Source Code for Module ML.Neural.Network

  1  ## Automatically adapted for numpy.oldnumeric Jun 27, 2008 by -c 
  2   
  3  # 
  4  #  Copyright (C) 2000-2008  greg Landrum 
  5  # 
  6  """ Contains the class _Network_ which is used to represent neural nets 
  7   
  8  **Network Architecture:** 
  9   
 10    A tacit assumption in all of this stuff is that we're dealing with 
 11    feedforward networks. 
 12   
 13    The network itself is stored as a list of _NetNode_ objects.  The list 
 14    is ordered in the sense that nodes in earlier/later layers than a 
 15    given node are guaranteed to come before/after that node in the list. 
 16    This way we can easily generate the values of each node by moving 
 17    sequentially through the list, we're guaranteed that every input for a 
 18    node has already been filled in. 
 19   
 20    Each node stores a list (_inputNodes_) of indices of its inputs in the 
 21    main node list. 
 22   
 23  """ 
 24  import numpy 
 25  import random 
 26   
 27  from ML.Neural import NetNode, ActFuncs 
 28   
 29  # FIX: this class has not been updated to new-style classes 
 30  # (RD Issue380) because that would break all of our legacy pickled 
 31  # data. Until a solution is found for this breakage, an update is 
 32  # impossible. 
 33 -class Network: 
 34    """ a neural network 
 35   
 36    """ 
 37 -  def ConstructRandomWeights(self,minWeight=-1,maxWeight=1): 
 38      """initialize all the weights in the network to random numbers 
 39   
 40        **Arguments** 
 41   
 42          - minWeight: the minimum value a weight can take 
 43   
 44          - maxWeight: the maximum value a weight can take 
 45           
 46      """ 
 47      for node in self.nodeList: 
 48        inputs = node.GetInputs() 
 49        if inputs: 
 50          weights = [random.uniform(minWeight,maxWeight) for x in range(len(inputs))] 
 51          node.SetWeights(weights) 
 52       
 53   
 54 -  def FullyConnectNodes(self): 
 55      """ Fully connects each layer in the network to the one above it 
 56   
 57   
 58       **Note** 
 59         this sets the connections, but does not assign weights 
 60          
 61      """ 
 62      nodeList = range(self.numInputNodes) 
 63      nConnections = 0 
 64      for layer in xrange(self.numHiddenLayers): 
 65        for i in self.layerIndices[layer+1]: 
 66          self.nodeList[i].SetInputs(nodeList) 
 67          nConnections = nConnections + len(nodeList) 
 68        nodeList = self.layerIndices[layer+1] 
 69   
 70      for i in self.layerIndices[-1]: 
 71        self.nodeList[i].SetInputs(nodeList) 
 72        nConnections = nConnections + len(nodeList) 
 73      self.nConnections = nConnections 
 74       
 75 -  def ConstructNodes(self,nodeCounts,actFunc,actFuncParms): 
 76      """ build an unconnected network and set node counts 
 77   
 78        **Arguments** 
 79   
 80          - nodeCounts: a list containing the number of nodes to be in each layer. 
 81             the ordering is: 
 82              (nInput,nHidden1,nHidden2, ... , nHiddenN, nOutput) 
 83   
 84      """ 
 85      self.nodeCounts = nodeCounts 
 86      self.numInputNodes = nodeCounts[0] 
 87      self.numOutputNodes = nodeCounts[-1] 
 88      self.numHiddenLayers = len(nodeCounts)-2 
 89      self.numInHidden = [None]*self.numHiddenLayers 
 90      for i in xrange(self.numHiddenLayers): 
 91        self.numInHidden[i] = nodeCounts[i+1] 
 92   
 93      numNodes = sum(self.nodeCounts) 
 94      self.nodeList = [None]*(numNodes) 
 95      for i in xrange(numNodes): 
 96        self.nodeList[i] = NetNode.NetNode(i,self.nodeList, 
 97                                           actFunc=actFunc, 
 98                                           actFuncParms=actFuncParms) 
 99   
100      self.layerIndices = [None]*len(nodeCounts) 
101      start = 0 
102      for i in xrange(len(nodeCounts)): 
103        end = start + nodeCounts[i] 
104        self.layerIndices[i] = range(start,end) 
105        start = end 
106         
107 -  def GetInputNodeList(self): 
108      """ returns a list of input node indices 
109      """ 
110      return self.layerIndices[0] 
111 -  def GetOutputNodeList(self): 
112      """ returns a list of output node indices 
113      """ 
114      return self.layerIndices[-1] 
115 -  def GetHiddenLayerNodeList(self,which): 
116      """ returns a list of hidden nodes in the specified layer 
117      """ 
118      return self.layerIndices[which+1] 
119     
120 -  def GetNumNodes(self): 
121      """ returns the total number of nodes 
122      """ 
123      return sum(self.nodeCounts) 
124   
125 -  def GetNumHidden(self): 
126      """ returns the number of hidden layers 
127      """ 
128      return self.numHiddenLayers 
129   
130 -  def GetNode(self,which): 
131      """ returns a particular node 
132      """ 
133      return self.nodeList[which] 
134 -  def GetAllNodes(self): 
135      """ returns a list of all nodes 
136      """ 
137      return self.nodeList 
138     
139 -  def ClassifyExample(self,example,appendExamples=0): 
140      """ classifies a given example and returns the results of the output layer. 
