Context Navigation

← Previous Change
Next Change →

Changeset 1338 for applications

Timestamp:

04/27/11 19:00:36 (13 years ago)

Author:

sindj

Message:

Zacatky experimentovani v robustlib. Zacleneni vypoctu polynomialni rovnice pro optimalni akci s vyuzitim funkce itpp::roots(). Odstranovani zbylych chyb. JS

Location:

applications

Files:

: 4 modified

robust/main.cpp (modified) (12 diffs)
robust/robustlib.cpp (modified) (1 diff)
robust/robustlib.h (modified) (19 diffs)
trading_models/trading_models_main.cpp (modified) (1 diff)

Legend:

: Unmodified
: Added
: Removed

applications/robust/main.cpp

r1337	r1338
12	12	#include <iostream>
13	13	#include <fstream>
14		#include <itpp/~~signal/poly~~.h>
	14	#include <itpp/itsignal.h>
15	15
16	16	using namespace itpp;
…	…
138	138	vector<vector<string>> strings;
139	139
140		char* file_strings[3] = {"c:\\dataADClosePercDiff", "c:\\ar_student_single","c:\\ar_cauchy_single"};
	140	char* file_strings[3] = {"c:\\dataCDClosePercDiff", "c:\\ar_student_single","c:\\ar_cauchy_single"};
141	141
142	142	for(int i = 0;i<3;i++)
…	…
172	172	RARX* my_rarx = new RARX(2,30,false);
173	173
174		/*
	174
175	175	mat V0 = 0.0001 * eye ( 3 );
176	176	ARX* my_arx = new ARX(0.97);
…	…
178	178	my_arx->set_constant ( false );
179	179	my_arx->validate();
180		*/
181
182		for(int k = 1;k<~~170~~;k++)
	180
	181
	182	for(int k = 1;k<strings[j].size();k++)
183	183	{
184	184	vec condition;
…	…
199	199	conditions[k-3].ins(0,strings[j][k]);
200	200
201		// ~~cout << "modi~~:" << conditions[k-3] << endl;
	201	//cout << "Condition:" << conditions[k-3] << endl;
202	202
203	203	my_rarx->bayes(conditions[k-3]);
	204	//my_rarx->posterior->step_me(1);
204	205
205	206	vec cond_vec;
206	207	cond_vec.ins(0,conditions[k-3][0]);
207	208
208		// my_arx->bayes(cond_vec,conditions[k-3].right(2));
	209	my_arx->bayes(cond_vec,conditions[k-3].right(2));
209	210
210	211
211		if(k>5)
	212	if(k>8)
212	213	{
213	214	//my_rarx->posterior->step_me(0);
…	…
219	220	{
220	221	vec lap_sample = conditions[k-3].left(2);
	222	//lap_sample.ins(lap_sample.size(),1.0);
221	223
222	224	lap_sample.ins(0,LapRNG());
…	…
225	227	}
226	228
	229
227	230	vec sample_pow = sample_prediction;
228	231	vec poly_coefs;
…	…
236	239	{
237	240	sample_pow = elem_mult(sample_pow,sample_prediction);
238		poly_coefs.ins(poly_coefs.size(),~~pow(-1,~~en)*poly_coef);
	241	poly_coefs.ins(poly_coefs.size(),((-1)^en)*poly_coef);
239	242	}
240	243	else
…	…
247	250	while(!stop_iteration);
248	251
249		cvec action =
250
251
252		cout << "MaxLik coords:" << my_rarx->posterior->minimal_vertex->get_coordinates() << endl;
	252	// cout << "Coefficients: " << poly_coefs << endl;
	253
	254	cvec actions = roots(poly_coefs);
	255	bool is_max = false;
	256	for(int t = 0;t<actions.size();t++)
	257	{
	258	if(actions[t].imag() == 0 && actions[t].real()>-1 && actions[t].real()<1)
	259	{
	260	cout << "Action:" << actions[t].real() << endl;
	261	is_max = true;
	262	}
	263	}
	264
	265	if(!is_max)
	266	{
	267	cout << "No maximum." << endl;
	268	}
	269
	270	// cout << "MaxLik coords:" << my_rarx->posterior->minimal_vertex->get_coordinates() << endl;
253	271
254	272	double prediction = 0;
…	…
290	308	}
291	309
292		/*
	310	cout << "Prediction: "<< prediction << endl;
	311
293	312	enorm<ldmat>* pred_mat = my_arx->epredictor(conditions[k-3].left(2));
294	313	double prediction2 = pred_mat->mean()[0];
295		*/
	314
296	315
297	316	ofstream myfile;
298	317	char fstring[80];
299		//char f2string[80];
	318	char f2string[80];
300	319	strcpy(fstring,file_strings[j]);
301		//strcpy(f2string,fstring);
	320	strcpy(f2string,fstring);
302	321
303	322	strcat(fstring,"pred.txt");
304		//strcat(f2string,"2pred.txt");
	323	strcat(f2string,"2pred.txt");
305	324
306	325
307	326	myfile.open(fstring,ios::app);
308	327
309		//myfile << my_rarx->posterior->minimal_vertex->get_coordinates()[0];
	328	// myfile << my_rarx->posterior->minimal_vertex->get_coordinates()[0];
310	329	myfile << prediction;
311	330
…	…
320	339	myfile.close();
321	340
322		/*
	341
323	342	myfile.open(f2string,ios::app);
324	343	myfile << prediction2;
…	…
333	352	}
334	353	myfile.close();
335		*/
	354
336	355
337	356	}

