In this paper we will analyze regulation of labour relations, speak about management problems in the companies looking from the law side. Besides analyze labor rights, their disputes, individual labour disputes and the labour disputes considered in courts.

Teisė  Referatai   (16 psl., 26,11 kB)

Referato formos rašinys įskaitos potemei. Kalba, visų pirma, yra tradicija, perduodama iš kartos į kartą, leidžianti bendrauti, susikalbėti, suprasti vieniem kitus skirtingų kartų atstovams, gyvenantiems tuo pačiu laikotarpiu. Žinoma, perduodama kalba kinta – ji gyva. Kiekviena karta keičia kalbą, pritaikydama ją prie gyvenimo aplinkybių, situacijų. Kalbos baigties – tradicijos nutraukimo – atvejai akivaizdžiai parodo, kad tarp kartų vis dėlto yra spraga.

Lietuvių kalba  Pagalbinė medžiaga   (5 psl., 20,33 kB)

Šie metodai turi labai lėtą konvergavimą, bet teoriškai jie gali įveikti vietinį minimumą (local minima.) Kitas trūkumas yra tas, kad vienas turi valdyti daugybę vidinių kintamųjų (kiekvienam svoriui nustatyti triukšmo periodus), kas nėra labai efektyvu. Arba apibrėžti tik išorinius kintamuosius - tokius kaip įėjimo signalas (input), norimas signalas ir žingsnio dydis. Iš pragmatiškos požiūrio pusės labai pageidaujami būtų taip vadinami on-line (tiesioginiai) algoritmai, t.y. algoritmai, kur atskiram pavyzdžiui svoriai kaskart būtų atnaujinami. Bet žvelgiant iš (annealing) modeliavimo pusės į stochastinę on-line atnaujinimo metodų prigimtį , jų realizavimas nebūtų efektyvus. Dėl šių priežasčių bus bandoma atlikti tokį eksperimentą: pridėti triukšmus prie norimo signalo ir eksperimentiškai ištirti tokios procedūros privalumus. Triukšmas taip pat buvo naudojamas gradiento perdavimo (descent) procedūrose. Holmstrom išanalizavo statinio BP algoritmo [Holmstrom and Koistinen, 1992] apibendrinimo galimybę, kuomet atsitiktinis triukšmas įvedamas į išorinius signalus. Šie bandymai parodė, kad apibendrinimas gali būti pagerintas naudojant bandomuosiusose (training) duomenyse papildomus triukšmus. Matsuoka pademonstravo, kad ir triukšmo įvedimas į vieną įėjimo signalą gali pagerinti apibendrinimą (generalization) [Matsuoka, 1992]. Abu autoriai susikoncentravo tik ties tinklo apibendrinimo galimybe, tačiau jie nenagrinėjo triukšmų poveikio mokymosi greičiui ir išėjimo iš local minima galimybės. II Mokymosi su papildomais triukšmais atitinkamame signale analizė II.1 Klasikinis stebimas mokymasis Šioje dalyje kaip mokymosi sistemos prototipas yra naudojamas daugiasluoksnis perceptronas (perceptron) (MLP) su dviem lygiais. Tačiau išvados gali būti atvaizduojamos atsikartojančiose topologijose. Šiame tinkle, xk aprašo iėjimo vektoriaus vieną elementą; yi yra išėjimo lygio i-tasis išėjimas; Wij nusako svorius tarp paslėpto ir išėjimo sluoksnių; Vjk yra svoris tarp įėjimo ir paslėpto sluoksnio; ir Pj nusako paslėpto sluoksnio aktyvavimą. Pateiktas čia apmokymo algoritmas - tai atgalinio mokymo (backpropagation) (BP) algoritmas [Rumelhart et al, 1986]. Tegul di(t) žymi kelis norimus išėjimo neurono i laiko momentu t atsakymus, kur t yra diskretaus laiko indeksas. Galima apibrėžti klaidos signalą, kaip skirtumą tarp norimo atsakymo di(t) ir turimo atsakymo yi(t). Tai nusakomo (1) formulė: Pagrindinis mokymosi tikslas yra minimizuoti kainos funkciją, kurią nusako klaidos signalas ei(t), taip, kad turimas kiekvieno išėjimo neurono atsakymas tinkle statistikine prasme artėtų prie