvotes: 1

Views: 2217

Answers (2)

Andi
Andi

Reputation: 1273

To make @BiRico's comment more explicit: You forgot brackets around the arrays you want to concatenate. Like this np.concatenate((iter_progress.x_it.copy(), x_copy)).

The first argument to np.concatenate should be an iterable of arrays to be concatenated. The extra-brackets make a tuple out of the argument so the code works fine then.

Upvotes: 2

Bi Rico
Bi Rico

Reputation: 25833

The second argument to concatenate should be the axis you want to concatenate on, numpy thinks you're trying to concatenate on axis x_copy.

Take a look here to see how concatenate should be used.

Upvotes: 3

Related Questions