d-ai37 776 research in algebraic ?rnipulation(u ... · d-ai37 776 research in algebraic...

14
D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U) MASSACHUSETTS I INST OF TECH CAMBRIDGE LAB FOR COMPUTER SCIENCE J MOSES 14 DEC 8B. AFOSR-TR-84 88 7 AFOSR-8-0250 UNCLASSIFIED F/G 12/1 N EEEEEEEEEEEEBIM

Upload: others

Post on 22-Sep-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U) MASSACHUSETTS IINST OF TECH CAMBRIDGE LAB FOR COMPUTER SCIENCEJ MOSES 14 DEC 8B. AFOSR-TR-84 88 7 AFOSR-8-0250

UNCLASSIFIED F/G 12/1 NEEEEEEEEEEEEBIM

Page 2: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

- -112.2-6.

1%.

% %*%i

.:.., I ".'.

~~~MICROCOPY RESOLUTION TEST CHART •""-

HATIONAI. BUJREAU OF STANDARDS-|963-A '

• . -. . ,, -. 0 )

- .. . - •

777..

% %

Page 3: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

AFOSR'.TR.

SECURITY CLASSIFICATION OF THIS PAGE

REPORT DOCUMENTATION PAGEis REPORT SECURITY CLASSIFICA1 ION lb. RESTRICTIVE MARKINGS

'-.'-TTNCLASS!FI!ED2& SECURITY CLASSIFICATION AUTHORITY 3. DISTRIBUTION/AVAILABILITY OF REPORT

____________________________________Approved for public release; distribution2b. DECLASSIF ICATION/DOWNGRADING SCHEDULE unlimited.

4. PERFORMING ORGANIZATION REPORT NUMBER(S) 5. MONITORING ORGANIZATION REPORT NUMBER(S)

AFOSR.TR, 76& NAME OF PERFORMING ORGANIZATION b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION

Massachusetts Institute of (If .pplicabiu) Air Force Office of Scientific ResearchTechnology ,,

- 6c. ADDRESS (Cit). Slate and ZIP Codu) 7b. ADDRESS (City. State and ZIP Code)

Laboratory for Computer Science, 545 Tech- Directorate of Mathematical & Information0 nology Square,; Cambridge MA 02139 Sciences, Bolling AFB DC 20332

"2Sm. NAME OF FUNDINGlSPONSORING lab. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER" 4. ORGANIZATION Of appUcable)

AFOSR NM AFOSR-80-0250&c. ADDRESS (City. State and ZIP Code) 10. SOURCE OF FUNDING NO.

PROGRAM PROJECT TASK WORK UNIT

Bolling AFB DC 20332 ELEMENT NO. NO. NO. NO.

11. TITLE lnceae Security Ciaaafication, I '*U % " "

RESEARCH IN ALGEBRAIC MANIPULATION; 3n up 9 61102F 2304 A412. PERSONAL AUTHOR(S)

Joel Moses13& TYPE OF REPORT 13b. TIME COVERED 114. DATE OF REPORT (Yr.. %lo. Day) 15. PAGE COUNTInterim FROMI/7/82 T030/6/83 14 DEC 83 9

1. SUPPLEMENTARY NOTATION

17. COSATI CODES 18. SUBJECT TERMS (Continue on reverse If neceuary aid identify b

FIELD GROUP SUB. GR.

19, ABSTRACT 'Continue on reverse if necesary and identify by block nurber"

-"-': L Professor Shunro Watanabe of Tapan has been visiting the group since March 1982. In the.'. past year he has been able, with the support of the group, to develop a major new pr'ogramLA- for solving in closed form a wide variety of ordinary differential equations.

""" The MACSYMA system uses two large packages for solving ODEs. The first package was writtenby Jeffrey Golden of the group. It solves equations by using a few well known techniques(e.g., separation of variables). It is quite weak in dealing with second order equations,

Lsuch as hypergeometric functions. The second package, due to Ed Lafferty of the MITRECorporation, incorporates many more techniques, but is still not a very general solver ofODEs.

Professor Watanabe's approach is to convert most second order ODEs into (CONTINUED)

