rmontanana/BayesNet
rmontanana/BayesNet
1
0
Fork 0
Code Issues 2 Pull Requests Packages Projects Releases 3 Wiki Activity

Add weights to BayesMetrics

Browse Source
This commit is contained in:
Ricardo Montañana Gómez 2023-08-13 12:56:06 +02:00
parent af0419c9da
commit f26ea1f0ac
Signed by: rmontanana
GPG Key ID: 46064262FD9A7ADE
5 changed files with 28 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
src/BayesNet/BayesMetrics.cc
View File

@ -32,7 +32,7 @@ namespace bayesnet {
}
return result;
}
torch::Tensor Metrics::conditionalEdge()
torch::Tensor Metrics::conditionalEdge(const torch::Tensor& weights)
{
auto result = vector<double>();
auto source = vector<string>(features);
@ -52,7 +52,7 @@ namespace bayesnet {
auto mask = samples.index({ -1, "..." }) == value;
auto first_dataset = samples.index({ index_first, mask });
auto second_dataset = samples.index({ index_second, mask });
auto mi = mutualInformation(first_dataset, second_dataset);
auto mi = mutualInformation(first_dataset, second_dataset, weights);
auto pb = margin[value].item<float>();
accumulated += pb * mi;
}
@ -70,15 +70,16 @@ namespace bayesnet {
return matrix;
}
// To use in Python
vector<float> Metrics::conditionalEdgeWeights()
vector<float> Metrics::conditionalEdgeWeights(vector<float>& weights_)
{
auto matrix = conditionalEdge();
const torch::Tensor weights = torch::tensor(weights_);
auto matrix = conditionalEdge(weights);
std::vector<float> v(matrix.data_ptr<float>(), matrix.data_ptr<float>() + matrix.numel());
return v;
}
double Metrics::entropy(const torch::Tensor& feature)
double Metrics::entropy(const torch::Tensor& feature, const torch::Tensor& weights)
{
torch::Tensor counts = feature.bincount();
torch::Tensor counts = feature.bincount(weights);
int totalWeight = counts.sum().item<int>();
torch::Tensor probs = counts.to(torch::kFloat) / totalWeight;
torch::Tensor logProbs = torch::log(probs);
@ -86,15 +87,15 @@ namespace bayesnet {
return entropy.nansum().item<double>();
}
// H(Y|X) = sum_{x in X} p(x) H(Y|X=x)
double Metrics::conditionalEntropy(const torch::Tensor& firstFeature, const torch::Tensor& secondFeature)
double Metrics::conditionalEntropy(const torch::Tensor& firstFeature, const torch::Tensor& secondFeature, const torch::Tensor& weights)
{
int numSamples = firstFeature.sizes()[0];
torch::Tensor featureCounts = secondFeature.bincount();
torch::Tensor featureCounts = secondFeature.bincount(weights);
unordered_map<int, unordered_map<int, double>> jointCounts;
double totalWeight = 0;
for (auto i = 0; i < numSamples; i++) {
jointCounts[secondFeature[i].item<int>()][firstFeature[i].item<int>()] += 1;
totalWeight += 1;
totalWeight += weights[i].item<float>();
}
if (totalWeight == 0)
return 0;
@ -115,9 +116,9 @@ namespace bayesnet {
return entropyValue;
}
// I(X;Y) = H(Y) - H(Y|X)
double Metrics::mutualInformation(const torch::Tensor& firstFeature, const torch::Tensor& secondFeature)
double Metrics::mutualInformation(const torch::Tensor& firstFeature, const torch::Tensor& secondFeature, const torch::Tensor& weights)
{
return entropy(firstFeature) - conditionalEntropy(firstFeature, secondFeature);
return entropy(firstFeature, weights) - conditionalEntropy(firstFeature, secondFeature, weights);
}
/*
Compute the maximum spanning tree considering the weights as distances

