Yury Nebieridze

Hi Chris,

Did your NASA experience influence the book?
Meaning - any specific optimization techniques, choice of CPU/GPU etc.
(you mentioned text over image in your YouTube presentation, for e.g.)

Hi Chris,

What's your cloud platform preference?
Since TensorFlow is made by Google, looks like GCP is a natural choice, providing it offers features like Cloud TPUs.
Please share your insight.

Welcome to the Ranch, Chris!

Welcome to the Ranch, Ken Youens-Clark!

Welcome to the Ranch, Rishal Hurbans!

Welcome Thushan Ganegedara!

Welcome to the Ranch, Nicolò and Gianluca!

Welcome, Alexander and Brandon.

Hi Jakub and Vladimir,

Are GANs solving a specific class of tasks, such as image recognition?
I ask it because the original use of this technique and its success in that field of data.

Also, looks like Generator very much depends on Discriminator, that is quality of latter's data set.
What if the potential subject area of samples (let's say - corpora) for the data set is vast enough (space level or sub electron level volumes)?
Wouldn't this produce special cases only as a result, if any?

Also, if there is a game analogy, GANs are like tennis (or criminal - detective).
Are there other heuristics in GANs, such as team play, meaning number of players more than 2?    

Thank you,
Yury

Thanks Stephen and Tim for sharing your Kubernetes sentiments . Understood.
Stephen, special thanks for the blog reference, I subscribed and looked through it. Will be reading it.

Hi Jeff and Stephen,

In your book you use Docker Swarm for services orchestration.
Do you prefer it to Kubernetes or Mesos due to it's Docker native, simpler?

Also, the subject of Orchestration probably needs a separate cookbook.
Orchestrating services / micro-services brings many tasks to be solved.

Welcome Jeff and Stephen!

This is true, Sathya. But true is also that even big companies strive to keep up with the new, advanced methods of SDLC and are trying to embrace a faster technology turnover. Yes, this is to take risk, but risk, with technologies like Kotlin can be pretty calculated and justified, IMHO. And also, there is another, much bigger, risk of being left off behind the competition. Think AI/ML/DL, cloud computing and many more, 5 years ago versus now. Every big name wants a piece of a pie and a new, robust, versatile language is a serious part of the game.

If I may humbly add to the comprehensive and great answer of Ken: in today's world there is no working application cast in stone, meaning it's a living organism, literally living online! Thus - it must be maintained daily, must improve / grow some functionality, add new one and, yes, get rid of obsolete features, similarly to the old skin and cells we lose daily. Thus, yes, in my opinion, you can use Kotlin in your project along with Java, first to augment (wrapping) features and then,gradually, moving to Kotlin only code. Sure, you have to start with good set of tests, based on your design goals.    

Agreed, Kotlin indeed is concise and usable, so I share your concurrency (and not only) manageability optimism. Need more practice and experience though .

Yes, I noticed code generation issues, like for JVM of javascript, so code consistency looks like an issue in multi-platform promise. But, same here, deeper insight is required. Thanks again for an interesting discussion.