20 DISTRIBUTION,AVAILABILITY OF ABSTRACT 21 ABSTRACT SECURITY CLASSIFICATION

'P_' -" UNCLASSIFIEOUNLIMITEO 7- SAME AS RPT :9 OTIC USERS - UIJCL,:J1I'

22. NAME OF RESPONSIBLE INDIVIDUAL 22L, TELEPHONE NUMBER 22L OFFICE SYMBOL.nude . rra Code o

Dr. Robert N. Buchal '0. 767- 4939EDT DIFORM 1473, 83 APR OF I JAN 73 I OBSOLEtE 1"' ''-* ---!'T,

1 7N'L !,l IC.ATIN 0I, THIS PAC.

V .

, ' ,q. , .,% , . .<_, , -. ,. ,, . ., .; ,. . . ,. - . .," ... .-. . , .... ,- . .. 1. .-. ,,

Page 4: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

UNCLASSIFI ED

SECURITY CLASSIFICATION OF THIS PAGE

ITEM #19, ABSTRACT, CONTINUED: instances of the so-called P-function. These functions,originally due to Riemann, have been extensively studied by the Japanese mathematicianFukuhara (also spelled Hukuhara). For the past year, Watanabe has implemented a packagethat transforms a large class of equations into P-functions. He has used Kaxe's tableas a test bed.

acssonFo

NTI GAA

DTC A

Unnnune

jutfiai

BNISS~~

Ditrb tin 100,**.I ' .. * ~ *

Page 5: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

* -- -N*. .* - * - * .. - . . . . . . ,.~ * • . . ,

AFOSR-TR.PROGRESS REPORT 1982/83

Professor Shunro Watanabe of Japan has been visiting our group since March,

1982. In the past year he has been able, with the support of our group, to

develop a major new program for solving in closed form a wide variety of

ordinary differential eqat .

The MACSYMA system uses two large packages for solving ODEs, The first

package was written by Jeffrey Golden of our group. It solves equations

- by using a few well known techniques (e.g., separation of variables). It

is quite weak in dealing with second order equations, such as hypergeometric

functions. The second package, due to Ed Lafferty of the MITRE Corporation,

incorporates many more techniques, but is still not a very general solver of

ODEs.

Professor Watanabe's approach is to convert most second order ODEs into

instances of the so-called P-function. These functions, originally due to

Riemann, have been extensively studied by the Japanese mathematician

Fukuhara (also spelled Hukuhara). For the past year, Watanabe has implemented

a package that transforms a large class of equations into P-functions. He has

used Kamke's table as a test bed. Several examples from Kamke using this

package, are given in Appendix I.

,.J

W ,. . -

- . .. . .

": ' ,: : " " " " " " ," " ' " ,r " ." " ' . " - ,. " . , , ' , - ., - ' - - . - . • • .. . . - . • . . .. . . . . . . .

Page 6: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

-P APPENDIX I

(D2) [DSK, SWATAN]

(C3) showtime:trueSTime: 5 msec.

(C4) loadfile(pmain,fasl);

PMAIN FASL DSK SWATAN being loaded Kamke examole 378aLoading doneTime= 411 msec.(D4) DONE

(C5) batch(exampl,2);

(C6) k378a:x,(x-l),(x+l)-2*'diff(y.x.2)+2*x-(x+l)*(x-3)*'diff(y,x)-2*(x-1).y=0;Time- 57 msec.

22 d Y dY

(D6) (X - 1) X (X + 1) --- + 2 (X - 3) X (X + 1) -- - 2 (X - 1) Y = 02 dX

dX

(C7) lode2(k378a);dY

2 (2 X - 6) --d Y dX Z Y

we solve --- + ---------------- = 02 2 3 2

dX X - 1 X + 2 X + X

SOLVE FASL DSK MACSYM being loadedLoading done

PREMAP FASL DSK SWATAN being loadedLoading done

U...j we use Y -

2(X + 1)

dU2- 2

2 U dX d Uthe result is ------------ -- 0

2 X - 1 2x - x dX

dY2 2 --dY dX 2Y

we solve --- ------- ------- 0

2 X- 1 2dX X - x

