Mercurial > hg > camir-aes2014

comparison toolboxes/FullBNT-1.0.7/nethelp3.3/mlp.htm @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 <html>
2 <head>
3 <title>
4 Netlab Reference Manual mlp
5 </title>
6 </head>
7 <body>
8 <H1> mlp
9 </H1>
10 <h2>
11 Purpose
12 </h2>
13 Create a 2-layer feedforward network.
14
15 <p><h2>
16 Synopsis
17 </h2>
18 <PRE>
19 net = mlp(nin, nhidden, nout, func)
20 net = mlp(nin, nhidden, nout, func, prior)
21 net = mlp(nin, nhidden, nout, func, prior, beta)
22 </PRE>
23
24
25 <p><h2>
26 Description
27 </h2>
28 <CODE>net = mlp(nin, nhidden, nout, func)</CODE> takes the number of inputs,
29 hidden units and output units for a 2-layer feed-forward network,
30 together with a string <CODE>func</CODE> which specifies the output unit
31 activation function, and returns a data structure <CODE>net</CODE>. The
32 weights are drawn from a zero mean, unit variance isotropic Gaussian,
33 with varianced scaled by the fan-in of the hidden or output units as
34 appropriate. This makes use of the Matlab function
35 <CODE>randn</CODE> and so the seed for the random weight initialization can be
36 set using <CODE>randn('state', s)</CODE> where <CODE>s</CODE> is the seed value.
37 The hidden units use the <CODE>tanh</CODE> activation function.
38
39 <p>The fields in <CODE>net</CODE> are
40 <PRE>
41
42 type = 'mlp'
43 nin = number of inputs
44 nhidden = number of hidden units
45 nout = number of outputs
46 nwts = total number of weights and biases
47 actfn = string describing the output unit activation function:
48 'linear'
49 'logistic
50 'softmax'
51 w1 = first-layer weight matrix
52 b1 = first-layer bias vector
53 w2 = second-layer weight matrix
54 b2 = second-layer bias vector
55 </PRE>
56
57 Here <CODE>w1</CODE> has dimensions <CODE>nin</CODE> times <CODE>nhidden</CODE>, <CODE>b1</CODE> has
58 dimensions <CODE>1</CODE> times <CODE>nhidden</CODE>, <CODE>w2</CODE> has
59 dimensions <CODE>nhidden</CODE> times <CODE>nout</CODE>, and <CODE>b2</CODE> has
60 dimensions <CODE>1</CODE> times <CODE>nout</CODE>.
61
62 <p><CODE>net = mlp(nin, nhidden, nout, func, prior)</CODE>, in which <CODE>prior</CODE> is
63 a scalar, allows the field <CODE>net.alpha</CODE> in the data structure
64 <CODE>net</CODE> to be set, corresponding to a zero-mean isotropic Gaussian
65 prior with inverse variance with value <CODE>prior</CODE>. Alternatively,
66 <CODE>prior</CODE> can consist of a data structure with fields <CODE>alpha</CODE>
67 and <CODE>index</CODE>, allowing individual Gaussian priors to be set over
68 groups of weights in the network. Here <CODE>alpha</CODE> is a column vector
69 in which each element corresponds to a separate group of weights,
70 which need not be mutually exclusive. The membership of the groups is
71 defined by the matrix <CODE>indx</CODE> in which the columns correspond to
72 the elements of <CODE>alpha</CODE>. Each column has one element for each
73 weight in the matrix, in the order defined by the function
74 <CODE>mlppak</CODE>, and each element is 1 or 0 according to whether the
75 weight is a member of the corresponding group or not. A utility
76 function <CODE>mlpprior</CODE> is provided to help in setting up the
77 <CODE>prior</CODE> data structure.
78
79 <p><CODE>net = mlp(nin, nhidden, nout, func, prior, beta)</CODE> also sets the
80 additional field <CODE>net.beta</CODE> in the data structure <CODE>net</CODE>, where
81 beta corresponds to the inverse noise variance.
82
83 <p><h2>
84 See Also
85 </h2>
86 <CODE><a href="mlpprior.htm">mlpprior</a></CODE>, <CODE><a href="mlppak.htm">mlppak</a></CODE>, <CODE><a href="mlpunpak.htm">mlpunpak</a></CODE>, <CODE><a href="mlpfwd.htm">mlpfwd</a></CODE>, <CODE><a href="mlperr.htm">mlperr</a></CODE>, <CODE><a href="mlpbkp.htm">mlpbkp</a></CODE>, <CODE><a href="mlpgrad.htm">mlpgrad</a></CODE><hr>
87 <b>Pages:</b>
88 <a href="index.htm">Index</a>
89 <hr>
90 <p>Copyright (c) Ian T Nabney (1996-9)
91
92
93 </body>
94 </html>