neighbor 0.1.0 → 0.1.1
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +4 -0
- data/README.md +14 -5
- data/lib/neighbor.rb +14 -1
- data/lib/neighbor/model.rb +9 -3
- data/lib/neighbor/vector.rb +1 -5
- data/lib/neighbor/version.rb +1 -1
- metadata +3 -3
checksums.yaml
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
---
|
2
2
|
SHA256:
|
3
|
-
metadata.gz:
|
4
|
-
data.tar.gz:
|
3
|
+
metadata.gz: 01df817b036ee9b4d0c54ddadfb3be7fb31b36bb81504321b8dfd3d7a1ba83a4
|
4
|
+
data.tar.gz: f7a1757914f2e1226bc3b95fe5e0df9837a530315d6e627062df565e331b1708
|
5
5
|
SHA512:
|
6
|
-
metadata.gz:
|
7
|
-
data.tar.gz:
|
6
|
+
metadata.gz: 8381834f92092cb13d2c8898b588f0758e99e48f84e3b506aab15acf54fb0b3f72449734ca68db10de466eecde4a7b277cceca72192240ce21567125ad0af52f
|
7
|
+
data.tar.gz: 79ddd953bf0c134d73c92c61a8051524c024e2e1c95710e346a191a8a7c098f55562f81263752ea2e2ccaef9ced8e376dbf6431a4e745931d256ee2ca854d309
|
data/CHANGELOG.md
CHANGED
data/README.md
CHANGED
@@ -20,6 +20,8 @@ rails generate neighbor:install
|
|
20
20
|
rails db:migrate
|
21
21
|
```
|
22
22
|
|
23
|
+
This enables the [cube extension](https://www.postgresql.org/docs/current/cube.html) in Postgres
|
24
|
+
|
23
25
|
## Getting Started
|
24
26
|
|
25
27
|
Create a migration
|
@@ -48,13 +50,13 @@ item.update(neighbor_vector: [1.0, 1.2, 0.5])
|
|
48
50
|
|
49
51
|
> With cosine distance (the default), vectors are normalized before being stored
|
50
52
|
|
51
|
-
|
53
|
+
Get the nearest neighbors
|
52
54
|
|
53
55
|
```ruby
|
54
56
|
item.nearest_neighbors.first(5)
|
55
57
|
```
|
56
58
|
|
57
|
-
##
|
59
|
+
## Distance
|
58
60
|
|
59
61
|
Specify the distance metric
|
60
62
|
|
@@ -64,19 +66,24 @@ class Item < ApplicationRecord
|
|
64
66
|
end
|
65
67
|
```
|
66
68
|
|
67
|
-
Supported
|
69
|
+
Supported values are:
|
68
70
|
|
69
71
|
- `cosine` (default)
|
70
72
|
- `euclidean`
|
71
73
|
- `taxicab`
|
72
74
|
- `chebyshev`
|
73
75
|
|
74
|
-
|
76
|
+
Records returned from `nearest_neighbors` will have a `neighbor_distance` attribute
|
75
77
|
|
76
78
|
```ruby
|
77
|
-
|
79
|
+
nearest_item = item.nearest_neighbors.first
|
80
|
+
nearest_item.neighbor_distance
|
78
81
|
```
|
79
82
|
|
83
|
+
## Dimensions
|
84
|
+
|
85
|
+
By default, Postgres limits the `cube` data type to 100 dimensions. See the [Postgres docs](https://www.postgresql.org/docs/current/cube.html) for how to increase this.
|
86
|
+
|
80
87
|
## Example
|
81
88
|
|
82
89
|
You can use Neighbor for online item recommendations with [Disco](https://github.com/ankane/disco). We’ll use MovieLens data for this example.