141   
142        **Arguments** 
143   
144          - example: the example to be classified 
145   
146        **NOTE:** 
147   
148          if the output layer is only one element long, 
149          a scalar (not a list) will be returned.  This is why a lot of the other 
150          network code claims to only support single valued outputs. 
151   
152      """ 
153      if len(example) > self.numInputNodes: 
154        if len(example)-self.numInputNodes > self.numOutputNodes: 
155          example = example[1:-self.numOutputNodes] 
156        else: 
157          example = example[:-self.numOutputNodes] 
158      assert len(example) == self.numInputNodes 
159      totNumNodes = sum(self.nodeCounts) 
160      results = numpy.zeros(totNumNodes,numpy.float64) 
161      for i in xrange(self.numInputNodes): 
162        results[i] = example[i] 
163      for i in xrange(self.numInputNodes,totNumNodes): 
164        self.nodeList[i].Eval(results) 
165      self.lastResults = results[:] 
166      if self.numOutputNodes == 1: 
167        return results[-1] 
168      else: 
169        return results 
170     
171 -  def GetLastOutputs(self): 
172      """ returns the complete list of output layer values from the last time this node classified anything""" 
173      return self.lastResults 
174     
175 -  def __str__(self): 
176      """ provides a string representation of the network """ 
177      outStr = 'Network:\n' 
178      for i in xrange(len(self.nodeList)): 
179        outStr = outStr + '\tnode(% 3d):\n'%i 
180        outStr = outStr + '\t\tinputs:  %s\n'%(str(self.nodeList[i].GetInputs())) 
181        outStr = outStr + '\t\tweights: %s\n'%(str(self.nodeList[i].GetWeights())) 
182   
183      outStr = outStr + 'Total Number of Connections: % 4d'%self.nConnections 
184      return outStr 
185       
186 -  def __init__(self,nodeCounts,nodeConnections=None, 
187                 actFunc=ActFuncs.Sigmoid,actFuncParms=(), 
188                 weightBounds=1): 
189      """ Constructor 
190   
191        This constructs and initializes the network based upon the specified 
192        node counts. 
193   
194        A fully connected network with random weights is constructed. 
195   
196        **Arguments** 
197   
198          - nodeCounts: a list containing the number of nodes to be in each layer. 
199             the ordering is: 
200              (nInput,nHidden1,nHidden2, ... , nHiddenN, nOutput) 
201   
202          - nodeConnections: I don't know why this is here, but it's optional.  ;-) 
203   
204          - actFunc: the activation function to be used here.  Must support the API 
205              of _ActFuncs.ActFunc_. 
206   
207          - actFuncParms: a tuple of extra arguments to be passed to the activation function 
208              constructor. 
209   
210          - weightBounds:  a float which provides the boundary on the random initial weights 
211   
212   
213     
214      """ 
215      self.ConstructNodes(nodeCounts,actFunc,actFuncParms) 
216      self.FullyConnectNodes() 
217      self.ConstructRandomWeights(minWeight=-weightBounds,maxWeight=weightBounds) 
218      self.lastResults = [] 
219   
220  if __name__ == '__main__': 
221   
222    print '[2,2,2]' 
223    net = Network([2,2,2]) 
224    print net 
225   
226    print '[2,4,1]' 
227    net = Network([2,4,1]) 
228    print net 
229   
230    print '[2,2]' 
231    net = Network([2,2]) 
232    print net 
233    input = [1,0] 
234    res = net.ClassifyExample(input) 
235    print input,'->',res 
236    input = [0,1] 
237    res = net.ClassifyExample(input) 
238    print input,'->',res 
239    input = [.5,.5] 
240    res = net.ClassifyExample(input) 
241    print input,'->',res 
242