Montag, 23. September 2013

Finished GSoC project Expresso

Finished GSoC project Expresso

Finished GSoC project Expresso


GSoC ends today and I can announce the 0.2.0 version of the expresso library. It is build on top of core.logic and core.matrix and provides symbolic manipulation of algebraic expressions.

What's there?

  1. An api/dsl for manipulation of algebraic expressions which doesn't get in your way. Expresso's expressions are just clojure s-expressions and can be manipulated with rich set of clojure sequence functions
  2. useful manipulations for mathematical expressions: simplify, multiply-out, differentiate, …
  3. An equation solver which is capable of solving a single equation and multiple equations for unknowns.
  4. An optimizer which transforms a mathematical expression to a semantically equivalent but performanter one
  5. An expression compiler to compile an expression to an efficient clojure function
  6. A semantic rule based translator on top of which many of expresso's features are implemented

The code is fully documented and I wrote a tutorial and showcase of expresso, the expresso-tutorial.

GSoC has been a really fun and valuable time for me. I learned a lot. Of course I will continue developing expresso! Expresso and core.matrix are the first steps in the direction of a full computer algebra system for clojure. I hope that it will help clojure to be an attractive choice for scientific computing projects in the future.

Showcase: Here are two examples of expresso's facility to manipulate mathematical expressions. They can be found and are explained in the expresso-tutorial 1.

  1. solving word problems:
(solve 'blue
  (ex (= pencils (+ green white blue red)))
  (ex (= (/ pencils 10) green))
  (ex (= (/ pencils 2) white))
  (ex (= (/ pencils 4) blue))
  (ex (= red 45))) ;=> #{{blue 75N}}
  1. Analyzing roots and extremata of functions. This code shows how easy one can implement tasks involving symbolic manipulation with expresso:
(defn roots
  "returns the set of roots of the expression in regard to var"
  [var expr]
  (solve var (ex (= ~expr 0))))


(defn extremata 
  "gets the extrema of the expression in regard to var. Returns a map with the
   keys :maxima and :minima"
  [var expr]
  (let [d1 (differentiate [var] expr)
 d2 (differentiate [var] d1)
 candidates (roots var d1)]
    (if (seq candidates)
      (let [extremata
     (->> candidates
   (map (fn [candidate] [candidate (evaluate d2 {var candidate})]))
   (remove #(== 0 (second %)))
   (group-by #(< 0 (second %))))]
 {:maxima (map first (get extremata false))
  :minima (map first (get extremata true))}))))


(defn analyse-function 
  "returns a map with the :roots, the :maxima and the :minima of the expression
   in regard to var"
  [var expr]
  (assoc (extremata var expr)
    :roots (roots var expr)))

(analyse-function 'x (ex (- (** x 4) (** x 2))))
;=> {:roots #{0 -1 1},
;;   :maxima (0),
;;   :minima (0.7071067811865476 -0.7071067811865476)}

Ideas/feedbacks etc are greatly appreciated! Enjoy, Maik