CODEWITHSUNDEEP
pythonoptimization
pir
pir

Reputation: 5923

Error caused by bounds when using fmin_l_bfgs_b

I get the error 'Too many values to unpack' when optimizing using fmin_l_bfgs_b. From stackoverflow I've found that something is wrong with the way I've defined the bounds. However, I cannot seem to find the right way to do it.

I've tried to create a minimal working example below to illustrate the issue. The input is a 28x28x1 greyscale image, which is bounded from 0 to 1. As I see it I therefore want a list of 784 pairs that each have the value (0,1). I've tried to implement that using the following code:

img = random.uniform(size=(28, 28))
constraintPairs = [(0, 1)]*(28*28)

def func(img):
    return img.mean()

imgOpt, cost = fmin_l_bfgs_b(func, img, approx_grad=1,bounds=constraintPairs)

What am I doing wrong? Thanks!

Upvotes: 1

Views: 188

Answers (1)

hitzg
hitzg

Reputation: 12701

The problem is simply the return value of fmin_l_bfgs_b (Documentation). It returns 3 objects and you only define two in your code. This should work:

img = random.uniform(size=(28, 28))
constraintPairs = [(0, 1)]*(28*28)

def func(img):
    return img.mean()

img = reshape(img, (1, 28*28))
imgOpt, cost, info = fmin_l_bfgs_b(func, img, approx_grad=1,bounds=constraintPairs)

imgOpt = reshape(imgOpt, (28,28))

Whether or not 768 dimensions are too many, is hard to tell. If so, you could consider downsampling the input image.

Upvotes: 1

Related Questions