|
List of Main Keywords
- ACTIVATION function of neurons.
- TRAINING Indicates method that will be
used to train neural network weights.
-
Seven methods are currently available.
-
ANALYSIS Indicates method that will be used
to analyze data.
-
Two methods, i.e. to fit data and to prune input variables, are currently
available.
-
ASNN Indicate to use Associative Neural Network algorithm for data analysis.
- AVOID Indicates input variables that will not
be considered in the analysis.
- The variables indicated in this keyword will be deleted from the input
data set before any analysis.
-
CLASSIFICATION Performs classification on classes and not the regression
analysis.
- CORRELATION Level of correlation to eliminate
highly correlated input variables.
- The second of each two highly correlated variables (with R^2>value
of this keyword) will be deleted before any analysis.
-
DISTANCE function for the ASNN algorithm.
- ENSEMBLE Indicates the number of neural networks
that will be averaged.
- Analysis of several networks estimates variability of calculated results.
-
INCLUDE Indicates input variables that will
be considered in the analysis.
- The default value is to include all variables.
-
INPUTS.
 Indicates
the number of input (independent) variables for neural network calculations.
- This parameter should be identified for all calculations.
-
ITERATIONS Maximal number of iterations for
neural network calculations.
- The network calculations will be stopped if maximal number of iterations
was exceeded.
-
KNN Maximal number of nearest neighbors for the ASNN algorithm.
- LIMIT Indicates RMSE error to stop neural network
training.
- A neural network calculation is stopped if RMSE error for the learning
set is less than value of this keyword.
-
LOO Performs validation of the approach using the
leave-one-out method.
- A data entry is removed and its value predicted after neural network
calculations. This procedure is repeated for all data entries.
-
MISSED If several output values are analyzed simultaneously, the missed
values should be substituted with the this value.
- MODELS Indicates if neural network weight should
be saved/loaded.
- This keyword is useful to save neural network models for further analysis
of new data.
-
NAMES.Indicates
if the first column of data table contains names.
- Names should not contain spaces and tabs.
-
NONZERO Indicate minimal number of non constant elements in a column.
- NEURONS Determines hidden layers of neural network.
- The numbers of neurons in each hidden layer are indicated.
-
OUTLIERS absolute value to detect outliers in the output values. Mainly
used to calculate statistic with and without outliers.
- OUTPUTS.Indicates
number of outputs (dependent) variables for neural network calculations.
- This number determines number of neurons on the last layer of neural
network.
-
PARALLEL Uses several computers (if available) to speed-up calculations.
- PARTITION Indicates method to subdivide data
on training and validation subsets.
- The initial training data set is divided on training and validation
set according to this criterion.
-
PRINT Indicates options for output information.
- The corresponding checkbox should be checked to activate each option.
-
PRUNE Performs optimization of input variables
using pruning methods.
- The variables will be pruned according to their sensitivities calculated
by neural network ensemble.
-
RANGE Initial range for weight initialization.
- REVERSED.Indicates
that the reversed order of input and target values is used in the data
file.
- The standard order is X, independent variables, are followed by Y,
targets or dependent variables.
-
SEED Random seed number.
- This number is used in to start sequence of random numbers for neural
network calculations.
-
SHOW Number of iterations to print detailed information
about neural network calculations.
- This keyword can be useful to select number of iterations required
for neural network training..
-
TYPE Type of neural network.
- Two methods, namely feed-forward back propagation and cascade correlation,
are currently available.
-
UPDATE mode: either after each sample or using the all samples (batch).
- VALIDATION Indicates which data entries (rows)
will be used as validation data set.
- Explicitly indicate which data entries should be used as the validation
set.
-
Indicates parameter
that should be always verified.
See FAQ if you have questions. How to cite this applet? Are you looking for a new job in chemoinformatics?
|
|
