public void Test_Cofi_Recommender()
{
var movies = new[] {
new { ID = 1, Name = "From Dawn til Dusk", Ratings = new int[] { 4, 0, 3, 4, 4, 5, 4 } },
new { ID = 2, Name = "The Hoarder", Ratings = new int[] { 0, 1, 5, 0, 0, 1, 0 } },
new { ID = 3, Name = "Cowboys and Cows", Ratings = new int[] { 2, 0, 3, 4, 4, 5, 4 } },
new { ID = 4, Name = "Small Fry Town", Ratings = new int[] { 0, 1, 2, 0, 1, 4, 0 } },
new { ID = 5, Name = "The White Knight", Ratings = new int[] { 4, 0, 3, 0, 4, 5, 3 } },
new { ID = 6, Name = "Love Me Tender", Ratings = new int[] { 0, 1, 1, 3, 0, 0, 0 } },
new { ID = 7, Name = "Total Groove", Ratings = new int[] { 0, 1, 5, 3, 0, 1, 0 } },
new { ID = 8, Name = "Action Chase", Ratings = new int[] { 0, 2, 3, 0, 0, 5, 4 } },
new { ID = 9, Name = "Underneath", Ratings = new int[] { 0, 4, 0, 0, 0, 5, 0 } },
new { ID = 10, Name = "Time Reinvented", Ratings = new int[] { 0, 0, 4, 3, 0, 3, 2 } },
};
// should predict (top 5): From Dawn til Dusk, The White Knight, Underneath, Cowboys and Cows, Action Chase
var descriptor = Descriptor.New("MOVIES")
.With("Ratings").AsEnumerable(7)
.Learn("ID").As(typeof(int));
var generator = new Recommendation.CofiRecommenderGenerator()
{
Ratings = new Math.Range(1, 5),
CollaborativeFeatures = 7,
Descriptor = descriptor,
LearningRate = 0.1,
Lambda = 1.0,
};
var model = Learner.Learn(movies, 1.0, 1, generator);
// get predictions for the movies of the first user
var predictions = ((Recommendation.CofiRecommenderModel)model.Model).Predict(0);
Assert.Equal(1d, predictions[0]);
// due to random initialisation one is favoured over the other at certain times
Assert.True(5d == predictions[1] || 9d == predictions[1]);
Assert.True(5d == predictions[2] || 9d == predictions[2]);
Assert.Equal(3d, predictions[3]);
Assert.Equal(8d, predictions[4]);
}
public void Test_Cofi_Recommender()
{
var movies = new[] {
new { ID = 1, Name = "From Dawn til Dusk", Ratings = new int[] { 4, 0, 3, 4, 4, 5, 4 } },
new { ID = 2, Name = "The Hoarder", Ratings = new int[] { 0, 1, 5, 0, 0, 1, 0 } },
new { ID = 3, Name = "Cowboys and Cows", Ratings = new int[] { 2, 0, 3, 4, 4, 5, 4 } },
new { ID = 4, Name = "Small Fry Town", Ratings = new int[] { 0, 1, 2, 0, 1, 4, 0 } },
new { ID = 5, Name = "The White Knight", Ratings = new int[] { 4, 0, 3, 0, 4, 5, 3 } },
new { ID = 6, Name = "Love Me Tender", Ratings = new int[] { 0, 1, 1, 3, 0, 0, 0 } },
new { ID = 7, Name = "Total Groove", Ratings = new int[] { 0, 1, 5, 3, 0, 1, 0 } },
new { ID = 8, Name = "Action Chase", Ratings = new int[] { 0, 2, 3, 0, 0, 5, 4 } },
new { ID = 9, Name = "Underneath", Ratings = new int[] { 0, 4, 0, 0, 0, 5, 0 } },
new { ID = 10, Name = "Time Reinvented", Ratings = new int[] { 0, 0, 4, 3, 0, 3, 2 } },
};
// should predict (top 5): From Dawn til Dusk, The White Knight, Underneath, Cowboys and Cows, Action Chase
var descriptor = Descriptor.New("MOVIES")
.With("Ratings").AsEnumerable(7)
.Learn("ID").As(typeof(int));
var generator = new Recommendation.CofiRecommenderGenerator()
{
Ratings = new Math.Range(1, 5),
CollaborativeFeatures = 7,
Descriptor = descriptor,
LearningRate = 0.1,
Lambda = 1.0,
};
var model = Learner.Learn(movies, 1.0, 1, generator);
// get predictions for the movies of the first user
var predictions = ((Recommendation.CofiRecommenderModel)model.Model).Predict(0);
Assert.Equal(1d, predictions[0]);
// due to random initialisation one is favoured over the other at certain times
Assert.True(5d == predictions[1] || 9d == predictions[1]);
Assert.True(5d == predictions[2] || 9d == predictions[2]);
Assert.Equal(3d, predictions[3]);
Assert.Equal(8d, predictions[4]);
}
