Re: Simple Neural Network DSL -- request for feedback

2010-11-14 Thread Carson
It might be interesting to add a Clojure frontend to Nengo. Having said that, Nengo is very "opinionated" in the sense that you have to buy in to the kind of neural simulation Eliasmith is doing (and the way he's doing it), which Eric may not be interested in doing. There are many other kinds of

Re: Simple Neural Network DSL -- request for feedback

2010-11-13 Thread Ross Gayler
You might also consider using your DSL as a frontend to the Nengo neural simulator (http://nengo.ca). Nengo (which is written in Java) has recently added a Python scripting interface (http://www.frontiersin.org/neuroinformatics/ 10.3389/neuro.11/007.2009/abstract). Nengo has a lot to recommend it

Re: Simple Neural Network DSL -- request for feedback

2010-11-13 Thread Eric Schulte
#+begin_src clojure (let [n {:phi identity :accum (comp (partial reduce +) (partial map *)) :weights [2 2 2]}] [(repeat 3 n) (repeat 5 n) (assoc n :weights (vec (repeat 5 1)))]) #+end_src would result in the following connection pattern [[file:/tmp/layers.png]] > Howe

Re: Simple Neural Network DSL -- request for feedback

2010-11-12 Thread Ross Gayler
On Nov 13, 9:12 am, "Eric Schulte" wrote: > Albert Cardona writes: > > > Your neural network DSL looks great. One minor comment: why use lists > > instead of sets? ... > > I used lists because I want to be able to specify a network in which (at > least initially) all neurons in a hidden layer a

Re: Simple Neural Network DSL -- request for feedback

2010-11-12 Thread Eric Schulte
Hi Albert, Albert Cardona writes: > Hi Eric, > > Your neural network DSL looks great. One minor comment: why use lists > instead of sets? In the webpage you state: > > "Lists are used to represent a unordered series" > I used lists because I want to be able to specify a network in which (at lea

Re: Simple Neural Network DSL -- request for feedback

2010-11-12 Thread Eric Schulte
Carson writes: > Hi! That looks interesting. I'm curious how big a network are you > intending to experiment with? (ie, # of layers, size of layers?). > I haven't really thought about limits on the size of the networks, although I suppose with very large networks it may become a good idea to

Re: Simple Neural Network DSL -- request for feedback

2010-11-12 Thread Ken Wesson
On Fri, Nov 12, 2010 at 9:15 AM, Albert Cardona wrote: > Hi Eric, > > Your neural network DSL looks great. One minor comment: why use lists > instead of sets? In the webpage you state: > > "Lists are used to represent a unordered series" > > ... but lists are generally considered data structures w

Re: Simple Neural Network DSL -- request for feedback

2010-11-12 Thread Albert Cardona
Hi Eric, Your neural network DSL looks great. One minor comment: why use lists instead of sets? In the webpage you state: "Lists are used to represent a unordered series" ... but lists are generally considered data structures whose elements are accessible by index. The closest representation to

Re: Simple Neural Network DSL -- request for feedback

2010-11-11 Thread Carson
Hi! That looks interesting. I'm curious how big a network are you intending to experiment with? (ie, # of layers, size of layers?). Carson On Nov 11, 8:17 am, "Eric Schulte" wrote: > Hi Saul, > > Saul Hazledine writes: > > On Nov 10, 11:20 pm, "Eric Schulte" wrote: > >> Hi, > > >> Inspired

Re: Simple Neural Network DSL -- request for feedback

2010-11-11 Thread Eric Schulte
Sean Grove writes: > Very elegant. From the example, it looks like it takes away a great > deal of the tedium of neural networks. > Thanks, that is certainly the goal. > > Do you have any more significant code examples than those listed on [3]? > At the bottom of [3] I link to both a perceptro

Re: Simple Neural Network DSL -- request for feedback

2010-11-11 Thread Eric Schulte
Hi Saul, Saul Hazledine writes: > On Nov 10, 11:20 pm, "Eric Schulte" wrote: >> Hi, >> >> Inspired by cgrand's regexp example [1], I've implemented a simple DSL >> for specifying neural networks using Clojure data types.   > > This is really clear. The web page documentation is awesome. > Grea

Re: Simple Neural Network DSL -- request for feedback

2010-11-10 Thread Saul Hazledine
On Nov 10, 11:20 pm, "Eric Schulte" wrote: > Hi, > > Inspired by cgrand's regexp example [1], I've implemented a simple DSL > for specifying neural networks using Clojure data types.   This is really clear. The web page documentation is awesome. > Construction of this simple language involved a

Re: Simple Neural Network DSL -- request for feedback

2010-11-10 Thread Greg
Awesome! Thanks for sharing! :-) On Nov 10, 2010, at 2:20 PM, Eric Schulte wrote: > Hi, > > Inspired by cgrand's regexp example [1], I've implemented a simple DSL > for specifying neural networks using Clojure data types. The code is > available in this gist [2], and a brief introduction with s

Re: Simple Neural Network DSL -- request for feedback

2010-11-10 Thread Sean Grove
Very elegant. From the example, it looks like it takes away a great deal of the tedium of neural networks. Do you have any more significant code examples than those listed on [3]? Sean On Nov 10, 2010, at 2:20 PM, Eric Schulte wrote: > Hi, > > Inspired by cgrand's regexp example [1], I've imp