class NeuralNet extends Predictor with Error
The NeuralNet
class supports basic 3-layer (input, hidden and output) Neural
Networks. Given several input and output vectors (training data), fit the weights
connecting the layers, so that for a new input vector 'zi', the net can predict
the output vector 'zo' ('zh' is the intermediate value at the hidden layer), i.e.,
zi --> zh = f (w * zi) --> zo = g (v * zh)
Note, w_0 and v_0 are treated as biases, so zi_0 and zh_0 must be 1.0.
- Alphabetic
- By Inheritance
- NeuralNet
- Error
- Predictor
- AnyRef
- Any
- Hide All
- Show All
- Public
- All
Instance Constructors
-
new
NeuralNet(x: MatrixD, y: MatrixD, h: Int, eta: Double = 1.0)
- x
the input matrix (training data consisting of m input vectors)
- y
the output matrix (training data consisting of m output vectors)
- h
the number of neurons in the hidden layer
- eta
the learning/convergence rate
Value Members
-
final
def
!=(arg0: Any): Boolean
- Definition Classes
- AnyRef → Any
-
final
def
##(): Int
- Definition Classes
- AnyRef → Any
-
final
def
==(arg0: Any): Boolean
- Definition Classes
- AnyRef → Any
-
final
def
asInstanceOf[T0]: T0
- Definition Classes
- Any
-
val
b: VectoD
- Attributes
- protected
- Definition Classes
- Predictor
-
def
backProp(): Unit
Use back propagation to adjust the weight matrices 'w' and 'v' to make the predictions more accurate.
Use back propagation to adjust the weight matrices 'w' and 'v' to make the predictions more accurate. First adjust the 'v' weights (hidden to output layer) and then move back to adjust the 'w' weights (input to hidden layer).
- See also
http://ufldl.stanford.edu/wiki/index.php/Backpropagation_Algorithm
http://www4.rgu.ac.uk/files/chapter3%20-%20bp.pdf
-
def
clone(): AnyRef
- Attributes
- protected[java.lang]
- Definition Classes
- AnyRef
- Annotations
- @throws( ... )
-
def
coefficient: VectoD
Return the vector of coefficient/parameter values.
Return the vector of coefficient/parameter values.
- Definition Classes
- Predictor
-
def
diagnose(yy: VectoD): Unit
Compute diagostics for the predictor.
Compute diagostics for the predictor. Override to add more diagostics. Note, for 'rmse', 'sse' is divided by the number of instances 'm' rather than degrees of freedom.
- yy
the response vector
- Definition Classes
- Predictor
- See also
en.wikipedia.org/wiki/Mean_squared_error
-
val
e: VectoD
- Attributes
- protected
- Definition Classes
- Predictor
-
final
def
eq(arg0: AnyRef): Boolean
- Definition Classes
- AnyRef
-
def
equals(arg0: Any): Boolean
- Definition Classes
- AnyRef → Any
-
def
finalize(): Unit
- Attributes
- protected[java.lang]
- Definition Classes
- AnyRef
- Annotations
- @throws( classOf[java.lang.Throwable] )
-
def
fit: VectorD
Return the quality of fit.
-
def
fit2: (MatrixD, MatrixD)
Return the fit (weight matrix 'w' and weight matrix 'v').
Return the fit (weight matrix 'w' and weight matrix 'v'). FIX - make compatible with
Predictor
-
def
fitLabels: Seq[String]
Return the labels for the fit.
-
final
def
flaw(method: String, message: String): Unit
- Definition Classes
- Error
-
final
def
getClass(): Class[_]
- Definition Classes
- AnyRef → Any
-
def
hashCode(): Int
- Definition Classes
- AnyRef → Any
-
final
def
isInstanceOf[T0]: Boolean
- Definition Classes
- Any
-
val
mae: Double
- Attributes
- protected
- Definition Classes
- Predictor
-
def
minimizeError(xx: MatrixD, yy: MatrixD, ww: MatrixD): Double
Minimize the error in the prediction by adjusting the weight vector 'w'.
Minimize the error in the prediction by adjusting the weight vector 'w'. The error 'eo' is simply the difference between the target value 'yi' and the predicted value 'zo'. Minimize 1/2 of the dot product of error with itself using gradient-descent.
- xx
the effective input layer training data/matrix
- yy
the effective output layer training data/matrix
- ww
the weights between these two layers
-
final
def
ne(arg0: AnyRef): Boolean
- Definition Classes
- AnyRef
-
final
def
notify(): Unit
- Definition Classes
- AnyRef
-
final
def
notifyAll(): Unit
- Definition Classes
- AnyRef
-
def
predict(zi: VectoD): Double
Given an input vector 'zi', predict the output/response scalar 'zo(0)'.
-
def
predict(z: VectoI): Double
Given a new discrete data vector z, predict the y-value of f(z).
Given a new discrete data vector z, predict the y-value of f(z).
- z
the vector to use for prediction
- Definition Classes
- Predictor
-
def
predictAll(zi: MatriD): MatrixD
Given several input vectors 'zi', predict the output/response vectors 'zo'.
Given several input vectors 'zi', predict the output/response vectors 'zo'.
- zi
the new input vectors (stored as rows in a matrix)
-
def
predictAll(zi: VectoD): VectoD
Given an input vector 'zi', predict the output/response vector 'zo'.
Given an input vector 'zi', predict the output/response vector 'zo'. For the hidden to output layer bias, prepend the hidden values with a one (_11).
- zi
the new input vector
-
val
rSq: Double
- Attributes
- protected
- Definition Classes
- Predictor
-
def
residual: VectoD
Return the vector of residuals/errors.
Return the vector of residuals/errors.
- Definition Classes
- Predictor
-
val
rmse: Double
- Attributes
- protected
- Definition Classes
- Predictor
-
def
setWeights(i: Int = 0): Unit
Set the initial weight matrices 'w' and 'v' randomly with a value in (0, 1) before training.
Set the initial weight matrices 'w' and 'v' randomly with a value in (0, 1) before training.
- i
the random number stream to use
-
def
setWeights(w0: MatrixD, v0: MatrixD): Unit
Set the initial weight matrices 'w and 'v' manually before training.
Set the initial weight matrices 'w and 'v' manually before training.
- w0
the initial weights for w
- v0
the initial weights for v
-
val
sse: Double
- Attributes
- protected
- Definition Classes
- Predictor
-
val
ssr: Double
- Attributes
- protected
- Definition Classes
- Predictor
-
val
sst: Double
- Attributes
- protected
- Definition Classes
- Predictor
-
final
def
synchronized[T0](arg0: ⇒ T0): T0
- Definition Classes
- AnyRef
-
def
toString(): String
- Definition Classes
- AnyRef → Any
-
def
train(): Unit
Given training data 'x' and 'y', fit the weight matrices 'w' and 'v'.
-
def
train(yy: VectoD): Unit
Given training data 'x' and 'y', fit the weight matrices 'w' and 'v'.
-
final
def
wait(): Unit
- Definition Classes
- AnyRef
- Annotations
- @throws( ... )
-
final
def
wait(arg0: Long, arg1: Int): Unit
- Definition Classes
- AnyRef
- Annotations
- @throws( ... )
-
final
def
wait(arg0: Long): Unit
- Definition Classes
- AnyRef
- Annotations
- @throws( ... )