16
src/BayesNet/BayesMetrics.h
View File

@ -12,16 +12,16 @@ namespace bayesnet {
vector<string> features;
string className;
int classNumStates = 0;
double entropy(const Tensor& feature, const Tensor& weights);
double conditionalEntropy(const Tensor& firstFeature, const Tensor& secondFeature, const Tensor& weights);
vector<pair<string, string>> doCombinations(const vector<string>&);
public:
Metrics() = default;
Metrics(const Tensor&, const vector<string>&, const string&, const int);
Metrics(const vector<vector<int>>&, const vector<int>&, const vector<string>&, const string&, const int);
double entropy(const Tensor&);
double conditionalEntropy(const Tensor&, const Tensor&);
double mutualInformation(const Tensor&, const Tensor&);
vector<float> conditionalEdgeWeights(); // To use in Python
Tensor conditionalEdge();
vector<pair<string, string>> doCombinations(const vector<string>&);
Metrics(const torch::Tensor& samples, const vector<string>& features, const string& className, const int classNumStates);
Metrics(const vector<vector<int>>& vsamples, const vector<int>& labels, const vector<string>& features, const string& className, const int classNumStates);
double mutualInformation(const Tensor& firstFeature, const Tensor& secondFeature, const Tensor& weights);
vector<float> conditionalEdgeWeights(vector<float>& weights); // To use in Python
Tensor conditionalEdge(const torch::Tensor& weights);
vector<pair<int, int>> maximumSpanningTree(const vector<string>& features, const Tensor& weights, const int root);
};
}

5
src/BayesNet/Classifier.h
View File

@ -14,13 +14,14 @@ namespace bayesnet {
Classifier& build(vector<string>& features, string className, map<string, vector<int>>& states);
protected:
bool fitted;
Network model;
int m, n; // m: number of samples, n: number of features
Tensor dataset; // (n+1)xm tensor
Network model;
Metrics metrics;
vector<string> features;
string className;
map<string, vector<int>> states;
Tensor dataset; // (n+1)xm tensor
Tensor weights;
void checkFitParameters();
virtual void buildModel() = 0;
void trainModel() override;

4
src/BayesNet/KDB.cc
View File

@ -32,10 +32,10 @@ namespace bayesnet {
vector <float> mi;
for (auto i = 0; i < features.size(); i++) {
Tensor firstFeature = dataset.index({ i, "..." });
mi.push_back(metrics.mutualInformation(firstFeature, y));
mi.push_back(metrics.mutualInformation(firstFeature, y, weights));
}
// 2. Compute class conditional mutual information I(Xi;XjIC), f or each
auto conditionalEdgeWeights = metrics.conditionalEdge();
auto conditionalEdgeWeights = metrics.conditionalEdge(weights);
// 3. Let the used variable list, S, be empty.
vector<int> S;
// 4. Let the DAG network being constructed, BN, begin with a single

6
src/BayesNet/TAN.cc
View File

@ -15,15 +15,15 @@ namespace bayesnet {
Tensor class_dataset = dataset.index({ -1, "..." });
for (int i = 0; i < static_cast<int>(features.size()); ++i) {
Tensor feature_dataset = dataset.index({ i, "..." });
auto mi_value = metrics.mutualInformation(class_dataset, feature_dataset);
auto mi_value = metrics.mutualInformation(class_dataset, feature_dataset, weights);
mi.push_back({ i, mi_value });
}
sort(mi.begin(), mi.end(), [](const auto& left, const auto& right) {return left.second < right.second;});
auto root = mi[mi.size() - 1].first;
// 2. Compute mutual information between each feature and the class
auto weights = metrics.conditionalEdge();
auto weights_matrix = metrics.conditionalEdge(weights);
// 3. Compute the maximum spanning tree
auto mst = metrics.maximumSpanningTree(features, weights, root);
auto mst = metrics.maximumSpanningTree(features, weights_matrix, root);
// 4. Add edges from the maximum spanning tree to the model
for (auto i = 0; i < mst.size(); ++i) {
auto [from, to] = mst[i];