|
Daniel@0
|
1 <html>
|
|
Daniel@0
|
2 <head>
|
|
Daniel@0
|
3 <title>
|
|
Daniel@0
|
4 Netlab Reference Manual mdn
|
|
Daniel@0
|
5 </title>
|
|
Daniel@0
|
6 </head>
|
|
Daniel@0
|
7 <body>
|
|
Daniel@0
|
8 <H1> mdn
|
|
Daniel@0
|
9 </H1>
|
|
Daniel@0
|
10 <h2>
|
|
Daniel@0
|
11 Purpose
|
|
Daniel@0
|
12 </h2>
|
|
Daniel@0
|
13 Creates a Mixture Density Network with specified architecture.
|
|
Daniel@0
|
14
|
|
Daniel@0
|
15 <p><h2>
|
|
Daniel@0
|
16 Synopsis
|
|
Daniel@0
|
17 </h2>
|
|
Daniel@0
|
18 <PRE>
|
|
Daniel@0
|
19 net = mdn(nin, nhidden, ncentres, dimtarget)
|
|
Daniel@0
|
20 net = mdn(nin, nhidden, ncentres, dimtarget, mixtype, ...
|
|
Daniel@0
|
21 prior, beta)
|
|
Daniel@0
|
22 </PRE>
|
|
Daniel@0
|
23
|
|
Daniel@0
|
24
|
|
Daniel@0
|
25 <p><h2>
|
|
Daniel@0
|
26 Description
|
|
Daniel@0
|
27 </h2>
|
|
Daniel@0
|
28 <CODE>net = mdn(nin, nhidden, ncentres, dimtarget)</CODE> takes the number of
|
|
Daniel@0
|
29 inputs,
|
|
Daniel@0
|
30 hidden units for a 2-layer feed-forward
|
|
Daniel@0
|
31 network and the number of centres and target dimension for the
|
|
Daniel@0
|
32 mixture model whose parameters are set from the outputs of the neural network.
|
|
Daniel@0
|
33 The fifth argument <CODE>mixtype</CODE> is used to define the type of mixture
|
|
Daniel@0
|
34 model. (Currently there is only one type supported: a mixture of Gaussians with
|
|
Daniel@0
|
35 a single covariance parameter for each component.) For this model,
|
|
Daniel@0
|
36 the mixture coefficients are computed from a group of softmax outputs,
|
|
Daniel@0
|
37 the centres are equal to a group of linear outputs, and the variances are
|
|
Daniel@0
|
38 obtained by applying the exponential function to a third group of outputs.
|
|
Daniel@0
|
39
|
|
Daniel@0
|
40 <p>The network is initialised by a call to <CODE>mlp</CODE>, and the arguments
|
|
Daniel@0
|
41 <CODE>prior</CODE>, and <CODE>beta</CODE> have the same role as for that function.
|
|
Daniel@0
|
42 Weight initialisation uses the Matlab function <CODE>randn</CODE>
|
|
Daniel@0
|
43 and so the seed for the random weight initialization can be
|
|
Daniel@0
|
44 set using <CODE>randn('state', s)</CODE> where <CODE>s</CODE> is the seed value.
|
|
Daniel@0
|
45 A specialised data structure (rather than <CODE>gmm</CODE>)
|
|
Daniel@0
|
46 is used for the mixture model outputs to improve
|
|
Daniel@0
|
47 the efficiency of error and gradient calculations in network training.
|
|
Daniel@0
|
48 The fields are described in <CODE>mdnfwd</CODE> where they are set up.
|
|
Daniel@0
|
49
|
|
Daniel@0
|
50 <p>The fields in <CODE>net</CODE> are
|
|
Daniel@0
|
51 <PRE>
|
|
Daniel@0
|
52
|
|
Daniel@0
|
53 type = 'mdn'
|
|
Daniel@0
|
54 nin = number of input variables
|
|
Daniel@0
|
55 nout = dimension of target space (not number of network outputs)
|
|
Daniel@0
|
56 nwts = total number of weights and biases
|
|
Daniel@0
|
57 mdnmixes = data structure for mixture model output
|
|
Daniel@0
|
58 mlp = data structure for MLP network
|
|
Daniel@0
|
59 </PRE>
|
|
Daniel@0
|
60
|
|
Daniel@0
|
61
|
|
Daniel@0
|
62 <p><h2>
|
|
Daniel@0
|
63 Example
|
|
Daniel@0
|
64 </h2>
|
|
Daniel@0
|
65 <PRE>
|
|
Daniel@0
|
66
|
|
Daniel@0
|
67 net = mdn(2, 4, 3, 1, 'spherical');
|
|
Daniel@0
|
68 </PRE>
|
|
Daniel@0
|
69
|
|
Daniel@0
|
70 This creates a Mixture Density Network with 2 inputs and 4 hidden units.
|
|
Daniel@0
|
71 The mixture model has 3 components and the target space has dimension 1.
|
|
Daniel@0
|
72
|
|
Daniel@0
|
73 <p><h2>
|
|
Daniel@0
|
74 See Also
|
|
Daniel@0
|
75 </h2>
|
|
Daniel@0
|
76 <CODE><a href="mdnfwd.htm">mdnfwd</a></CODE>, <CODE><a href="mdnerr.htm">mdnerr</a></CODE>, <CODE><a href="mdn2gmm.htm">mdn2gmm</a></CODE>, <CODE><a href="mdngrad.htm">mdngrad</a></CODE>, <CODE><a href="mdnpak.htm">mdnpak</a></CODE>, <CODE><a href="mdnunpak.htm">mdnunpak</a></CODE>, <CODE><a href="mlp.htm">mlp</a></CODE><hr>
|
|
Daniel@0
|
77 <b>Pages:</b>
|
|
Daniel@0
|
78 <a href="index.htm">Index</a>
|
|
Daniel@0
|
79 <hr>
|
|
Daniel@0
|
80 <p>Copyright (c) Ian T Nabney (1996-9)
|
|
Daniel@0
|
81 <p>David J Evans (1998)
|
|
Daniel@0
|
82
|
|
Daniel@0
|
83 </body>
|
|
Daniel@0
|
84 </html> |
