不存在的人 深度学习_这个人不存在

不存在的人 深度学习

Yes, you heard me right. This man does not exist. He was created by a Generative Adversarial Network.

是的，你明白我的意思了。 这个人不存在。 他是由对抗性网络创建的。

什么是一般对抗网络？ (What is a General Adversarial Network?)

General Adversarial Networks, or GANs for short, are a type of neural network for generative machine learning. They are able to accurately recreate similar, but not identical, content to what they are fed in.

通用对抗网络(简称GAN)是一种用于生成机器学习的神经网络。 他们能够准确地重新创建与其所提供内容相似但不完全相同的内容。

Let me give an example,

让我举个例子吧，

http://yann.lecun.com/exdb/mnist/ //yann.lecun.com/exdb/mnist/

This is the MNIST handwritten digit dataset. This dataset includes over 70,000 images of handwritten digits. I trained a GAN that I created on this data, to generate random handwritten digits.

这是MNIST手写数字数据集。 该数据集包括超过70,000张手写数字图像。 我训练了在此数据上创建的GAN，以生成随机的手写数字。

GAN如何工作？ (How do GANs work?)

A GAN consists of two parts: A generator and a discriminator.

GAN由两部分组成：生成器和鉴别器。

The generator is a Neural Network that takes in random values and returns a long array of pixel values, that can be reconstructed to form images. The discriminator is another separate Neural Network that compares “real” and “fake” images, and tries to guess if they are real or fake.

生成器是一个神经网络，它接收随机值并返回一长列像素值，可以将其重建以形成图像。 鉴别器是另一个单独的神经网络，用于比较“真实”和“伪造”图像，并试图猜测它们是真实的还是伪造的。

The adversarial part of the GAN is how they work together and feed into each other: When training the GAN, the loss value for the generator is how accurate the discriminator is. The worse the discriminator performs, the better the generator is performing. On the other hand, the loss value of the discriminator is based on the accuracy of the predictions.

GAN的对抗部分是它们如何协同工作并相互馈送：训练GAN时，发生器的损耗值就是鉴别器的准确性。 鉴别器的性能越差，生成器的性能就越好。 另一方面，鉴别器的损失值是基于预测的准确性。

This means that the two Neural Networks are competing against each other: One is trying to trick the other, while the other tries to avoid being tricked.

这意味着两个神经网络正在相互竞争：一个试图欺骗另一个，而另一个试图避免被欺骗。

GAN的优势： (Advantages of GANs:)

• Unsupervised Learning
  无监督学习

Although the GAN itself is a form of Supervised Learning, the relationship between the generator and discriminator is unsupervised. This means that less data is required at every level of the network.

尽管GAN本身是监督学习的一种形式，但生成器和判别器之间的关系是不受监督的。 这意味着在网络的每个级别上需要的数据更少。

• Highly applicable
  高度适用

Since the generator and discriminator of the data have convolutional layers as their input layers, the data for GANs usually come in the form of images. Since images can be expressed as a long array of numbers, most numerical data can be composed into images and are therefore compatible with GANs.

由于数据的生成器和鉴别器具有卷积层作为其输入层，因此GAN的数据通常以图像形式出现。 由于图像可以表示为一长串数字，因此大多数数值数据都可以组成图像，因此与GAN兼容。

GAN的缺点： (Disadvantages of GANs:)

• Long Computation Time
  计算时间长

Because of the nested neural networks within the GAN, it can take a long time to train it. A good GPU is a necessity for training GANs.

由于GAN中嵌套了神经网络，因此训练它可能需要很长时间。 良好的GPU是训练GAN的必要条件。

• Possible Collapse
  可能的崩溃

The balance between the generator and the discriminator is very fragile. If there is a local minima for the generator, it might start creating unrecognizable generations that, by coincidence, perfectly fool the discriminator. This would happen more for images in which there is no clear pattern, as it would pick it up on false signals.

生成器和鉴别器之间的平衡非常脆弱。 如果生成器存在局部最小值，则它可能会开始创建无法识别的世代，巧合的是，它们完全愚弄了鉴别器。 对于没有清晰图案的图像，这种情况会更多发生，因为它会在错误信号上拾取它。

• No True Way to Evaluate Model
  没有真正的模型评估方法

On the premise that GANs would generate original image, there would be no objective, numerical way to check how accurate the recreations of the GAN is. One can only hope that the GAN will do its job.

在GAN会生成原始图像的前提下，将没有客观，数值的方式来检查GAN的休闲度有多准确。 人们只能希望GAN能够完成工作。

结论： (Conclusion:)

Credit to: https://thispersondoesnotexist.com/

致谢： https : //thispersondoesnotexist.com/

Although the advances of the GANs might seem promising, we need to be aware of the obstacles that GANs need to overcome. I would like to send you off with this funny and slightly unsettling fail from thispersondoesnotexist:

尽管GAN的进步似乎很有希望，但我们需要意识到GAN需要克服的障碍。 我想送你一个来自thispersondoesnotexist的有趣而令人不安的失败：

It might seem like there are more disadvantages than advantages, but I promise you that the GANs have large potential in the future to change the field of machine learning forever.

劣势似乎多于优势，但我向您保证，GAN在未来有巨大潜力来永久改变机器学习领域。

Thank you for reading my article, I hope you learnt. something!

感谢您阅读我的文章，希望您能学到。 东西！

翻译自: https://medium.com/analytics-vidhya/this-man-does-not-exist-c9be2cf985e2

不存在的人 深度学习