numpy.dot() gives TypeError: can't multiply sequence by non-int of type 'float'

Question

I have just begun learning Machine Learning using Python. I have written the following class which gives an error:

TypeError: can't multiply sequence by non-int of type 'float'

class Perceptron(object):
    def __init__(self, eta=0.01, n_iter=10):
        self.eta = eta                          # Learning Rate
        self.n_iter = n_iter                    # Number of iteration over the training dataset

    def fit(self, x, y):
        self.w_ = np.zeros(1 + x.shape[1])      # Initialize Weights to zero initially                                                # x = {array-like} : shape[no_of_samples, no_of_features]
        self.errors_ = []                       # No errors in the beginning of the computation
        for _ in range(self.n_iter):
            errors = 0
            for xi, target in zip(x, y):
                update = self.eta * (target - self.predict(xi))
                self.w_[1:] += update * xi
                self.w_[0] += update
                errors += int(update != 0.0)
            self.errors_.append(errors)

        return self

    def net_input(self, x):
        return np.dot(x, self.w_[1:]) + self.w_[0]

    def predict(self, x):
        return np.where(self.net_input(x) >= 0.0, 1, -1)

I am getting an error in the net_input() method at np.dot(). I am using the following dataset : https://raw.githubusercontent.com/uiuc-cse/data-fa14/gh-pages/data/iris.csv

is `x` of the correct dimensions? see [documentation](https://docs.scipy.org/doc/numpy/reference/generated/numpy.dot.html) — Carlos Gonzalez, Aug 18 '17 at 12:07

score 1 · Answer 1 · answered May 24 '19 at 23:44

If you're reading the training data (data and predictions) from the file iris.csv

sepal_length,sepal_width,petal_length,petal_width,species
5.1,3.5,1.4,0.2,setosa
4.9,3,1.4,0.2,setosa

with:

data = pd.read_csv("iris.csv")

make sure that you define x as the first four columns, otherwise it will contain the strings from the last column:

X = data.iloc[:,0:4]

And prediction values:

y = data.iloc[:,5]
y = y.values.reshape(150,1)

Ishant Mrinal · Answer 2 · 2017-08-18T12:46:27.467

0

Following changes would help.

def fit(self, x, y):
    ...
    for xi, target in zip(x, y):
        update = self.eta * (target - self.predict(xi.reshape(1, x.shape[1]))
        ...

# Here if you want to implement perceptron, use matmul not dot product
def net_input(self, x):
    return np.matmul(x, self.w_[1:]) + self.w_[0]

edited Aug 18 '17 at 12:46

answered Aug 18 '17 at 12:40

Ishant Mrinal

4,898
3
29
47

`np.dot` is element-wise multiplication `np.matmul` is matrix multiplication. Here in this special case `(1, n)(n, 1)` it would be same; NOT OTHERWISE – Ishant Mrinal Aug 18 '17 at 13:17
For element-wise multiplication dimensions of the matrices have to be same. `np.multiply` is the element-wise multiplication. (Operator is `*`). For matrix multiplication result of `np.dot` and `np.matmul` (Operator is `@`) are same. – Ramesh-X Aug 18 '17 at 13:29

score 0 · Answer 3 · answered Aug 18 '17 at 13:08

0

Check the shape of x.

If it is (a, 1) where a is a number, use this:

def net_input(self, x):
    return np.dot(x.T, self.w_[1:]) + self.w_[0]

If it is (1, a) use this:

def net_input(self, x):
    return np.dot(x, self.w_[1:].T) + self.w_[0]

answered Aug 18 '17 at 13:08

Ramesh-X

4,853
6
46
67

hpaulj · Answer 4 · 2017-08-18T16:04:51.420

My guess is that x is an object dtype array of lists.

If I define an object dtype array:

In [45]: x=np.empty((2,),object)
In [46]: x[:]=[[1,2],[3,4]]
In [49]: x
Out[49]: array([list([1, 2]), list([3, 4])], dtype=object)

I get the same error with a list (or array) of floats:

In [50]: np.dot(x, [1.2,4.5])
...
TypeError: can't multiply sequence by non-int of type 'float'

If instead I give it integers, it works - sort of

In [51]: np.dot(x, [1,2])
Out[51]: [1, 2, 3, 4, 3, 4]

What it has actually done is [1,2]*1 and [3,4]*2, list replication. This is not numeric multiplication.

That's the only combination of variables that makes sense of the error message.

So you need to figure out why x is an object array. Often that's the result of building an array from lists that differ in length

In [54]: x = np.array([[1,2],[3,4,5]])
In [55]: x
Out[55]: array([list([1, 2]), list([3, 4, 5])], dtype=object)

So the basic question when faced with an error like this, what's the shape and dtype of the variables.

numpy.dot() gives TypeError: can't multiply sequence by non-int of type 'float'

4 Answers4

Linked