Hey everyone, so I finally got the courage to say "hi" and post in the forums. I am really glad they are finally back.
As one of my first posts, I am thinking to start blog or thread. The blog will focus on the development of an algorithm to improve Twitch pulse and recommendations. As a way to give back to the community, I was going to blog so to make the re-write of the algorithm and concepts open source. I need some advice though as I have never written a good thread before...
Where would be a good place to start such a thread/blog in the forums?
How can I write a good thread? Does anyone have tips for beginners to blogging or writing threads?
More Details:
- [+] Show
- Way back when I was younger and stupid in 2010, I made an algorithm (or process) to make Youtube's search and later recommendations more accurate for me. Over the years, I have tried to improve it. As I watched new around "Youtube" get more heated, I discovered that the algorithm fixed, on my account, many the problems people were having(**). When Cohh-Carnage talked Twitch's later last month, he inspired me to think... "What if I tried to apply the algorithm to Twitch? Could it improve Twitch Pulse or Recommendations?" Twitch both its community and tech support have been very welcoming, so it might be very workable. Over the past month, I have manually walked through the algorithm in order to develop training data. I think the concepts are at the professional level for me to share with others.
The only problem is I have never written a blog or thread, so I was hoping for some tips on how to write a good one?
P.S. Since 2010, I would come to later learn that the process (I made) is called discriminative network in GAN algorithm. Here is a quick current diagram of the algorithm process, but applied to Twitch:
P.P.S. (*)Specifically, for the test account I have applied it to my recommendation, feeds only have malicious/fake news after 57th ranked video, I always see notifications, and video are never suppressed from my subscription feed. I do feel like the test account misses out though as what is "trending" or "popular" rarely appears in that account's recommendation feeds.
