ExploringMLNet
A set of small projects to explore ML.NET library. This repo is a part of my Futurice's DotNeat talk on ML.NET.
Problem 1: Regression
Say, we have a huge list of sets of 5 numbers that participate in some expression and its result:
A,B,C,D,E,Result
25,76,60,75,34,14791
25,86,71,98,17,15194
21,29,20,15,49,4215
84,50,40,30,7,18446
...
The idea of this scenario is to train a model using this data, evaluate it, and guess a result based on a provided sample set. For example, feeding numbers like 19, 24, 64, 12, 10 should give us something close to 2609.
Keep in mind that the machine doesn't know the actual formula of that expression. (Although if we do some math or check CalculateExpression method we would know that the expression is (a + b)ˆ2 + (c + d) * e.)
To run the program and see sample based predictions run the following commands:
cd ./ExploringMLNet.Expression
dotnet restore
dotnet runProblem 2: Binary Classification
In this scenario we have a lot of user reviews from Rotten Tomatoes. Each review has either a positive or negative rating. Then again, we build and evaluate a model, but this time round we need binary classification for that. What we want though is to determine a rating based on a provided text.
To run the program and see sample based predictions run the following commands:
cd ./ExploringMLNet.RottenTomatoes
dotnet restore
dotnet run