kumulant

com.eignex.kumulant/stat/regression/SoftmaxRegressionResult

SoftmaxRegressionResult

@Serializable

@SerialName(value = "SoftmaxRegressionResult")

data class SoftmaxRegressionResult(val featureSize: Int, val numClasses: Int, val weights: DenseMatrix, val biases: DenseVector, val totalWeights: Double, val step: Long, val crossEntropy: Double) : HasObservationCount(source)

Snapshot from SoftmaxRegressionStat: per-class linear-model parameters plus cumulative bookkeeping. The K-by-p weights matrix and length-K biases vector define the linear predictors eta[k] = biases[k] + weights[k] . x; the predicted class probability is the softmax over the K logits.

Constructors

SoftmaxRegressionResult

constructor(featureSize: Int, numClasses: Int, weights: DenseMatrix, biases: DenseVector, totalWeights: Double, step: Long, crossEntropy: Double)(source)

Properties

biases

val biases: DenseVector(source)

Per-class intercept; length numClasses.

crossEntropy

val crossEntropy: Double(source)

Accumulated weighted negative log-likelihood (cross-entropy) over the stream.

featureSize

val featureSize: Int(source)

Number of input features (columns of weights).

numClasses

val numClasses: Int(source)

Number of classes (rows of weights and length of biases).

step

val step: Long(source)

Number of update calls absorbed.

totalWeights

open override val totalWeights: Double(source)

Cumulative observation weight folded in.

weights

val weights: DenseMatrix(source)

K-by-p weight matrix; weights[k][i] is the coefficient on feature i for class k.

open val isEmpty: Boolean

True when no observation has been folded in, so every other field is a placeholder.

Functions

logit

fun logit(x: VectorView, k: Int): Double(source)

Linear predictor for class k: biases[k] + weights[k] . x.

predict

fun predict(x: VectorView): Int(source)

Argmax class index for x.

probabilities

fun probabilities(x: VectorView): DoubleArray(source)

Softmax probabilities across all classes for x; length numClasses.