applications/robust/robustlib.cpp

r1335	r1338
33	33	{
34	34	double cur_prob = 0;
	35
	36	/*
	37	if(should_integrate&&new_simplex->vertices.size()!=3)
	38	{
	39	cout << "Error: Wrong vertex count for integration!";
	40	}
	41	*/
35	42
36	43	if(should_integrate)

applications/robust/robustlib.h

r1337	r1338
883	883	{
884	884	current_parent->totallyneutralgrandchildren.insert(sender->totallyneutralchildren.begin(),sender->totallyneutralchildren.end());
885
	885
886	886	for(set<polyhedron*>::iterator tot_child_ref = sender->totallyneutralchildren.begin();tot_child_ref!=sender->totallyneutralchildren.end();tot_child_ref++)
887	887	{
888	888	(*tot_child_ref)->grandparents.insert(current_parent);
	889	}
	890
	891	if(current_parent->totally_neutral == NULL)
	892	{
	893	current_parent->totally_neutral = sender->totally_neutral;
	894	}
	895	else
	896	{
	897	current_parent->totally_neutral = current_parent->totally_neutral && sender->totally_neutral;
889	898	}
890	899
…	…
898	907	current_parent->neutralchildren.push_back(sender);
899	908	current_parent->positiveneutralvertices.insert(sender->positiveneutralvertices.begin(),sender->positiveneutralvertices.end());
900		current_parent->negativeneutralvertices.insert(sender->negativeneutralvertices.begin(),sender->negativeneutralvertices.end());
901
902		if(current_parent->totally_neutral == NULL)
903		{
904		current_parent->totally_neutral = sender->totally_neutral;
905		}
906		else
907		{
908		current_parent->totally_neutral = current_parent->totally_neutral && sender->totally_neutral;
909		}
	909	current_parent->negativeneutralvertices.insert(sender->negativeneutralvertices.begin(),sender->negativeneutralvertices.end());
910	910
911	911	if(sender->totally_neutral)
…	…
922	922
923	923	if(is_last)
924		{
	924	{
925	925
926		/// \TODO Nechapu druhou podminku, zda se mi ze je to spatne.. Nemela by byt jen prvni? Nebo se jedna o nastaveni totalni neutrality?
927		if((current_parent->negativechildren.size()>0&&current_parent->positivechildren.size()>0)\|\|
928		(current_parent->neutralchildren.size()>0&&current_parent->totally_neutral==false))
	926	if((current_parent->negativechildren.size()>0&&current_parent->positivechildren.size()>0)
	927	\|\|(current_parent->neutralchildren.size()>0&&current_parent->totallyneutralchildren.empty()))
929	928	{
930	929	for_splitting[level+1].push_back(current_parent);
…	…
981	980	current_parent->negativechildren.clear();
982	981	current_parent->neutralchildren.clear();
983		current_parent->totallyneutralchildren.clear();
	982	//current_parent->totallyneutralchildren.clear();
984	983	current_parent->totallyneutralgrandchildren.clear();
985	984	// current_parent->grandparents.clear();
…	…
1002	1001
1003	1002	}
