Fork me on GitHub
#announcements
<
2019-05-02
>
chrisn16:05:23

The http://tech.ml system is upgrade from the ground up. Literally from the ground. ## tech.datatype This is the core of the system. It provides the ability to create native and jvm based contiguous contains of data, accelerated copying of data to/from native and jvm, and some basic mathematics. It also provides a graduated pathway from lisp sequence based programming to apl/numeric array-based programming. See the docs if you want to get into how that works. Finally it has sparse support at the base level so that moving towards NLP is easier. * Generic math/tensor support at the base level. * [apl version of game of life](https://github.com/techascent/tech.datatype/blob/master/test/tech/v2/apl/game_of_life.clj) * [sparse game of life](https://github.com/techascent/tech.datatype/blob/master/test/tech/v2/apl/sparse_game_of_life.clj) * kixi.stats and apache commons-math3 are included * [docs and cheatsheet](https://github.com/techascent/tech.datatype/tree/master/docs) * [accelerated blas support](https://github.com/techascent/tech.datatype/blob/master/src/tech/v2/tensor/impl.clj#L902) ## tech.ml.dataset Major refactor to remove the data-based etl abstraction (it is now functions) in an effort to reduce 'magic' and enable a more exploratory pathway. Given the datatype upgrade above it is also just less necessary; you can do some of your exploration without the dataset library. * [walkthrough](https://github.com/techascent/tech.ml.dataset) ## http://tech.ml Mainly upgraded to sit on top of both previously mentioned libraries. Same great flavor, lots more horsepower especially in the exploratory case: * [walkthrough](https://github.com/techascent/tech.ml) * [real world example](https://github.com/cnuernber/ames-house-prices/blob/master/src/clj_ml_wkg/ames_house_prices.clj) Scroll down to end of real world example for demonstration of moving from exploratory phase to actual production.

👏 48
😮 16
👍 4
otfrom18:05:06

Great stuff. Looking forward to digging into it especially to see how the kixi.stats stuff will work

chrisn16:05:37

The gist is with kixi that anything produced via any vector statement involving an iterable or a reader implements both java.lang.Iterable and potentially java.util.List and java.util.RandomAccess. So kixi works with the lazy readers and such out of the box with no further changes. As do most of the clojure functions (count, vec, etc).

danielcompton19:05:14

This quarter Clojurists Together is funding Fireplace and Boot $9k each! boot-clj https://www.clojuriststogether.org/news/q2-2019-funding-announcement/

👍 84
💯 16
💥 8