Perceptron train

[lordpba]

Hi!

I have done the Perceptron as in 1.6

but when I do the training I have the error: TypeError: can't multiply sequence by non-int of type 'float'

this is my code:

class Perceptron:
def init(self, num_features):
self.num_features = num_features
self.weights = [0.0 for _ in range(num_features)]
self.bias = 0.0

def forward(self, x):
    weighted_sum_z = self.bias
    for i, _ in enumerate(self.weights):
        weighted_sum_z += x[i] * self.weights[i] #qui vengono moltiplicate tutte le features (x) per i corrispondenti pesi (w)
    
    if weighted_sum_z > 0.0: #qui applichiamo il 'treshold' dell'attivazione 0 o 1 se la somma dei pesi e' appunto 0 o 1
        prediction = 1
    else:
        prediction = 0
    
    return prediction

def update(self, x, true_y):
    prediction = self.forward(x)
    error = true_y - prediction
    # da qui abbiamo l'aggiornamento dei pesi del perceptron dati dall'errore x la feature
    self.bias += error
    for i, _ in enumerate(self.weights):
        self.weights[i] += error * x[i]

    return error

def train(model, all_x, all_y, epochs):

for epoch in range(epochs):
    error_count = 0

    for x, y in zip(all_x, all_y):
        error = model.update(x, y)
        error_count += abs(error)
    
    print(f"Epoch {epoch+1} errors {error_count}")

---

ppn = Perceptron(num_features=2)

train(model=ppn, all_x=X_train, all_y=y_train, epochs=5)

TypeError Traceback (most recent call last)
Cell In[13], line 3
1 ppn = Perceptron(num_features=2)
----> 3 train(model=ppn, all_x=X_train, all_y=y_train, epochs=5)

Cell In[11], line 7, in train(model, all_x, all_y, epochs)
4 error_count = 0
6 for x, y in zip(all_x, all_y):
----> 7 error = model.update(x, y)
8 error_count += abs(error)
10 print(f"Epoch {epoch+1} errors {error_count}")

Cell In[9], line 20, in Perceptron.update(self, x, true_y)
19 def update(self, x, true_y):
---> 20 prediction = self.forward(x)
21 error = true_y - prediction
22 # da qui abbiamo l'aggiornamento dei pesi del perceptron dati dall'errore x la feature

Cell In[9], line 10, in Perceptron.forward(self, x)
8 weighted_sum_z = self.bias
9 for i, _ in enumerate(self.weights):
---> 10 weighted_sum_z += x[i] * self.weights[i] #qui vengono moltiplicate tutte le features (x) per i corrispondenti pesi (w)
12 if weighted_sum_z > 0.0: #qui applichiamo il 'treshold' dell'attivazione 0 o 1 se la somma dei pesi e' appunto 0 o 1
13 prediction = 1

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

[rasbt]

Hmm, to help finding out what's going on, can you modify this

    for i, _ in enumerate(self.weights):
        weighted_sum_z += x[i] * self.weights[i] 

to

    for i, _ in enumerate(self.weights):
        print(x[i])
        weighted_sum_z += x[i] * self.weights[i] 

to find out what could be causing this?

[lordpba]

got it! it was an error by me in loading the dataset:

before:

X_train = df[['x1','x2']]
y_train = df['label']

after (correct):

X_train = df[["x1", "x2"]].values
y_train = df["label"].values

now all ok!

Thanks for quick reply!!!

[rasbt]

Nice, glad you could fix it!