org.soulspace.math.gradient — org.soulspace.clj/math.core 0.9.0

direction clj/s

Returns the direction of the vector w (which is w normalized to length 1)

directional-variance clj/s

(directional-variance m w)

Calculates the variance of the matrix m in the direction of w.

directional-variance-gradient clj/s

(directional-variance-gradient m w)

Calculates the contribution of the matrix m to the gradient of the variance in direction w.

directional-variance-gradient-i clj/s

(directional-variance-gradient-i x-i w)

Calculates the contribution of row x-i to the gradient of the variance in direction w.

directional-variance-i clj/s

(directional-variance-i x-i w)

Calculates the variance of the row x-i in the direction of w.

estimate-gradient clj/s

(estimate-gradient f v)

(estimate-gradient f v h)

Estimates the gradient of f at v.

maximize-batch clj/s

(maximize-batch target-fn gradient-fn theta-0)

(maximize-batch target-fn gradient-fn theta-0 tolerance)

Calculates the theta that minimizes the target function by gradient descent.

maximize-stochastic clj/s

(maximize-stochastic target-fn gradient-fn x y theta-0)

(maximize-stochastic target-fn gradient-fn x y theta-0 alpha-0)

Calculates .

minimize-batch clj/s

(minimize-batch target-fn gradient-fn theta-0)

(minimize-batch target-fn gradient-fn theta-0 tolerance)

Calculates .

minimize-stochastic clj/s

(minimize-stochastic target-fn gradient-fn x y theta-0)

(minimize-stochastic target-fn gradient-fn x y theta-0 alpha-0)

Calculates .

minimize-sum-of-squares clj/s

(minimize-sum-of-squares start-v)

Calculates the minimization of the sum of squares.

negated-all-fn clj/s

(negated-all-fn f)

Returns a function that calculates the negated values (multiplied with -1) of f for functions returning a sequence of values.

negated-fn clj/s

(negated-fn f)

Returns a function that calculates the negated value (multiplied with -1) of f.

partial-difference-quotient clj/s

(partial-difference-quotient f v i h)

Calculates the i-th partial difference quotient of f at v.

safe-apply clj/s

(safe-apply f & args)

Safely applies f so that it returns infinity whenever f returns an error.

safe-fn clj/s

(safe-fn f)

Returns a function that is the same as f except that it returns infinity whenever f returns an error.

step clj/s

(step v direction step-size)

Returns the vector v moved step-size in direction.

step-sizes clj/s

sum-of-squares-gradient clj/s

(sum-of-squares-gradient v)

Calculates the gradient of the sum of squares of v.