dmm | Entries tagged with my papers

The project aiming to produce a Clojure implementation of DMMs (dataflow matrix machines) is now underway and has already produced some very nice fruits (a new vector space of recurrent maps, a nice Clojure implementation of linear algebra for those, a new interesting architecture for neurons (to be elaborated upon later)).

A talk on DMMs was given to NEPLS on Oct 7. Slides are posted.

A theoretical preprint on dataflow matrix machines was posted to arxiv.org on Oct 4 (this concludes the current series of preprints on dataflow matrix machines as a generalization of recurrent neural nets).

A preprint on the relationships between the rational-imaginary values for the "quantum of area" (Barbero-Immirzi parameter) and the finite-dimensional non-unitary representations of the Lorenz group was posted to arxiv.org on Oct 2.

It turns out that the best way to create self-referential generalized recurrent neural networks (dataflow matrix machines) is to allow the network to work with streams of matrices describing connectivity patterns and weights of the networks of this kind:

http://arxiv.org/abs/1605.05296

Today we mark the emergence of the first simple program written in the experimental dialect of the new Fluid programming language (a language for specifying dataflow matrix machines).

We now have a new formalism describing a large class of data flow graphs as matrices of real numbers. This allows for continuous program transformations and interesting higher-order techniques.

A preprint, open source prototype and videos are now available (github, user anhinga, repository fluid).

We now have an open-sourced dynamic dataflow code, with the capability of containing the changing dataflow graph within one of its own nodes; so the graph is visualized as it changes together with the animations it produces; there are also videos of that on YouTube now, linked from the open source project page.

We now also have a draft of the paper describing the dynamic data flow and almost continuous transformations, and it should be posted as a preprint within a week and also submitted to a conference.

A preprint, "Linear models of computation and program learning", was written.

Some progress in the meditation practice during that period; a trip to NYC to give a math talk took place on Apr.9-11; and now we are on the eve of a car trip from Boston to Charleston, Savanna, and Florida.

S	M	T	W	T	F	S
			1	2	3	4
5	6	7	8	9	10	11
12	13	14	15	16	17	18
19	20	21	22	23	24	25
26	27	28	29	30

Dataflow matrix machines (by Anhinga anhinga)

Entries tagged with my papers

Current events

Self-referential neural networks; first program in the new experimental programming language Fluid

Data flow graphs as matrices

The state of things

Update since the last post

Profile

April 2026

Syndicate

Most Popular Tags

Page Summary

Active Entries

Style Credit

Expand Cut Tags