1004
	1003
	1004	sender->totallyneutralchildren.clear();
1005	1005	}
1006	1006	}
…	…
1049	1049	{
1050	1050
1051
	1051	/*
1052	1052	if(i==statistic.size()-1)
1053	1053	{
…	…
1055	1055	cout << "Order:" << ((toprow*)horiz_ref)->condition_order << endl;
1056	1056	}
	1057	*/
	1058
	1059	cout << "Stepped." << endl;
1057	1060
1058	1061	/*
…	…
1259	1262	current_vertex->negativeneutralvertices.insert(current_vertex);
1260	1263	current_vertex->positiveneutralvertices.insert(current_vertex);
1261		}
	1264	}
	1265	else
	1266	{
	1267	current_vertex->totally_neutral = false;
	1268	}
1262	1269	}
1263	1270
…	…
1296	1303	if(should_remove)
1297	1304	{
	1305	/*
1298	1306	for(int i = 0;i<for_merging.size();i++)
1299	1307	{
…	…
1305	1313	cout << endl;
1306	1314	}
	1315	*/
	1316
	1317	cout << "Merging." << endl;
1307	1318
1308	1319	set<vertex*> vertices_to_be_reduced;
…	…
1407	1418	}
1408	1419
	1420	/*
1409	1421	current_positive->totallyneutralchildren.insert(current_negative->totallyneutralchildren.begin(),current_negative->totallyneutralchildren.end());
1410	1422
1411	1423	current_positive->totallyneutralchildren.erase(*merge_ref);
	1424	*/
1412	1425
1413	1426	current_positive->totallyneutralgrandchildren.insert(current_negative->totallyneutralgrandchildren.begin(),current_negative->totallyneutralgrandchildren.end());
…	…
1423	1436	current_positive->negativechildren.clear();
1424	1437	current_positive->neutralchildren.clear();
1425		current_positive->totallyneutralchildren.clear();
	1438	// current_positive->totallyneutralchildren.clear();
1426	1439	current_positive->totallyneutralgrandchildren.clear();
1427	1440	current_positive->positiveneutralvertices.clear();
…	…
1558	1571	}
1559	1572
	1573	/*
1560	1574	vector<int> sizevector;
1561	1575	for(int s = 0;s<statistic.size();s++)
…	…
1564	1578	cout << statistic.row_size(s) << ", ";
1565	1579	}
1566
1567		cout << endl;
	1580	*/
	1581
	1582	//cout << endl;
	1583
	1584	// this->step_me(2);
1568	1585
1569	1586	if(should_add)
…	…
1722	1739	}
1723	1740
1724
	1741	/*
1725	1742	sizevector.clear();
1726	1743	for(int s = 0;s<statistic.size();s++)
…	…
1731	1748
1732	1749	cout << endl;
	1750	*/
1733	1751
1734	1752	/*
…	…
1943	1961
1944	1962	//// cout << "Sigma: " << sigma << endl;
1945		//// cout << "Nr. of runs: " << number_of_runs << endl;
	1963	//// cout << "Nr. of sigma runs: " << number_of_runs << endl;
1946	1964
1947	1965	int dimension = (*s_ref)->vertices.size()-1;
…	…
2087	2105	sample_mat.ins_col(0,sample_coordinates);
2088	2106
2089		cout << sample_mat.cols() << ",";
	2107	// cout << sample_mat.cols() << ",";
2090	2108	}
2091	2109
…	…
2096	2114	}
2097	2115
2098		cout << endl;
	2116	// cout << endl;
2099	2117	return sample_mat;
2100	2118	}

