Half Space Trees HST classifier .Net C# implemetation

Classifier.cs

    public class ClassifierSettings
    {
        public int WindowSize { get; set; } = 250;

        public int Estimators { get; set; } = 25;

        public int MaxDepth { get; set; } = 15;

        public int Features { get; set; } = 0;

        public double MinLimit { get; set; } = 0d;

        public double MaxLimit { get; set; } = 1d;
    }

    public class Classifier
    {
        private readonly Node[] _nodes;

        private readonly Random _random = new Random(42);

        private readonly ClassifierSettings _settings;

        private int _count;

        public Classifier(ClassifierSettings settings)
        {
            _settings = settings;

            _nodes = new Node[_settings.Estimators];
        }

        private Node Build(Limit[] limits, int depth)
        {
            if (depth == _settings.MaxDepth)
            {
                return new Node
                {
                    Type = NodeType.External,
                    Depth = depth
                };
            }

            var feature = _random.NextChoice(
                Enumerable.Range(0, _settings.Features)
                    .ToDictionary(i => i, i => limits[i].Max - limits[i].Min));

            var limit = limits[feature];

            var max = limit.Max;
            var min = limit.Min;

            const double padding = 0.15d;

            var value = _random.NextDouble(
                limit.Min + padding * (limit.Max - limit.Min),
                limit.Max - padding * (limit.Max - limit.Min));

            limit.Max = value;
            limit.Min = min;

            var left = Build(limits, depth + 1);

            limit.Max = max;
            limit.Min = value;

            var right = Build(limits, depth + 1);

            return new Node
            {
                Left = left,
                Right = right,
                Type = NodeType.Internal,
                Depth = depth,
                Feature = feature,
                Value = value
            };
        }

        private void Initialise()
        {
            var features = _settings.Features;

            for (var i = 0; i < _settings.Estimators; i++)
            {
                var limits = Enumerable.Range(0, features)
                    .Select(_ =>
                        new Limit
                        {
                            Min = _settings.MinLimit,
                            Max = _settings.MaxLimit
                        })
                    .ToArray();

                _nodes[i] = Build(limits, 0);
            }
        }

        public void Fit(IData data)
        {
            if (_count == 0)
            {
                Initialise();
            }

            foreach (var root in _nodes)
            {
                foreach (var node in Path(root, n =>
                    data.Features[n.Feature] < n.Value))
                {
                    if (_count < _settings.WindowSize)
                    {
                        node.RightMass++;
                    }

                    node.LeftMass++;
                }
            }

            _count++;

            if (_count % _settings.WindowSize == 0)
            {
                foreach (var root in _nodes)
                {
                    Iterate(root, n =>
                    {
                        n.RightMass = n.LeftMass;
                        n.LeftMass = 0;
                    });
                }
            }
        }

        private IEnumerable<Node> Path(Node node, Func<Node, bool> evaluator)
        {
            while (node != null)
            {
                yield return node;

                node = evaluator.Invoke(node) ? node.Left : node.Right;
            }
        }

        private void Iterate(Node node, Action<Node> action)
        {
            action.Invoke(node);

            if (node.Type == NodeType.External)
            {
                return;
            }

            Iterate(node.Left, action);
            Iterate(node.Right, action);
        }

        public double Score(IData data)
        {
            var size = Math.Min(_settings.WindowSize, _count);

            var max = _settings.Estimators * size *
                      (Math.Pow(2, _settings.MaxDepth + 1) - 1);

            var limit = 0.1d * _settings.WindowSize;

            var score = 0d;

            foreach (var root in _nodes)
            {
                foreach (var node in Path(root, n =>
                    data.Features[n.Feature] < n.Value))
                {
                    score += node.RightMass * Math.Pow(2, node.Depth);

                    if (node.RightMass < limit)
                    {
                        break;
                    }
                }
            }

            return 1 - score / max;
        }

        private class Limit
        {
            public double Min { get; set; }

            public double Max { get; set; }
        }

        private class Node
        {
            public NodeType Type { get; set; }

            public Node Left { get; set; }

            public Node Right { get; set; }

            public int LeftMass { get; set; }

            public int RightMass { get; set; }

            public int Feature { get; set; }

            public double Value { get; set; }

            public int Depth { get; set; }
        }
    }