def crossvalidation(ds,
trainer,
model,
nfolds=10,
verbose=1,
shuffle=False,
error=eval.errorrate,
seed=None):
n = ds.n
ds.shuffle(seed=seed)
folds = ds.split(nfolds)
err = []
for foldidx in range(nfolds):
if verbose > 1:
print("--------------------")
print("Fold-%d" % (foldidx + 1))
print("--------------------")
lm = copy.deepcopy(model)
t1 = time()
dtest = folds[foldidx]
trainidxs = range(nfolds)
del trainidxs[foldidx]
dtrain = MemoryDataset.merge(folds[trainidxs])
trainer.train(lm, dtrain, verbose=(verbose - 1), shuffle=shuffle)
e = error(lm, dtest)
if verbose > 0:
fid = ("%d" % (foldidx + 1)).ljust(5)
print("%s %s" % (fid, ("%.2f" % e).rjust(5)))
err.append(e)
if verbose > 1:
print "Total time for fold-%d: %f" % (foldidx + 1, time() - t1)
return np.array(err)
def crossvalidation(ds, trainer, model, nfolds=10, verbose=1, shuffle=False,
error=eval.errorrate, seed=None):
n = ds.n
ds.shuffle(seed = seed)
folds = ds.split(nfolds)
err = []
for foldidx in range(nfolds):
if verbose > 1:
print("--------------------")
print("Fold-%d" % (foldidx+1))
print("--------------------")
lm = copy.deepcopy(model)
t1 = time()
dtest = folds[foldidx]
trainidxs = range(nfolds)
del trainidxs[foldidx]
dtrain = MemoryDataset.merge(folds[trainidxs])
trainer.train(lm, dtrain,
verbose = (verbose-1),
shuffle = shuffle)
e = error(lm,dtest)
if verbose > 0:
fid = ("%d" % (foldidx+1)).ljust(5)
print("%s %s" % (fid , ("%.2f"%e).rjust(5)))
err.append(e)
if verbose > 1:
print "Total time for fold-%d: %f" % (foldidx+1, time()-t1)
return np.array(err)