Avi Bryant
Avi Bryant
load.bzl
The README refers to `load.bzl` but doesn't mention that you have to write it yourself, or link to any default implementation. If you copy the one from this repo's 3rdparty,...
This is just a start, and shouldn't be merged yet, but: defining some Bufferable implementations for Algebird objects so that we don't have to use Kryo for long-term serialized storage.
See http://comonad.com/reader/2008/linear-bloom-filters/: > So to recap, we took a normal (or counted or spectral) Bloom filter, crossbred it with Litwin's linear hash table and found that the mutant offspring is...
See eg http://www.cse.cuhk.edu.hk/~taoyf/course/5705f10/lec8.pdf for how this works. I have a basic working implementation of the DyadicRange logic (for positive integer keys, which you can map whatever else to), but it...
HyperLogLog has an updateInto method which updates a mutable buffer representing its current state, instead of doing immutable operations. This allows for an efficient implementation of sumOption. It would be...
SummingCache could buffer up multiple values for a key and sum them in one pass before flush using sumOption; this would trade memory use for performance. If we standardized a...
For counts-based data, this may be a better approach than the current exponential decay. See http://word.bitly.com/post/41284219720/forget-table
Check out this JSONProtobuf conversion package: https://github.com/turn/shapeshifter Would be neat to have a Bijection that wrapped this.
This makes a few related changes to `Evaluator`: - it decouples it from `Split` and just has it directly operate on `Iterable[T]`, which was the only part of a `Split`...
attn @tixxit @non The intent here is to capture the basic mechanics of various training steps - updateTargets, expand, prune, etc - in a way that can be reused in...