GPN18:Good Patterns for Deep Learning with Tensorflow

aus dem Wiki des Entropia e.V., CCC Karlsruhe
Version vom 5. Mai 2018, 22:34 Uhr von Fahrplanbot (Diskussion | Beiträge)
(Unterschied) ← Nächstältere Version | Aktuelle Version (Unterschied) | Nächstjüngere Version → (Unterschied)
Wechseln zu:Navigation, Suche

Ein Vortrag von penguinmeanc3 auf der GPN18.

The talk will introduce a split into model, loss, dataset and estimator that keeps your code clean and easy to understand. For each of those 4 topics details how to implement it in an efficient, reusable way are explained.

The talk will assume basic knowledge about deep learning, since it will focus on implementation and no theory.

You will learn:

  • Writing good quality deep learning code
  • Implementing a model from scratch
  • Implementing a loss
  • Loading a dataset
  • Training a model

You will *not* learn:

  • What to use Deep Learning for?
  • What is a CNN, GAN, RNN, etc.?
  • That latest super weird feature of tensorflow.

One more thing: "The answer to the ultimate question of life, the universe and everything is 42." [The Hitchhiker's Guide to the Galaxy]