radumg / dynai Goto Github PK

Add

• SVM
• Decision tree
• Logistic

Note : this should only be done after #2 and #7 are complete

Add

• SimpleLinearRegression

In linear regression, the model specification is that the dependent variable, y is a linear combination of the parameters (but need not be linear in the independent variables).

• MultipleLinearRegression

• MultivariateLinearRegression

• LogisticRegression

It is a generalized linear model used for binomial regression.
Like many forms of regression analysis, it makes use of several predictor variables that may be either numerical or categorical.
For example, the probability that a person has a heart attack within a specified time period might be predicted from knowledge of the person's age, sex and body mass index.
Logistic regression is used extensively in the medical and social sciences as well as marketing applications such as prediction of a customer's propensity to purchase a product or cease a subscription.

Note : this should only be done after #2 and #7 are complete

The Machine.LoadModel node returns null now - no errors provided though.
It should also be renamed to Machine.Load 😛 maybe.

This enhancement would decouple a Machine from any particular algorithm, instead allowing us to specify the algorithm as an input.

Requires
#2 to be completed
abstracting Machine to use Accord.NET interfaces

it was available before but when refactoring in #11, it was left out.

using JSON as implemented in #3

The Machine.Learn node returns null instead of the trained Machine.

Note, the node does work at least once as the Machine.Predict node runs, so the underlying algorithm is trained.

under the Release configuration at least, it should build a Dynamo-compatible package folder at repo top level

add support for at least 1 clustering algorithm, such as

• K-means

see Accord documentation for samples :
https://github.com/accord-net/framework/wiki/Clustering

Should be able to

- serialise
- deserialise

to a specified type from any

- JSON string
- JSON file

Description

I'm using lots of dynamic types to marshall input data from a Machine.Predict() to its associated Algorithm.Predict() method.
The input ends up malformed in Algorithm.Predict()

Why do it this way

The Predict method is defined on the IAlgorithm interface but since each algorithm has a different return type, it declares its inputs/outputs as dynamic.

Limitations

That wouldn't be a problem in C# land, but Dynamo doesn't like an input port with dynamic as type and unwraps lists to feed individual values - madenning ! Solved with [Arbitrary...] tag in the end so Dynamo leaves lists alone.

Dynamo also can't resolve nodes that take interfaces as inputs to ports, despite feeding it classes that implement those interfaces.

Also can't use generics like `IAlgorithm<Tinput, Toutput> as they're not known and the algorithm is stored on the machine.

What to do about it

• potentially find a way to use generics - i wasted several hours on it already but Dynamo won't play ball
• use delegates instead of generics
• not return the value from Predict directly, but you'd still have the data types conversion issues.