added linear regression #4
Conversation
pom.xml
Outdated
| @@ -103,6 +103,21 @@ | |||
| <scope>test</scope> | |||
| </dependency> | |||
|
|
|||
| <dependency> | |||
There was a problem hiding this comment.
let's not do this but use embedded testing for the library, it's faster than going through server + driver
src/main/java/regression/LR.java
Outdated
| } | ||
|
|
||
| @Procedure(value = "regression.linear.storeModel", mode = Mode.WRITE) | ||
| public Stream<ModelResult> storeModel(@Name("model") String model) { |
There was a problem hiding this comment.
Perhaps it's easier to just have a function instead that turns the model into a byte[] so the user can decide themselves what to do with it? i.e. where to store it.
There was a problem hiding this comment.
Yes I agree. This was my original idea, but I ran into issues because I was trying to return byte[] as part of a stream (this is silly because I should have just written a user function). I will fix this, and also convert predict into a function that returns a single value rather than a stream.
src/main/java/regression/LR.java
Outdated
| } | ||
|
|
||
| @Procedure(value = "regression.linear.createFromStorage", mode = Mode.READ) | ||
| public Stream<ModelResult> createFromStorage(@Name("model") String model) { |
There was a problem hiding this comment.
Just consume byte[] and turn into Model
| } | ||
|
|
||
| //Serializes the object into a byte array for storage | ||
| public static byte[] convertToBytes(Object object) throws IOException { |
There was a problem hiding this comment.
For future proof it would probably be better at some point to serialize using some explicit data from the model.
For now it's ok
There was a problem hiding this comment.
Yea I agree, it's just difficult because you cannot create a new SimpleRegression object using the information it stores (because is doesn't actually store individual data points). Maybe I should consider using a different library?
src/main/java/regression/LR.java
Outdated
| } | ||
|
|
||
| @Procedure(value = "regression.linear.removeModel", mode = Mode.READ) | ||
| public Stream<ModelResult> removeModel(@Name("model") String model) { |
src/main/java/regression/LR.java
Outdated
| @Procedure(value = "regression.linear.create", mode = Mode.READ) | ||
| public Stream<ModelResult> create(@Name("model") String model) { | ||
| return Stream.of((new LRModel(model)).asResult()); | ||
| } |
There was a problem hiding this comment.
Add an info/details function or procedure that just returns data on the model by name.
| try{ byte[] serializedR = LR.convertToBytes(R); | ||
| Map<String, Object> parameters = new HashMap<>(); | ||
| parameters.put("name", name); | ||
| ResourceIterator<Entity> n = db.execute("MERGE (n:LRModel {name:$name}) RETURN n", parameters).columnAs("n"); |
There was a problem hiding this comment.
Change to embedded API or just return byte[]
src/test/java/regression/LRTest.java
Outdated
| assertThat(actualPrediction, equalTo(expectedPrediction)); | ||
| } | ||
|
|
||
| session.run("CALL regression.linear.storeModel('work and progress')"); |
src/test/java/regression/LRTest.java
Outdated
| "r.time) YIELD prediction SET r.predictedProgress = prediction"); | ||
|
|
||
| //replicate the creation and updates of the model | ||
| R.removeData(1.0, 1.345); |
There was a problem hiding this comment.
are there some in-between asserts missing?
src/test/java/regression/LRTest.java
Outdated
|
|
||
| assertThat( actualPrediction, equalTo( expectedPrediction ) ); | ||
| } | ||
|
|
The functionality of adding/removing data with SimpleRegression is a bit different than your implementation in ML.java (ex. no need for a separate "train" step) so I created my own package "regression" for these procedures. I also had to add some dependencies to pom.xml. Tests are located in LRTest.java.