PEQPI FASL DSK SWATAN being loadedLoading done

PHYPGN FASL DSK SWATAN being loadedLoading done ....

_e ' J'

* " 'i " % " " % % " -"" "' ' "" . . . . " - •" " ". . . .. • . - ,

Page 7: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

.4o 4

" (D2) [DSK SWATAN] a e-%a-re 344

*, (C3) showtime:trueSTime- 4 msec.

* .:(C4) batch(exampl,7);

- - (C5) K344:X-4*'DIFF(Y,X.2) (EXP(2/X)-V-2).Y=O;Time- 65 msec.

24 d Y 2/X 2

(D5) X --- + (%E - V ) Y 02

dX

(C6) loadfile(pmain,fasl);

-,'. . PIAIN FASL DSK SWATAN being loadedLoading doneTime- 413 msec.

(D6) DONE

d (C7) lode2(k344);

*,2 2/X 2d Y (%E -V )Y

we solve -- ---------------- 0

2 4dX X

PIVTR FAS DSK SWATAN being loadedLoading done

4.

PTRINV FASL DSK SWATAN being loadedLoading done

we useT -

x".-. dY

2 2 --dY dT 2T 2

the result is --- + + (%E -V )Y 02 T

- dT

2 2--d Y dX 2 X 2

we solve --- + ---- + (%E -V )Y 0

2 XdX

PIVTR2 FASL DSK SWATAN being loadedLoading done

PTREXP FASL DSK SWATAN being loadedLoading done

PDVTR FASL DSK SWATAN being loaded

4..,v ,..,- ,• " " . . . . . . . . .. .. . .,9.'' - - , , , " ' " ,", "4 , -" . "-""" '" ,,. " • ,,--" ."." ." "•"•"-" "" - " - "

Page 8: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

the type is hypergeometricthe solution may be written by Riemann's P-functions as follows

,0 1 INF)

y:p[0 0 2 (x)

i,'..T [ 1 3 1

PNOPM FASL DSK SWATAN being loaded*: Loading done

we solve[0 1 INF J[ J

y:P [ 1 0 - 1 ) (x)[ )[0 3 -2

/ 2C (X- 1)

y- (K2 I -------- dX + KI) X'I J 2

] xcontinue? type y or n

, Y;

SIN FASL DSK MACSYM being loadedLoading done

SININT FASt DSK NACSYM being loadedLoading done

SCHATC FASL DSK MACSYM being loadedLoading doneTime= 21253 msec.

2

(D7) - 2 K2 X LOG(X) + K2 X + Ki X - K2

Time= 23672 msec.(D8) BATCH DONE

(C9) closefile(buffersave);

.5 e

.oI'5 ,- ,- ,.. . .

Page 9: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

Loading done

Loading done" SOLVE FASL DSK MACSYM being 1: ded

"= U

we use Yx

22X 2 dU

the result is U (%E - V ) + ... 0

2dX

2d Y 2 X 2

we solve --- + (%E - V ) Y - 02

dXx

we use T - %EdY

2 -- 2 2d Y dT (V - T ) Y

the result is --- +- -- - ----------- 02 T 2

dT TdY

2 -- 2 2dY dX (X -V)Y

we solve --- + ---------- 0

2 X 2dX X

PCONFL FASL DSK SWATAN being loadedLoading donethe equation is confluent type

PCNEQO FASL DSK SWATAN being loadedLoading donethe solution may be representable by Fukuhara's P-function.

[ INF INF 0 J[ V

y-P [ 2 (x)

I)% ]( XI - v ]( 2J

PCFPTM FASL DSK SWATAN being loadedLoading doney- Y (X)

-, B, ABS(V)Times 15604 msec.

q.(D7) Y MXB. ABS(V)

%- - . - -° - - J.. ..

Page 10: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

W .,

(D2) [DSK. SWATANI

(C3) showtime trueSTimez 5 msec.

(C4) batchiexamIl.12); Kamke exam.'!e .06

(C5) k406: 16 (X'3-1)2-' DIFF(YX.2)-2? XY=O;Time= 41 mrsec.

23 2 d Y

(D5) 16 (X - 1) + 27 X Y 0

dX..

(C6) k406t:48*x-2,(x-1)-2 'diff(y,x,2)+32,x*(x-I)^2*'diff(yx)+9*x~y=O;Time- 50 msec.

22 2 dY 2 dY

(D6) 48 (X 1) X --- + 32 (X - 1) X -- 9 X Y 02 dX

dX

"* (C7) loadfile(pmain.,fasl);

PMAIN FASL DSK SWATAN being loaded

5..Loading doneTime- 426 msec.

. (D7) DONE

(C8) lode2(k406t);dY

2 2--dY dX 3 Y

we solve --- - + - .......-------------- 0

2 3 X 3 2dX 1C X - 32 X + 16 X

SOLVE FASL DSK MACSYM being loaded

Loading done

PHYPGN FASL DSK SWATAN being loaded