|
@@ -119,6 +126,8 @@ movie = Movie.find_by(name: "Star Wars (1977)")
|
|
119
126
|
movie.nearest_neighbors.first(5).map(&:name)
|
120
127
|
```
|
121
128
|
|
129
|
+
[Complete code](examples/disco.rb)
|
130
|
+
|
122
131
|
## History
|
123
132
|
|
124
133
|
View the [changelog](https://github.com/ankane/neighbor/blob/master/CHANGELOG.md)
|
data/lib/neighbor.rb
CHANGED
@@ -21,7 +21,20 @@ ActiveSupport.on_load(:active_record) do
|
|
21
21
|
|
22
22
|
extend Neighbor::Model
|
23
23
|
|
24
|
-
# prevent unknown OID warning
|
25
24
|
require "active_record/connection_adapters/postgresql_adapter"
|
25
|
+
|
26
|
+
# ensure schema can be dumped
|
27
|
+
ActiveRecord::ConnectionAdapters::PostgreSQLAdapter::NATIVE_DATABASE_TYPES[:cube] = {name: "cube"}
|
28
|
+
|
29
|
+
# ensure schema can be loaded
|
30
|
+
if ActiveRecord::VERSION::MAJOR >= 6
|
31
|
+
ActiveRecord::ConnectionAdapters::TableDefinition.send(:define_column_methods, :cube)
|
32
|
+
else
|
33
|
+
ActiveRecord::ConnectionAdapters::TableDefinition.define_method :cube do |*args, **options|
|
34
|
+
args.each { |name| column(name, :cube, options) }
|
35
|
+
end
|
36
|
+
end
|
37
|
+
|
38
|
+
# prevent unknown OID warning
|
26
39
|
ActiveRecord::ConnectionAdapters::PostgreSQLAdapter.prepend(Neighbor::RegisterCubeType)
|
27
40
|
end
|
data/lib/neighbor/model.rb
CHANGED
@@ -4,8 +4,12 @@ module Neighbor
|
|
4
4
|
distance = distance.to_s
|
5
5
|
raise ArgumentError, "Invalid distance: #{distance}" unless %w(cosine euclidean taxicab chebyshev).include?(distance)
|
6
6
|
|
7
|
+
# TODO make configurable
|
8
|
+
# likely use argument
|
9
|
+
attribute_name = :neighbor_vector
|
10
|
+
|
7
11
|
class_eval do
|
8
|
-
attribute
|
12
|
+
attribute attribute_name, Neighbor::Vector.new(dimensions: dimensions, distance: distance)
|
9
13
|
|
10
14
|
define_method :nearest_neighbors do
|
11
15
|
return self.class.none if neighbor_vector.nil?
|
@@ -20,9 +24,11 @@ module Neighbor
|
|
20
24
|
"<->"
|
21
25
|
end
|
22
26
|
|
27
|
+
quoted_attribute = "#{self.class.connection.quote_table_name(self.class.table_name)}.#{self.class.connection.quote_column_name(attribute_name)}"
|
28
|
+
|
23
29
|
# important! neighbor_vector should already be typecast
|
24
30
|
# but use to_f as extra safeguard against SQL injection
|
25
|
-
order = "
|
31
|
+
order = "#{quoted_attribute} #{operator} cube(array[#{neighbor_vector.map(&:to_f).join(", ")}])"
|
26
32
|
|
27
33
|
# https://stats.stackexchange.com/questions/146221/is-cosine-similarity-identical-to-l2-normalized-euclidean-distance
|
28
34
|
# with normalized vectors:
|
@@ -35,7 +41,7 @@ module Neighbor
|
|
35
41
|
self.class
|
36
42
|
.select(*self.class.column_names, "#{neighbor_distance} AS neighbor_distance")
|
37
43
|
.where.not(self.class.primary_key => send(self.class.primary_key))
|
38
|
-
.where.not(
|
44
|
+
.where.not(attribute_name => nil)
|
39
45
|
.order(Arel.sql(order))
|
40
46
|
end
|
41
47
|
end
|
data/lib/neighbor/vector.rb
CHANGED
@@ -13,11 +13,7 @@ module Neighbor
|
|
13
13
|
raise Error, "Expected #{@dimensions} dimensions, not #{value.size}" unless value.size == @dimensions
|
14
14
|
|
15
15
|
if @distance == "cosine"
|
16
|
-
norm =
|
17
|
-
value.each do |v|
|
18
|
-
norm += v * v
|
19
|
-
end
|
20
|
-
norm = Math.sqrt(norm)
|
16
|
+
norm = Math.sqrt(value.sum { |v| v * v })
|
21
17
|
value.map { |v| v / norm }
|
22
18
|
else
|
23
19
|
value
|
data/lib/neighbor/version.rb
CHANGED
metadata
CHANGED
@@ -1,14 +1,14 @@
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
2
2
|
name: neighbor
|
3
3
|
version: !ruby/object:Gem::Version
|
4
|
-
version: 0.1.
|
4
|
+
version: 0.1.1
|
5
5
|
platform: ruby
|
6
6
|
authors:
|
7
7
|
- Andrew Kane
|
8
8
|
autorequire:
|
9
9
|
bindir: bin
|
10
10
|
cert_chain: []
|
11
|
-
date: 2021-02-
|
11
|
+
date: 2021-02-17 00:00:00.000000000 Z
|
12
12
|
dependencies:
|
13
13
|
- !ruby/object:Gem::Dependency
|
14
14
|
name: activerecord
|
@@ -58,7 +58,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
|
|
58
58
|
- !ruby/object:Gem::Version
|
59
59
|
version: '0'
|
60
60
|
requirements: []
|
61
|
-
rubygems_version: 3.
|
61
|
+
rubygems_version: 3.2.3
|
62
62
|
signing_key:
|
63
63
|
specification_version: 4
|
64
64
|
summary: Nearest neighbor search for Rails and Postgres
|