Skip to content

Commit

Permalink
Addition Local Outlier Factor functions
Browse files Browse the repository at this point in the history 
Add
* `local_outlier_factor`
* `local_outlier_factor_table`
* `local_outlier_factor_projection`
* `local_outlier_factor_projection_table`
  • Loading branch information
@rasmushenningsson
rasmushenningsson committed Jun 20, 2024
1 parent e2e0658 commit a6e8788
Show file tree
Hide file tree
Showing 2 changed files with 108 additions and 18 deletions.
3 changes: 3 additions & 0 deletions CHANGELOG.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

## [Unreleased]

### Added
* `local_outlier_factor`, `local_outlier_factor_table`, `local_outlier_factor_projection`, `local_outlier_factor_projection_table`

## [0.4.0] - 2024-06-20

### Breaking
Expand Down
123 changes: 105 additions & 18 deletions src/local_outlier_factor.jl
View file
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,15 @@ function _local_outlier_factor(adj::AbstractSparseMatrix, lrdX::AbstractVector,
end


function _local_outlier_factor(data::DataMatrix, full::DataMatrix; k)
adj = knn_adjacency_matrix(data; k, make_symmetric=false)
dists = adjacency_distances(adj, full)
kdists = vec(maximum(dists.matrix; dims=1))
lrd = local_reachability_density(dists.matrix, kdists);
_local_outlier_factor(dists.matrix, lrd)
end





Expand Down Expand Up @@ -86,17 +95,65 @@ julia> reduced = svd(normalized; nsv=10)
julia> local_outlier_factor!(reduced, normalized; k=10)
```
See also: [`local_outlier_factor_projection!`](@ref)
See also: [`local_outlier_factor`](@ref), [`local_outlier_factor_table`](@ref), [`local_outlier_factor_projection!`](@ref)
"""
function local_outlier_factor!(data::DataMatrix, full::DataMatrix; k=10, col="LOF")
adj = knn_adjacency_matrix(data; k, make_symmetric=false)
dists = adjacency_distances(adj, full)
kdists = vec(maximum(dists.matrix; dims=1))
lrd = local_reachability_density(dists.matrix, kdists);
data.obs[:,col] = _local_outlier_factor(dists.matrix, lrd)
data.obs[:,col] = _local_outlier_factor(data, full; k)
data
end


"""
local_outlier_factor(data::DataMatrix, full::DataMatrix; k=10, col="LOF", matrix=:keep, var=:copy)
See `local_outlier_factor!` for documentation. This version does not modify `data` in place.
See also: [`local_outlier_factor!`](@ref), [`local_outlier_factor_table`](@ref), [`local_outlier_factor_projection`](@ref)
"""
function local_outlier_factor(data::DataMatrix, full::DataMatrix; k=10, col="LOF", matrix=:keep, var=:copy, kwargs...)
obs = copy(data.obs)
obs[:,col] = _local_outlier_factor(data, full; k)
X = matrix == :keep ? data.matrix : copy(data.matrix)
var = _update_annot(data.var, var, size(matrix,1))
DataMatrix(X, var, obs, data.models; kwargs...)
end


"""
local_outlier_factor_table(data::DataMatrix, full::DataMatrix; k=10, col="LOF")
See `local_outlier_factor!` for documentation. This returns a DataFrame with observation IDs and a column `col` with LOF values.
See also: [`local_outlier_factor!`](@ref), [`local_outlier_factor`](@ref), [`local_outlier_factor_projection_table`](@ref)
"""
function local_outlier_factor_table(data::DataMatrix, full::DataMatrix; k=10, col="LOF")
df = select(data.obs, 1)
df[:,col] = _local_outlier_factor(data, full; k)
df
end





function _local_outlier_factor_projection(data::DataMatrix, full::DataMatrix,
base::DataMatrix, base_full::DataMatrix;
k=10)
adj_X = knn_adjacency_matrix(base; k, make_symmetric=false)
adj_XY = knn_adjacency_matrix(base, data; k)

dists_X = adjacency_distances(adj_X, base_full)
dists_XY = adjacency_distances(adj_XY, base_full, full)

kdists = vec(maximum(dists_X.matrix; dims=1))

lrd_X = local_reachability_density(dists_X.matrix, kdists);
lrd_XY = local_reachability_density(dists_XY.matrix, kdists);

_local_outlier_factor(dists_XY.matrix, lrd_X, lrd_XY)
end


"""
local_outlier_factor_projection!(data::DataMatrix, full::DataMatrix, base::DataMatrix, base_full::DataMatrix; k=10, col="LOF_projection")
Expand Down Expand Up @@ -152,23 +209,53 @@ julia> reduced = project(normalized, base_reduced);
julia> local_outlier_factor!(reduced, normalized, base_reduced, base_normalized; k=10)
```
See also: [`local_outlier_factor!`](@ref)
See also: [`local_outlier_factor_projection`](@ref), [`local_outlier_factor_projection_table`](@ref), [`local_outlier_factor!`](@ref)
"""
function local_outlier_factor_projection!(data::DataMatrix, full::DataMatrix,
base::DataMatrix, base_full::DataMatrix;
k=10,
col="LOF_projection")
adj_X = knn_adjacency_matrix(base; k, make_symmetric=false)
adj_XY = knn_adjacency_matrix(base, data; k)
data.obs[:,col] = _local_outlier_factor_projection(data, full, base, base_full; k)
data
end

dists_X = adjacency_distances(adj_X, base_full)
dists_XY = adjacency_distances(adj_XY, base_full, full)

kdists = vec(maximum(dists_X.matrix; dims=1))

lrd_X = local_reachability_density(dists_X.matrix, kdists);
lrd_XY = local_reachability_density(dists_XY.matrix, kdists);

data.obs[:,col] = _local_outlier_factor(dists_XY.matrix, lrd_X, lrd_XY)
data
end

"""
local_outlier_factor_projection(data::DataMatrix, full::DataMatrix, base::DataMatrix, base_full::DataMatrix; k=10, col="LOF_projection", matrix=:keep, var=:copy)
See `local_outlier_factor_projection!` for documentation. This version does not modify `data` in place.
See also: [`local_outlier_factor_projection!`](@ref), [`local_outlier_factor_projection_table`](@ref), [`local_outlier_factor`](@ref)
"""
function local_outlier_factor_projection(data::DataMatrix, full::DataMatrix,
base::DataMatrix, base_full::DataMatrix;
k=10,
col="LOF_projection",
matrix = :keep,
var = :copy,
kwargs...)
obs = copy(data.obs)
obs[:,col] = _local_outlier_factor_projection(data, full, base, base_full; k)
X = matrix == :keep ? data.matrix : copy(data.matrix)
var = _update_annot(data.var, var, size(matrix,1))
DataMatrix(X, var, obs, data.models; kwargs...)
end


"""
local_outlier_factor_projection_table(data::DataMatrix, full::DataMatrix, base::DataMatrix, base_full::DataMatrix; k=10, col="LOF_projection", matrix=:keep, var=:copy)
See `local_outlier_factor_projection!` for documentation. This returns a DataFrame with observation IDs and a column `col` with LOF values for the projection.
See also: [`local_outlier_factor_projection!`](@ref), [`local_outlier_factor_projection`](@ref), [`local_outlier_factor_table`](@ref)
"""
function local_outlier_factor_projection_table(data::DataMatrix, full::DataMatrix,
base::DataMatrix, base_full::DataMatrix;
k=10,
col="LOF_projection")
df = select(base.obs, 1)
df[:,col] = _local_outlier_factor_projection(data, full, base, base_full; k)
df
end

0 comments on commit a6e8788

Please sign in to comment.