http://www.massapi.com/method/org/jblas/DoubleMatrix.mmul.html
mul是矩阵和元素相乘。
mmul是矩阵和矩阵相乘。
源码：
public DoubleMatrix mmuli(double v) {
return this.mmuli(v, this);
}
public DoubleMatrix mmul(double v) {
return this.mmuli(v, new DoubleMatrix(this.rows, this.columns));
}
public DoubleMatrix mmuli(DoubleMatrix other, DoubleMatrix result) {
if (other.isScalar()) {
return this.muli(other.scalar(), result);
} else if (this.isScalar()) {
return other.muli(this.scalar(), result);
} else {
this.assertMultipliesWith(other);
if (result.rows != this.rows || result.columns != other.columns) {
if (result == this || result == other) {
throw new SizeException("Cannot resize result matrix because it is used in-place.");
}
result.resize(this.rows, other.columns);
}
if (result != this && result != other) {
if (other.columns == 1) {
SimpleBlas.gemv(1.0D, this, other, 0.0D, result);
} else {
SimpleBlas.gemm(1.0D, this, other, 0.0D, result);
}
} else {
DoubleMatrix temp = new DoubleMatrix(result.rows, result.columns);
if (other.columns == 1) {
SimpleBlas.gemv(1.0D, this, other, 0.0D, temp);
} else {
SimpleBlas.gemm(1.0D, this, other, 0.0D, temp);
}
SimpleBlas.copy(temp, result);
}
return result;
}
}
@Override
public void update(Boolean label, double[] features) {
if (this.weights == null || this.variance == null) {
this.init(features.length);
}
DoubleMatrix weightsVector = new DoubleMatrix(1, this.weights.length, this.weights);
DoubleMatrix varianceMatrix = new DoubleMatrix(this.variance);
DoubleMatrix featuresVector = new DoubleMatrix(1, features.length, features);
double margin = weightsVector.dot(featuresVector);
double labelAsDouble = label ? 1.0 : -1.0;
if (margin * labelAsDouble < 1) {
double confidence = featuresVector.dot(featuresVector.mmul(varianceMatrix));
double beta = 1 / (confidence + this.r);
double alpha = Math.max(0, beta * (1 - labelAsDouble * margin));
DoubleMatrix delta = featuresVector.mmul(varianceMatrix).mul(alpha * labelAsDouble);
boolean zeroVector = MathUtil.isZeros(delta);
if (!zeroVector) {
this.weights = weightsVector.add(delta).toArray();
// Matrix library needed!
this.variance = varianceMatrix.sub(featuresVector.mmul(varianceMatrix).transpose().mmul(featuresVector).mul(beta).mmul(varianceMatrix)).toArray2();
}
}
}jblas is a fast linear algebra library for Java. jblas is based on BLAS and LAPACK, the de-facto industry standard for matrix computations, and uses state-of-the-art implementations like ATLAS for all its computational routines, making jBLAS very fast.
主页地址：http://jblas.org/
该库的API文档地址：http://jblas.org/javadoc/index.html
在Java工程中我们只要 引入jar包文件就可以了。
下面是其小册子：
jblas - Fast matrix computations for Java
import org.jblas.*; import static org.jblas.DoubleMatrix.*; import static org.jblas.MatrixFunctions.*;