Loading donethe type is hypergeometricthe solution may be written by Riemann's P-functions as follows

[ 0 1 INF )[ J( 13 1)

y-P 3 4 3 J (x)(1 ]

.(0 - 03[ 4 ]

PHGHP FASL DSK SWATAN being leaded

Loading done

Li z" -. ""

Page 11: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

W -jr -. 1 -0' CZ -w- 0!w

PALGS F-,SL CSK SAATAN being loadedLoadi ig done

we solve,0 1 INF ]

l[ )•.[1 1 1 )

1/4 - - ... ]y= (X - 1) P [ 3 2 12 ] (x)

-[ ]

0 0 -

(4)

PALG4 FASL DSK SWATAN being loaded

Loading done

P0H - '.SZL I.SK SWATAN being loadedLoading done

1/4 2 1/4 1/4KI T (2 SQRT(T + T + 1) + SQRT(3) (T + 1)) (X - 1)

y3 - - - - - - - - -- - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -3 1/12

(T - 1)

1/4 2 1/4 1/4K2 T (SQRT(3) (T + 1) - 2 SQRT(T + T + 1)) (X - 1)

-------------------------------------------------------------------------------------

3 1/12(T - 1)

" where t=x-(1/3)Time= 13449 msec.

1/4 2 1/4 1/4KI T (2 SQRT(T + T + 1) + SQRT(3) (T 1 1)) (X - 1)

(D8) ------------------------------------------------------------3 1/12

(T-1

1/4 2 1/4 1/4K2 7 (SQRT(3) (T + 1) - 2 SQRT(T + T + 1)) (X - 1)

---- --------------------------------------------------------------3 1/12

(T - 1)

Time- 15278 msec.(D9) BATCH DONE

(ClO) closefile(buffer,save);

-.

Page 12: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

* -- . . . -. . . . . , o O .- ,o - --. o'. - *,

I,, :(D2) [DSK. SWATAN]

(C3) showtie:true$Time= 5 msec.

Kamke exam~le 'SO

(C4) batch(exampl,13);

(C5) K180:X-2*'DIFF(YX.2)+3,X,'DIFF(Y,X.I)+(X-2+1-V-2)*Y=F(X):Time= 47 msec.

22 d Y dY 2 2

(D5) X --- + 3 X -- + (X - V + 1) Y = F(X)2 dX

dX

(C6) loadfile(pmain,fasl);

PMAIN FASL DSK SWATAN being loadedLoading doneTime= 413 msec.(D6) DONE

(C7) lode2(k180);dY

2 3d 2 2

d Y dX (X - V + 1) Y F(X)we solve --- + -----------------

2 X 2 2dX X x

dY2 3 -- 2 2d Y dX (X - V + 1) Y

first we solve --- +---- + -- ---------- 02 X 2

dX X

SOLVE FASL DSK MACSYM being loadedLoading done

PCONFL FASL DSK SWATAN being loadedLoading done

the equation is confluent type

PCNEQO FASL DSK SWATAN being loadedLoading done

the solution may be representable by Fukuhara's P-function.[ INF INF 0 J[ ][ 3 ]-I - v I)

ylP[ 2 ](x)

C% V* - -, IE 2*F [ 2 b

PCFPTM FASL DSK SWATAN bigloaded

• ~~~~~~...,o.... ... %....................... .......... .. ... ...-.......... . *... . . .,.

Page 13: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

Loading doneY (X)B. ABS(V)

yz------------------

xY (X)

B. ABS(V)the solution of the homog. eq. is-------------

x

PNONH FASL DSK SWATAN being loaded

Loading donea special sol. for the nonhom. eq. is

x/

I (%PI F(T) J (T) Y (X) - %PI F(T) Y (T) J (X)) dT) ABS(V) ABS(V) ABS(V) ABS(V)

4-. /

0

2Xcontinue? type y or nn;Time: 9943 msec.

x/[I (%PI F(T) J (T) Y (X) %PI F(T) Y (T) J (X)) dTJ ABS(V) ABS(V) ABS(V) ABS(V)I

0(07) -----------------------------------------------------------------------

2X

Y (x)B, ABS(V)

Time= 12131 msec.(D8) BATCH DONE

(C9) closefile(buffer,save):

.,

%

u4t ' ' . , ' .r . ' ."-' . ' . '" ' '. . . .' ,', -' . .". " . . - . .' ' ". . ." " " .. .' -.

Page 14: D-AI37 776 RESEARCH IN ALGEBRAIC ?RNIPULATION(U ... · d-ai37 776 research in algebraic ?rnipulation(u) massachusetts i inst of tech cambridge lab for computer science j moses 14

4 0I

IN

Iz

DTI