def crossValidate(data, meta, folds, topology, iterations, weights, graph=None):
"k-fold cross validation"
if folds <= 1:
raise Exception("Cross validation folds must be > 1")
averageError = 0.0
for counter, (training, validation) in enumerate(
crossValidateIndices(items=range(data.shape[0]), k=folds, randomize=True)
):
# setup training and validation matricies
train = data.ix[training].reset_index(drop=True)
validate = data.ix[validation].reset_index(drop=True)
trainingFeatures = train.drop(meta.categoricalLabelColumns, axis=1) # remove output columns
validationFeatures = validate.drop(meta.categoricalLabelColumns, axis=1) # remove output columns
trainingLabels = train[meta.categoricalLabelColumns] # use only output columns
validationLabels = validate[meta.categoricalLabelColumns] # use only output columns
# setup MLP and start training
li(
"Fold {2}/{3} - Training with {1}/{4} rows ({0} epochs)".format(
iterations, trainingFeatures.shape[0], counter + 1, folds, data.shape[0]
)
)
mlp = MLP()
mlp.trainingIterations = iterations
mlp.initalWeightsMultiplier = weights
mlp.features = (
trainingFeatures.values
) # convert from pandas dataframe to numpy arrays. they are faster for the computationally intensive training phase.
mlp.labels = trainingLabels.values
mlp.validationFeatures = validationFeatures # for validation, send in pandas dataframes.
mlp.validationLabels = validationLabels
mlp.meta = meta
mlp.topology = topology
if graph:
mlp.trackLearning = True
mlp.setupHiddenLayers()
mlp.train()
if graph:
li("Plotting Learning to file '{0}'".format(graph))
mlp.plotLearning(graph)
# validate model
li(
"Fold {0}/{1} - Testing with {2}/{3} rows".format(
counter + 1, folds, validationFeatures.shape[0], data.shape[0]
)
)
error = mlp.validateModel(printToScreen=True)
averageError += error
averageError = averageError / folds
li("Average error across all folds: {0}".format(averageError))
return averageError
def crossValidate(data, meta, folds, topology, iterations, weights, graph=None):
"k-fold cross validation"
if folds <= 1:
raise Exception("Cross validation folds must be > 1")
averageError = 0.0
for counter, (training, validation) in enumerate(crossValidateIndices(items=range(data.shape[0]), k=folds, randomize=True)):
# setup training and validation matricies
train = data.ix[training].reset_index(drop=True)
validate = data.ix[validation].reset_index(drop=True)
trainingFeatures = train.drop(meta.categoricalLabelColumns, axis=1) # remove output columns
validationFeatures = validate.drop(meta.categoricalLabelColumns, axis=1) # remove output columns
trainingLabels = train[meta.categoricalLabelColumns] # use only output columns
validationLabels = validate[meta.categoricalLabelColumns] # use only output columns
#setup MLP and start training
li("Fold {2}/{3} - Training with {1}/{4} rows ({0} epochs)".format(iterations, trainingFeatures.shape[0], counter + 1, folds, data.shape[0]))
mlp = MLP()
mlp.trainingIterations = iterations
mlp.initalWeightsMultiplier = weights
mlp.features = trainingFeatures.values # convert from pandas dataframe to numpy arrays. they are faster for the computationally intensive training phase.
mlp.labels = trainingLabels.values
mlp.validationFeatures = validationFeatures # for validation, send in pandas dataframes.
mlp.validationLabels = validationLabels
mlp.meta = meta
mlp.topology = topology
if graph:
mlp.trackLearning = True
mlp.setupHiddenLayers()
mlp.train()
if graph:
li("Plotting Learning to file '{0}'".format(graph))
mlp.plotLearning(graph)
#validate model
li("Fold {0}/{1} - Testing with {2}/{3} rows".format(counter + 1, folds, validationFeatures.shape[0], data.shape[0]))
error = mlp.validateModel(printToScreen=True)
averageError += error
averageError = averageError / folds
li("Average error across all folds: {0}".format(averageError))
return averageError
averageError += error
averageError = averageError / folds
li("Average error across all folds: {0}".format(averageError))
return averageError
if __name__ == "__main__":
if not args["data"]:
data = raw_input("Path to dataset: ")
else:
data = args["data"]
loader = DataLoader(data)
li("Loading Dataset")
data, meta = loader.load()
li(meta)
li("Dataset has Rows: {0}, Columns: {1}".format(len(data), len(meta.names())))
if args["command"] in ("info"):
sys.exit(0)
if not args["labels"]:
labels = raw_input("Which columns to use as labels? [{0}-{1}]: ".format(1, len(meta.names())))
else:
labels = args["labels"]
setLabels(data, meta, labels)
if len(meta.labelColumns) < 1:
raise Exception("Specify atleast 1 label column")
li("Label Columns: {0}".format(meta.labelColumns))
averageError += error
averageError = averageError / folds
li("Average error across all folds: {0}".format(averageError))
return averageError
if __name__ == '__main__':
if not args['data']:
data = raw_input('Path to dataset: ')
else:
data = args['data']
loader = DataLoader(data)
li("Loading Dataset")
data, meta = loader.load()
li(meta)
li("Dataset has Rows: {0}, Columns: {1}".format(len(data), len(meta.names())))
if args['command'] in ("info"):
sys.exit(0)
if not args['labels']:
labels = raw_input('Which columns to use as labels? [{0}-{1}]: '.format(1, len(meta.names())))
else:
labels = args['labels']
setLabels(data, meta, labels)
if len(meta.labelColumns) < 1:
raise Exception("Specify atleast 1 label column")
li("Label Columns: {0}".format(meta.labelColumns))