applications/trading_models/trading_models_main.cpp

r1266	r1338
7	7	*/
8	8
	9	//<<<<<<< .mine
	10	#include "trading_models_lib.h"
	11	#include <iostream>
	12	#include "estim/arx.h"
	13	//=======
9	14	#include "trading_models_lib.h"
10	15	#include "estim/arx.h"
	16	#include <vector>
	17	#include <string>
	18	#include <sstream>
	19	#include <fstream>
	20
	21
	22	//>>>>>>> .r1284
11	23	using namespace bdm;
12	24
13
14
15		int main ( int argc, char* argv[] ) {
16
17		for(int i=0;i<10000;i++)
	25	/*
	26	string cislo(string s) // tato procedura spracova tie cisla z toho textoveho suboru tak aby sa odstranilo to e-001 a boli v spravnom
	27	//formate. Lebo predtym to vyhadyovalo divne cisla, nevedel som ci je to zato, ze bayes si stym neporadi, ale yjavne nepomohlo
	28	{
	29	int b=s.find('e');
	30	string t=s.substr(0,b);
	31	string poz=s.substr(b+1,s.size());
	32
	33	istringstream g(poz);
	34	int pozc;
	35	g>>pozc;
	36	istringstream k(t);
	37	double csl;
	38	k>>csl;
	39	if (pozc<0) //pozc moze byt niekedy aj kladne
18	40	{
19		cout << i;
	41	for(int i=1;i<=-pozc;i++)
	42	csl/=10;
	43	} else
	44	{
	45	for(int j=1;j<=pozc;j++)
	46	csl*=10;
20	47	}
21		//prior
22		mat V0 = 0.00001 * eye ( 2 );
23		V0 ( 0, 0 ) = 0.1; //
24		ARX Ar;
25		Ar.set_statistics ( 1, V0 ); //nu is default (set to have finite moments)
26		Ar.set_constant ( true );
27		Ar.validate();
28		// forgetting is default: 1.0
29		mat Data = concat_vertical ( randn ( 1, 100 ), ones ( 1, 100 ) );
30		Ar.bayes_batch ( Data );
31
32
33		}
34
	48
	49	ostringstream o;
	50	o<<csl;
	51	return o.str();
	52	} */
	53
	54	double sumastlpec(int k,vector<vec> pole,vector<vec> pravd) { //robi sumu k-teho stlpca, pouzivam na konci pri ratani pravdepodobnosti
	55	double r=0;
	56	for (int i=0;i<pole.size();i++)
	57	{
	58	r+=pole[i][k]*pravd[i][k];
	59	}
	60	return r;
	61	}
	62
	63	int main () {
	64	vector<vector<string>> ADdata; //nacitavanie dat do pola ADdata -funguje spravne
	65	ifstream myfile("C:\\AD_dataupravene.txt");
	66	if (myfile.is_open())
	67	{
	68	while ( myfile.good() )
	69	{
	70	string line;
	71	getline(myfile,line);
	72	vector<string> parsed_line;
	73	while(line.find(' ') != string::npos) //jeden kanal je jeden riadok, na zaciatku a na konci {,}, data oddelene ciarkou a medzerou.
	74	{
	75	line.erase(0,1); //toto nie je yrovna peknz sposob,ale pri poslednom nacitani cisla v riadku sme uz nemali ziadnu medyeru a cyklus by sa posledny krat nevykonal, tak tu medzeru odstranujeme vzdy tu
	76	int loc = line.find(','); //ale pri poslednom cisla to nenajde ziadnu ciaarku, tak potom co prida do parsed_line?
	77	parsed_line.push_back(line.substr(0,loc));
	78	line.erase(0,loc+1); //odstranujeme ciarku za kazdym cislom
	79	}
	80	ADdata.push_back(parsed_line); //3927 dat v riadku, 6 riadkov
	81	}
	82	}
	83	myfile.close(); //konec nacitavania dat
	84
	85	vector<vec> norm; //do norm zapisujeme normalizacne faktory
	86	for (int h=1;h<=2;h++) //cyklus ktory ovplzvnuje konstantu h=1- model s konstantou, h=2, bez konstanty
	87	{
	88	bool b; //b pouzivame pri set_constant
	89	if(h==2)
	90	b=false;
	91	else
	92	b=true;
	93	int g=2;
	94	while (g<=4) //cyklus co meni rozmery matice V
	95	{
	96	mat V0 = 0.0001 * eye ( g ); // aj tato matica ma vplyv na normalizacny faktor, nemoze byt aj preto taky velky, ako inak by sa dala zvolit?
	97
