![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Research updates
Sep. 10th, 2016 08:54 pm1) Dataflow matrix machines were presented at SumTopo conference in Leicester (early August).
2) It turned out that LSTM and Gated Recurrect Unit networks can be thought of as "vanilla RNNs using linear and bilinear neurons together with sigmoid neurons" (mid August).
3) Prototypes of "lightweight pure dataflow matrix machines" were published in the Fluid project repository (late August).
4) There is a new effort ongoing for the recent weeks, which is a collaboration around trying to use Clojure for data flow matrix machines. Clojure seems to be a nice fit for this architecture. So there might be more Clojure posts in this blog (cf. a series of Clojure-related posts in February 2015). One Clojure-related thing to note: the Nightcode editor progressed a lot, and its Web site moved here: https://sekao.net/nightcode/
2) It turned out that LSTM and Gated Recurrect Unit networks can be thought of as "vanilla RNNs using linear and bilinear neurons together with sigmoid neurons" (mid August).
3) Prototypes of "lightweight pure dataflow matrix machines" were published in the Fluid project repository (late August).
4) There is a new effort ongoing for the recent weeks, which is a collaboration around trying to use Clojure for data flow matrix machines. Clojure seems to be a nice fit for this architecture. So there might be more Clojure posts in this blog (cf. a series of Clojure-related posts in February 2015). One Clojure-related thing to note: the Nightcode editor progressed a lot, and its Web site moved here: https://sekao.net/nightcode/
