Deep Learning with PyTorch
上QQ阅读APP看书,第一时间看更新

PyTorch non-linear activations

PyTorch has most of the common non-linear activation functions implemented for us already and it can be used like any other layer. Let's see a quick example of how to use the ReLU function in PyTorch:

sample_data = Variable(torch.Tensor([[1,2,-1,-1]]))
myRelu = ReLU()
myRelu(sample_data)

Output:

Variable containing:
1 2 0 0
[torch.FloatTensor of size 1x4]

In the preceding example, we take a tensor with two positive values and two negative values and apply a ReLU on it, which thresholds the negative numbers to 0 and retains the positive numbers as they are.

Now we have covered most of the details required for building a network architecture, let's build a deep learning architecture that can be used to solve real-world problems. In the previous chapter, we used a simple approach so that we could focus only on how a deep learning algorithm works. We will not be using that style to build our architecture anymore; rather, we will be building the architecture in the way it is supposed to be built in PyTorch.