[eggie5]

Interesting work on Online Learning:

There are many empirical studies which show for feature hashing, a few collisions don't have a sig impact on perf (https://youtu.be/ARjNMdCzN-Q?t=599).

However, for some archs, the impact is catastrophic. Eg matrix factorization. Any collision leads to an incorrect item. Zero Collision Hashing addresses the problem of mapping collisions. One technique is to introduce state into the hashing fn using the current id assignments.

real time publishing protocol:

* minute-level weight syncing * delta pushes only * ignore machine failures and rely on (possibly stale) full snapshot loading to bootstrap the new hosts

[somuchfordonor]

On the other hand, the end user can only distinguish two kinds of recommendations: (1) what is most popular over a period, and (2) what is similar to your most recent few likes/buys.

Besides, in a social media app, you can do whatever you want. To overcome limitations in (1), you can show whatever view counts you want. To overcome (2), you can take a video recorded a week ago, whose similarity you learned in a batch 3 days ago, and show it to a new user with a tag that says "1h ago."