python - Binary classification with PyTorch

Question

Below is code I've written for binary classification in PyTorch:

%reset -f

import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
import numpy as np
import matplotlib.pyplot as plt
import torch.utils.data as data_utils
import torch.nn as nn
import torch.nn.functional as F


device = 'cpu'
num_epochs = 10

hidden_size = 500 
num_classes = 2

learning_rate = .001

torch.manual_seed(24)

x1 = np.array([0,0])
x2 = np.array([0,1])
x3 = np.array([1,0])
x4 = np.array([1,1])
x = torch.tensor([x1,x2,x3,x4]).float()
y = torch.tensor([1,0,1,1]).long()

train = data_utils.TensorDataset(x,y)
train_loader = data_utils.DataLoader(train , batch_size=2 , shuffle=True)



input_size = len(x[0])

def weights_init(m):
    if type(m) == nn.Linear:
        m.weight.data.normal_(0.0, 1)

class NeuralNet(nn.Module) : 
    def __init__(self, input_size, hidden_size, num_classes) : 
        super(NeuralNet, self).__init__()
        self.fc1 = nn.Linear(input_size , hidden_size)
        self.fc2 = nn.Linear(hidden_size , 100)
        self.fc3 = nn.Linear(100 , num_classes)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x) : 
        out = self.fc1(x)
        out = self.sigmoid(out)
        out = self.fc2(out)
        out = self.sigmoid(out)
        out = self.fc3(out)
        out = self.sigmoid(out)
        return out

model = NeuralNet(input_size, hidden_size, num_classes).to(device)
model.apply(weights_init)

criterionCE = nn.BCELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
loss_values = []

for i in range(0 , 140) :
    print('in i' , i)
    total_step = len(train_loader)
    for epoch in range(num_epochs) :
        for i,(images , labels) in enumerate(train_loader) : 
            images = images.to(device)
            labels = labels.to(device)

            outputs = model(images)
            print('outputs' , outputs)
            loss = criterionCE(outputs , labels)
            loss_values.append(loss)
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()

    outputs = model(x)

    print(outputs.data.max(1)[1])

This code throws an error :

~/opt/miniconda3/envs/ds1/lib/python3.8/site-packages/torch/nn/functional.py in binary_cross_entropy(input, target, weight, size_average, reduce, reduction)
   2067                       stacklevel=2)
   2068     if input.numel() != target.numel():
-> 2069         raise ValueError("Target and input must have the same number of elements. target nelement ({}) "
   2070                          "!= input nelement ({})".format(target.numel(), input.numel()))
   2071 

ValueError: Target and input must have the same number of elements. target nelement (2) != input nelement (4)

I printed where I think is the cause of the error:

print('outputs' , outputs)

renders

outputs tensor([[9.9988e-01, 1.4011e-05],
        [9.9325e-01, 1.2087e-05]], grad_fn=<SigmoidBackward>)

So outputs should be a 2x1 instead of a 2x2 result. Have I not defined the setup of the model correctly? As I'm using sigmoid shouldn't a 2x1 output be created instead of 2x2?

question from:https://stackoverflow.com/questions/66052668/binary-classification-with-pytorch

Answer 1

If you are working on a binary classification task your model should only output one logit. Since you've set self.fc3 to have 2 neurons, you will get 2 logits as the output. Therefore, you should set self.fc3 as nn.Linear(100 , 1).