Isolation Forest feature max features < 1 #8318

itay955 · 2017-02-08T15:03:04Z

Regarding Isolation Forest released recently,
I found a bug on the decision_function method.
if you use max features less than 1 than the trees doesn't fit with the test.
it needs to first feature select the features for that specific tree.

instead of:
for i, tree in enumerate(self.estimators_):
leaves_index = tree.apply(X)
node_indicator = tree.decision_path(X)

it should be:
for i, tree in enumerate(self.estimators_):
X_i = X[:,self.estimators_features_[i]]
leaves_index = tree.apply(X_i)
node_indicator = tree.decision_path(X_i)

jnothman · 2017-02-08T20:18:36Z

Thanks. This should be fixed in master. See #5732

kuttire · 2017-02-21T23:52:52Z

Hi, This does not seem to be fixed. The module still gives an error if max features < 1

jnothman · 2017-02-22T00:40:32Z

In 0.18.1, or in master?

…

On 22 February 2017 at 10:52, kuttire ***@***.***> wrote: Hi, This does not seem to be fixed. The module still gives an error if max features < 1 — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#8318 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz6ykX4hrE05SgcJGX3Gr7QTuPgnIjks5re3jVgaJpZM4L67V_> .

jnothman closed this as completed Feb 8, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Isolation Forest feature max features < 1 #8318

Isolation Forest feature max features < 1 #8318

Isolation Forest feature max features < 1 #8318

Isolation Forest feature max features < 1 #8318

Comments