	98	int p=0;
	99	while (p<=1) //tento cyklus prechadza vacsinou len raz, vtedy p=0 a nic to neovplvni, ale pri AR(2) modely to bude vykonavat 2 krat aj pre p=1, ked bude brat do condition aj rozne kanale z toho isteho casu
	100	{
	101	int i=0;
	102	while(i < ADdata.size()-p) //niekedy sa ten cyklus ma vykonat len raz, preto nepouzivam for cyklus
	103	{
	104	int j=p*(i+1); //j=0 alebo j=i+1
	105
	106	while(j < ADdata.size())
	107	{
	108	ARX Ar;
	109	Ar.set_statistics ( 1, V0 ); //nu is default (set to have finite moments)
	110	Ar.set_constant ( b );
	111	Ar.validate(); // forgetting is default: 1.0
	112	vec pomocka; //pri kazdej jednej hypoteze zapisujeme normalizacne faktory do pomocky, tu potom ako riadok pridame do norm
	113	for(int k = 0;k<341;k++) //prechadyame "po riadkoch", teda v case. Nejake hodnoty su len po index 340, dalej uz #INF000
	114	{
	115	vec condition;
	116	vec predikce;
	117	predikce.ins(0,ADdata[3][k+2]); //predpovede nacitavame a zadavame do Bayes zvlast
	118	condition.ins(0,ADdata[i][k+1]);
	119	condition.ins(0,ADdata[j][k+p]);//zmena i -> j aby to bralo regresory z roznych riadkov, ak p=1 bereme data z toho isteho casu
	120
	121	cout << "Pred:" << predikce << ", ";
	122	cout << "Cond:" << condition << endl;
	123
	124	Ar.bayes(predikce,condition.right(g+h-3)); //z condition berem len urcity pocet prvkov, bud 0, 1,alebo 2, lebo nepotrebujem vzdy vsetky (AR(1) model)
	125	pomocka.ins(pomocka.size(),Ar.posterior().lognc()); //nie je tu exponenciala! -aby to bolo mensie
	126	}
	127	norm.push_back(pomocka);
	128	if ((g==3 && h==2) \|\| (g==4)) //tento cyklus sa bude opakovat, len ak mame maticu V0 roymeru 4x4, to je AR(2) model s konst, alebo podobne len g=3, h=2, teda AR(2)bez kons
	129	{
	130	j++;
	131	} else
	132	{
	133	j=ADdata.size(); //priradenim tejto hodnoty do j sa cyklus uz viac krat nevykona
	134	}
	135	}
	136	if (b==true && g==2) //pre model AR(0) s konstantou robi tento cyklus len raz, v ostatnych pripadoch viac-krat
	137	{
	138	i=ADdata.size();
	139	} else
	140	{
	141	i++;
	142	}
	143	}
	144	if ((g==3 && h==2) \|\| (g==4) )
	145	{p++;} else {p=2;}
	146	}
	147	if (h==2 && g==3) //pripad g=4, a konstanta zaroven nas uz nezaujima, vtedy to ukoncime
	148	{
	149	g=5; //ak priradime takuto hodnotu, cyklus while sa uz nevykona
	150	} else
	151	{
	152	g++;
	153	}
	154	}
	155	}
	156
	157	/* //tu je to povodne
	158	mat V0 = 0.0001 * eye ( 2 ); //pre pripad samotnej konstanty
	159	ARX Ar;
	160	Ar.set_statistics ( 1, V0 ); //nu is default (set to have finite moments)
	161	Ar.set_constant ( true );
	162	Ar.validate(); // forgetting is default: 1.0
	163	vector<double> pom1;
	164	for(int k = 0;k<140;k++) //prechadyame "po riadkoch"
	165	{
	166	vec predikce;
	167	vec cond;
	168	cond.ins(0,ADdata[3][3]);
	169	predikce.ins(0,ADdata[3][k+2]); //predpovede nacitavame a zadavame do Bayes zvlast
	170	Ar.bayes(predikce,cond.right(0));
	171	pom1.push_back(exp(Ar.posterior().lognc()));
	172	}
	173	norm.push_back(pom1);
	174
	175	for (int a=2;a<=3;a++) //AR(1) bez a potom s konstantou
	176	{
	177	bool b=false; //b pouzivame pri set_constant
	178	if(a==3)
	179	b=true;
	180	mat V0 = 0.0001 * eye ( a ); //pre pripad samotnej konstanty
	181
	182	for (int p=0;p < ADdata.size();p++)
	183	{
	184	ARX Ar;
	185	Ar.set_statistics ( 1, V0 ); //nu is default (set to have finite moments)
