I remember SCaNN. Are there easy to use implementations available now?
You also seem unsure.
Personally, I love a cheat-sheet like in Slide 7 of the presentation. It's a game-tree for what ANN implementation to use, given your problem setting and the assumption you don't want to do groundbreaking ANN research; rather, you have an ANN problem you need to solve on the path to your actual problem. Maybe that's why SCaNN wasn't cited?
A quick google reveals that SCaNN is available on pypi from Google: https://pypi.org/project/scann/
Do any users have any feedback or experiences to share?
[edit: GPT4, comparing hnwslib and SCaNN: "Note that SCaNN recently ceased to be supported by Google Research, and there won't be any further improvements, bug fixes, or documentation updates from them. However, you can still use and get support from the community. This recent change in support could shift preference towards hnswlib for future applications."]
You also seem unsure.
Personally, I love a cheat-sheet like in Slide 7 of the presentation. It's a game-tree for what ANN implementation to use, given your problem setting and the assumption you don't want to do groundbreaking ANN research; rather, you have an ANN problem you need to solve on the path to your actual problem. Maybe that's why SCaNN wasn't cited?
A quick google reveals that SCaNN is available on pypi from Google: https://pypi.org/project/scann/ Do any users have any feedback or experiences to share?
[edit: GPT4, comparing hnwslib and SCaNN: "Note that SCaNN recently ceased to be supported by Google Research, and there won't be any further improvements, bug fixes, or documentation updates from them. However, you can still use and get support from the community. This recent change in support could shift preference towards hnswlib for future applications."]