score_one method modifies anomaly.LocalOutlierFactor internal state unintentionally?

[MarekWadinger]

Versions

river version: development
Python version: Python 3.10.1
Operating system: macOS Sonoma 14.0 (23A344)

Describe the bug

Hello 👋

I found out that score_one method of anomaly.LocalOutlierFactor changes internal state of the detector.

If this is not intentional I'd like to assist in resolving this issue. I tried to trace back the error and found out that the problem is connected to expand_objects function, which creates only a references to original variables. Following modification to lines 349-369 in anomaly.LocalOutlierFactor HERE, helped me to preserve the state over score_one calls.

        (
            nm,
            x_list_copy,
            neighborhoods,
            rev_neighborhoods,
            k_dist,
            reach_dist,
            dist_dict,
            local_reach,
            lof,
        ) = expand_objects(
            self.x_scores,
            x_list_copy,
            self.neighborhoods,
            self.rev_neighborhoods,
            self.k_dist,
            self.reach_dist,
            self.dist_dict,
            self.local_reach,
            self.lof,
        )

to

        import copy
        (
            nm,
            x_list_copy,
            neighborhoods,
            rev_neighborhoods,
            k_dist,
            reach_dist,
            dist_dict,
            local_reach,
            lof,
        ) = expand_objects(
            self.x_scores,
            x_list_copy,
            self.neighborhoods.copy(),
            self.rev_neighborhoods.copy(),
            self.k_dist.copy(),
            copy.deepcopy(self.reach_dist),
            copy.deepcopy(self.dist_dict),
            self.local_reach.copy(),
            self.lof.copy(),
        )

Please, let me know if that makes sense, and I'd be happy to elaborate on any related issues or comments.

Thank you 🙏

Steps/code to reproduce

from river import anomaly

lof = anomaly.LocalOutlierFactor()

X = [{"a": 1, "b": 1}, {"a": 1, "b": 1}]
for x in X:
    lof.learn_one(x)

print(
    lof.x_list,
    lof.neighborhoods,
    lof.rev_neighborhoods,
    lof.k_dist,
    lof.reach_dist,
    lof.dist_dict,
    lof.local_reach,
    lof.lof, sep="\n"
    )

lof.score_one({"a": 0.5, "b": 1})

print(
    lof.x_list,
    lof.neighborhoods,
    lof.rev_neighborhoods,
    lof.k_dist,
    lof.reach_dist,
    lof.dist_dict,
    lof.local_reach,
    lof.lof, sep="\n"
    )

returns

[{'a': 1, 'b': 1}, {'a': 1, 'b': 1}]
{0: [1], 1: [0]}
{0: [1], 1: [0]}
{0: 0.0, 1: 0.0}
{0: {1: 0.0}, 1: {0: 0.0}}
{0: {1: 0.0}, 1: {0: 0.0}}
{0: 0, 1: 0}
{0: 0, 1: 0}

[{'a': 1, 'b': 1}, {'a': 1, 'b': 1}]
{0: [1, 2], 1: [0, 2], 2: [0, 1]}
{0: [1, 2], 1: [0, 2], 2: [0, 1]}
{0: 0.5, 1: 0.5, 2: 0.5}
{0: {1: 0.5, 2: 0.5}, 1: {0: 0.5, 2: 0.5}, 2: {0: 0.5, 1: 0.5}}
{0: {1: 0.0, 2: 0.5}, 1: {0: 0.0, 2: 0.5}, 2: {0: 0.5, 1: 0.5}}
{0: 2.0, 1: 2.0, 2: 2.0}
{0: 1.0, 1: 1.0, 2: 1.0}

[MaxHalford]

Good job spotting this! Indeed score_one should be stateless.

[smastelini]

Can we close this?

[MaxHalford]

Yep go for it :)

[smastelini]

Closed via #1330.