	186	Ar.set_constant ( b );
	187	Ar.validate(); // forgetting is default: 1.0
	188	vector<double> pom1;
	189	for(int k = 0;k<140;k++) //prechadyame "po riadkoch"
	190	{
	191	vec predikce;
	192	vec condition;
	193	condition.ins(0,ADdata[p][k]);
	194	predikce.ins(0,ADdata[3][k+1]); //predpovede nacitavame a zadavame do Bayes zvlast
	195	Ar.bayes(predikce,condition);
	196	pom1.push_back(exp(Ar.posterior().lognc()));
	197	}
	198	norm.push_back(pom1); //normalizacne faktory pre urcitu kombinaciu regresorov(teda po kazdom riadku) ulozi do pola norm
	199	}
	200	}
	201
	202
	203	for (int g=3;g<=4;g++) //tento cyklus je az do konca, raz to robime s konstantou, raz bez.
	204	{
	205	bool b; //b pouzivame pri set_constant
	206	if(g==3)
	207	b=false;
	208	else
	209	b=true;
	210	mat V0 = 0.0001 * eye ( g );
	211	for(int i = 0;i < ADdata.size();i++) //po pocet riadkov, co bz malo byt 6
	212	{
	213	for(int j = i+1; j<ADdata.size();j++)
	214	{
	215	ARX Ar;
	216	Ar.set_statistics ( 1, V0 ); //nu is default (set to have finite moments)
	217	Ar.set_constant ( b );
	218	Ar.validate();
	219	// forgetting is default: 1.0
	220	vector<double> pomocka;
	221	for(int k = 0;k<140;k++) //prechadyame "po riadkoch"
	222	{
	223	vec condition;
	224	vec predikce;
	225	predikce.ins(0,ADdata[3][k+1]); //predpovede nacitavame a zadavame do Bayes zvlast
	226	condition.ins(0,ADdata[i][k]);
	227	condition.ins(0,ADdata[j][k]);//zmena i -> j qby to bralo regresory z roznych riadkov
	228	Ar.bayes(predikce,condition);
	229	pomocka.push_back(exp(Ar.posterior().lognc()));
	230	}
	231	norm.push_back(pomocka);
	232	}
	233	}
	234	for(int i = 0;i < ADdata.size();i++) //po pocet riadkov, co bz malo byt 6
	235	{
	236	for(int j = 0; j<ADdata.size();j++)
	237	{
	238	ARX Ar;
	239	Ar.set_statistics ( 1, V0 ); //nu is default (set to have finite moments)
	240	Ar.set_constant ( b );
	241	Ar.validate();
	242	// forgetting is default: 1.0
	243	vector<double> pomocka;
	244	for(int k = 0;k<140;k++) //prechadyame "po riadkoch"
	245	{
	246	vec condition;
	247	vec predikce;
	248	predikce.ins(0,ADdata[3][k+2]); //predpovede nacitavame a zadavame do Bayes zvlast
	249	condition.ins(0,ADdata[i][k]);
	250
	251	condition.ins(0,ADdata[j][k+1]);//zmena i -> j qby to bralo regresory z roznych riadkov
	252	Ar.bayes(predikce,condition);
	253	pomocka.push_back(Ar.posterior().lognc());
	254	}
	255
	256	norm.push_back(pomocka);
	257	}
	258	}
	259	}*/
	260	vector<vec> prsti; //hypotez je 85
	261	int m,n,p;
	262	for(p=0;p<115;p++) //inicializuem apriorne pravdepodobnosti
	263	{
	264	vec k;
	265	k.ins(0,1/115.);
	266	prsti.push_back(k);
	267	}
	268	// v ramci riadku v poli norm su hodnoty pre jednu hypotezu v roznych casoch, pocitanie pravdepodobnosti z norm. faktorov
	269	for (m=0;m<norm[1].size();m++)
	270	{ double k=sumastlpec(m,norm,prsti);
	271	for(n=0;n < norm.size();n++)
	272	{
	273	prsti[n].ins(prsti[n].size(),norm[n][m]*prsti[n][m]/k);
	274	}
	275	}
	276	ofstream file; //zapis pravdepodobnosti do suboru
	277	file.open("prsti_hypot.txt");
	278	for(int i=0;i < prsti.size();i++)
	279	{
	280	for(int j=0;j < prsti[i].size();j++)
	281	{
	282	if(j!=prsti[i].size()-1)
	283	{
	284	file << prsti[i][j]<<" ";
	285	}else
	286	{
	287	file<<prsti[i][j]<<endl;
	288	}
	289	}
	290	}
	291	file<<endl;
	292	file.close();
	293
	294	}
	295
	296
	297