{"id":4842,"date":"2020-05-15T15:00:00","date_gmt":"2020-05-15T13:00:00","guid":{"rendered":"https:\/\/www.codemotion.com\/magazine\/?p=4842"},"modified":"2022-08-03T12:01:48","modified_gmt":"2022-08-03T10:01:48","slug":"how-to-build-a-gan-in-python","status":"publish","type":"post","link":"https:\/\/www.codemotion.com\/magazine\/ai-ml\/deep-learning\/how-to-build-a-gan-in-python\/","title":{"rendered":"How to build a GAN in Python"},"content":{"rendered":"\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\tTable Of Contents\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t
  1. Introduction<\/a>
  2. What is a Generative Adversarial Network?<\/a>
  3. A homemade GAN<\/a>
  4. An artificial dataset<\/a>
  5. Our GAN in small pieces<\/a>
  6. How is a GAN trained?<\/a>
  7. Conclusions<\/a><\/ol>\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\n\n\n

    Introduction<\/h2>\n\n\n\n

    Generative Adversarial Networks<\/span><\/strong> (GANs<\/strong>) are a hot topic in machine learning<\/span> for several good reasons. Here are three of the best:<\/p>\n\n\n\n

    1. GANs can provide astonishing results, creating new things (images, texts, sounds, etc.) by imitating samples they have previously been exposed to.<\/li>
    2. A GAN offers a new paradigm in machine learning – a generative<\/em> one – that combines pre-existing techniques to provide both current and brand new ideas and results.<\/li>
    3. GANs are a recent (2014) creation of Ian Goodfellow, the former Google, now Apple, researcher (also the co-author of a standard reference in deep learning with Joshua Bengio and Aaron Courville).<\/li><\/ol>\n\n\n\n

      It is likely that readers will already have encountered some of the impressive results GANs are capable of, especially in the realm of image processing<\/span>. Such networks are able, upon request, to draw a picture<\/span> of a red flower, a black bird or even a violet cat. Furthermore, that flower, bird, or cat does not exist at all in reality, but is entirely the product of the network’s ‘imagination’.<\/p>\n\n\n

      \n
      \"\"
      These images are not photos of real people – they have been generated by a properly trained GAN!<\/figcaption><\/figure><\/div>\n\n\n

      How is this possible, and can we share in the fun? This article endeavours to address both questions, using functional Python<\/strong> code that can be run on your laptop. You may need to add some packages that are missing from your Python<\/span> installation, but that\u2019s what Pip is there for\u2026<\/p>\n\n\n\n

      What is a Generative Adversarial Network?<\/h2>\n\n\n\n

      Neural networks<\/span> (NNs) were devised as prediction and classification<\/span> models<\/span>. They are powerful, non-linear optimizers which can be trained to evolve their inner parameters<\/span> (neuron weights<\/span>) to fit the training data<\/span>. This will enable the NN to predict and classify<\/span> unknown data<\/span> of the same kind.<\/p>\n\n\n\n

      We all know how impressive the data<\/span> approximations<\/span> of neural networks, in which ‘data<\/span>‘ can mean just about anything, can be. However, the features<\/span> of such algorithms also suggest some of their drawbacks, such as:<\/p>\n\n\n\n