PyTorch

When we using PyTorch, a useful Python deep learning framework for model training, maybe sometimes we forget to “store” the trained model, even unaware of that. (Such as me in the past)

Generative Adversarial Network (GAN) is a famous neural network model, its function is to input a set of noise and then generate a set of fake pictures through the Generator, and then use the Discriminator to distinguish whether it is a real picture.

If we have both the model classification results and correct answers, we can calculate the Binary Cross Entropy, it is a famous loss function.

When using sigmoid function in PyTorch as our activation function, for example it is connected to the last layer of the model as the output of binary classification. After all, sigmoid can compress the value between 0-1, we only need to set a threshold, for example 0.5 and you can divide the value into two categories.

Today if you are preprocessing some machine learning data, maybe you need to convert PyTorch tensor to one-hot encoding type. There is a intuitive method that is convert TENSOR to NUMPY-ARRAY, and then convert NUMPY-ARRAY to one-hot encoding type, just like this article: [Python] Convert the value to one-hot type in Numpy

NLLLoss is a loss function commonly used in multi-classes classification tasks. Its meaning is to take log the probability value after softmax and add the probability value of the correct answer to the average.

Today I want to record how to use Deep generative Adversarial Network (DCGAN) to implement a simple generate picture model. I wanted to demo with delicious snack pictures, but the effect was not very good, I downloaded half a million snack pictures in vain.

Finally, I used the official demo CelebA dataset.

Just as torchvision is a module in PyTorch that specializes in processing pictures, torchaudio to be recorded today is a module in PyTorch that specializes in processing audio.

Today we challenged the classifiers of different data sets again. This time, CIFAR-19 is a more difficult problem than MNIST handwriting recognition. In addition to the size of the picture becoming 32×32, CIFAR-10 is no longer a pure grayscale value, but a picture with the three primary colors of RGB.

“Use a toy dataset to train a classification model” is a simplest deep learning practice.

Today I want to record how to use MNIST A HANDWRITTEN DIGIT RECOGNITION dataset to build a simple classifier in PyTorch.