proyecto software

6
 Compet - Competitive transfer function. Syntax A = compet(N,FP) info = compet('code') Description Compet is a neural transfer function.  T ransfer fu nctions calculate a layer's output fr om its net input. A = compet(N,FP) takes N and optional function parameters, N S!y " matri# of net input (column) $ectors FP Struct of function parameters (i%nored) and returns t&e S!y" matri# A it& a in eac& column &ere t&e same column of N &as its ma#imum $alue, and else&ere. info = compet('code') returns information accordin% to t&e code strin% speci*ed+ compet('name') returns t&e name of t&is function. compet('out put',FP) returns t&e min ma#- output ran%e. compet('acti$e',FP) returns t&e min ma#- acti$e input ran%e. compet('fullderi$') returns or , dependin% on &et&er dAdN is S!yS!y " or S!y". compet('fpnames') returns t&e names of t&e function parameters. compet('fpdefaults') returns t&e default function parameters.

Upload: juangabyto

Post on 14-Jan-2016

5 views

Category:

Documents


0 download

DESCRIPTION

Consulta sobre herramientas de matlab.

TRANSCRIPT

Page 1: Proyecto Software

7/18/2019 Proyecto Software

http://slidepdf.com/reader/full/proyecto-software-5696aeabf1d78 1/6

Compet - Competitive transfer function.

Syntax

A = compet(N,FP)

info = compet('code')

Description

Compet is a neural transfer function.

 Transfer functions calculate a layer's output from its net input.

A = compet(N,FP) takes N and optional function parameters,

N S!y" matri# of net input (column) $ectors

FP Struct of function parameters (i%nored)

and returns t&e S!y" matri# A it& a in eac& column &ere t&e samecolumn of N &as its ma#imum $alue, and else&ere.

info = compet('code') returns information accordin% to t&e code strin%speci*ed+

compet('name') returns t&e name of t&is function.

compet('output',FP) returns t&e min ma#- output ran%e.

compet('acti$e',FP) returns t&e min ma#- acti$e input ran%e.

compet('fullderi$') returns or , dependin% on &et&er dAdN is S!yS!y" or S!y".

compet('fpnames') returns t&e names of t&e function parameters.

compet('fpdefaults') returns t&e default function parameters.

Page 2: Proyecto Software

7/18/2019 Proyecto Software

http://slidepdf.com/reader/full/proyecto-software-5696aeabf1d78 2/6

Example.

/ere you de*ne a net input $ector N, calculate t&e output, and plot !ot&it& !ar %rap&s.

n = 0 0 .10 .1-0

a = compet(n)0

su!plot(2,,), !ar(n), yla!el('n')

su!plot(2,,2), !ar(a), yla!el('a')

Assi%n t&is transfer function to layer i of a netork.

Elliotsig - Elliot symmetric sigmoid transfer function.

Syntax

A = elliotsi%(N)

Page 3: Proyecto Software

7/18/2019 Proyecto Software

http://slidepdf.com/reader/full/proyecto-software-5696aeabf1d78 3/6

Description

 Transfer functions con$ert a neural netork layer's net input into its netoutput.

A = elliotsi%(N) takes an S!y" matri# of S Nelement net input column$ectors and returns an S!y" matri# A of output $ectors, &ere eac&element of N is s3uas&ed from t&e inter$al inf inf- to t&e inter$al -it& an 4Ss&aped4 function.

 T&e ad$anta%e of t&is transfer function o$er ot&er si%moids is t&at it is fastto calculate on simple computin% &ardare as it does not re3uire anye#ponential or tri%onometric functions. 5ts disad$anta%e is t&at it only6attens out for lar%e inputs, so its e7ect is not as local as ot&er si%moid

functions. T&is mi%&t result in more trainin% iterations, or re3uire moreneurons to ac&ie$e t&e same accuracy.

Example.

Calculate a layer output from a sin%le net input $ector+

n = 0 0 .10 .1-0

a = elliotsi%(n)0

Plot t&e transfer function+

n = 1+.+10

plot(n, elliotsi%(n))

set(%ca,'dataaspectratio', -,'#%rid','on','y%rid','on')

For a netork you &a$e already de*ned, c&an%e t&e transfer function forlayer i+

Page 4: Proyecto Software

7/18/2019 Proyecto Software

http://slidepdf.com/reader/full/proyecto-software-5696aeabf1d78 4/6

Hardlims - Symmetric hard-limit transfer function.

Syntax

A = &ardlims(N,FP)

Description

&ardlims is a neural transfer function. Transfer functions calculate a layer'soutput from its net input.

A = &ardlims(N,FP) takes N and optional function parameters,

N S!y" matri# of net input (column) $ectors

FP Struct of function parameters (ignored)

and returns A, t&e S!y" 89: matri# it& 8s &ere N ; .

info = &ardlims('code') returns information accordin% to t&e code strin%speci*ed+

&ardlims('name') returns t&e name of t&is function.

&ardlims('output',FP) returns t&e min ma#- output ran%e.

&ardlims('acti$e',FP) returns t&e min ma#- acti$e input ran%e.

&ardlims('fullderi$') returns or , dependin% on &et&er dAdN is S!yS!y" or S!y".

&ardlims('fpnames') returns t&e names of t&e function parameters.

&ardlims('fpdefaults') returns t&e default function parameters.

Example.

/ere is &o to create a plot of t&e &ardlims transfer function.

n = 1+.+10

Page 5: Proyecto Software

7/18/2019 Proyecto Software

http://slidepdf.com/reader/full/proyecto-software-5696aeabf1d78 5/6

a = &ardlims(n)0

plot(n,a)

Assi%n t&is transfer function to layer i of a netork.

Netinv - Inverse transfer function.

Syntax

A = netin$(N,FP)

Description

netin$ is a transfer function. Transfer functions calculate a layer's outputfrom its net input.

A = netin$(N,FP) takes inputs.

N S!y" matri# of net input (column) $ectors

FP Struct of function parameters (i%nored)

Page 6: Proyecto Software

7/18/2019 Proyecto Software

http://slidepdf.com/reader/full/proyecto-software-5696aeabf1d78 6/6

and returns 9N.

info = netin$('code') returns information a!out t&is function. T&e folloin%codes are supported+

netin$('name') returns t&e name of t&is function.

netin$('output',FP) returns t&e min ma#- output ran%e.

netin$('acti$e',FP) returns t&e min ma#- acti$e input ran%e.

netin$('fullderi$') returns or , dependin% on &et&er dAdN is S!yS!y"or S!y".

netin$('fpnames') returns t&e names of t&e function parameters.

netin$('fpdefaults') returns t&e default function parameters.

Example.

/ere you de*ne *$eelement net input $ectors N and calculate A.

n = 2 < -0

a = netin$(n)0

Assi%n t&is transfer function to layer i of a netork.