norimo atsakymo. Kriterijus naudojamas kainos funkcijai yra Vidurkio-Kvadrato-Klaidos (Mean-Square-Error) (MSE) kriterijus, apibrėžiamas kaip klaidos kvadrato sumos vidurkio-kvadrato reikšmė [Haykin, 1994]: Kur E yra statistikinis tikimybės operatorius ir sumuojami visi išėjimo sluoksnio neuronai (i=1,…,M). Kainos funkcijos J minimizavimas atsižvelgiant į tinklo parametrus lengvai g.b. formuluojamas gradiento mažinimo (gradient descent) metodu. Šios optimizavimo procedūros problema yra ta, kad jai reikia žinių apie neapibrėžtų procesų, generuojančių pavyzdžius, statistikines charakteristikas. Praktiškai tai gali būti apeita, optimizavimo problemai surandant artimą sprendinį. Klaidos kvadratų sumos momentinė reikšmė (Instantaneous value of the sum of Squared Errors) (ISE) yra pasirinkimo kriterijus [Haykin, 1994]: Po to tinklo parametrai (svoriai) yra pritaikomi ε(t). Faktiškai ši procedūra vadovaujasi taip vadinamu LMS algoritmu, kuomet svoriai yra atnaujinami kartu su kiekvienu pavyzdžiu [Widrow and Hoff, 1960]. II.1 Mokymasis su norimu triukšmingu signalu Vietoj to, kad svorių pritaikymui naudoti norimą signalą di(t), kaip norimas signalas išėjimo neuronui i imamas naujas signalas di(t)+ ni(t), kur ni(t) yra triukšmo periodas. Šiam triukšmo periodui priskiriamas nulinės reišmės baltas triukšmas su σ2 pokyčiu (variance) , nepriklausančiu nei nuo įėjimo signalo xk(t) nei nuo norimų signalų di(t). Neapibrėžtas triukšmo perdavimas yra priskiriamas Gauso ar vienarūšiam perdavimui. Čia norima įrodyti, kad šis naujas norimas signalas neįtakoja galutinės svorių reikšmės statistikine prasme. Tai užtikrina, kad nauja savybė sprendžia originalią optimizavimo problemą. Turint naujus norimus signalus, MSE (4) lygties gali būti perrašyta taip: Nėra sunku įrodyti [Richard and Lippmann 1991; White, 1989; Haykin, 1994], kad (4) lygtis yra lygi Kur ‘|’ simbolis žymi sąlygines galimybes (probabilities), ir 'var' yra kitimų (variance) sutrumpinimas. Pastebėkite, kad antras periodas dešinėje (5) lygties pusėje prisidės prie bendros klaidos J ir koks ir bebūtų mokymosi progresas, jis neįtakos galutinės svorių reikšmės, kadangi jis nėra tinklo svorių funkcija. Optimali svorių reikšmė yra apsprendžiama tiktai pirmo (5) lygties periodo. Kuomet triukšmas yra nulinės reikšmės baltas triukšmas ir jis nepriklauso nei nuo norimo, nei nuo įėjimo signalų, mes turime (6) lygtis rodo, kad triukšmas iš lygties, kuri apibrėš galutines svorių reikšmes, dingsta, taigi mokymassi su norimo triukšmo signalu duos rezultatų, originalios optimizavimo problemos sprendimo prasme, t.y. be triukšmo pridėjimo prie norimo signalo. (learning with the noisy desired signal will yield in the mean the solution for the originaloptimization problem, i.e., without the noise added to the desired signal.) Reiktų konstatuoti, kad ši išvada galioja visoms architektūrų rūšims.Atlikimo funkcijai apibrėžti reikalingi tik išoriniai matavimai (MSE), ir tai nėra susiję nei su topologija nei su kainos funkcijos apibrėžimo būdu (statiniu ar kintamu). Nors šis sprendimas yra patenkinamas, reikia prisiminti, kad mus domina on-line algoritmas, kur yra mokymosi dinamika, t.y. kaip mokymosi progresas yra veikiamas triukšmų. II.3 On-line algoritmas mokymuisi veikiant norimam triukšmingam signalui Reiktų pažymėti, kad atliekamos, modifikacijos, jokiais būdais neveikia atgalinio mokymosi algoritmo realizacijos, kadangi yra modifikuojamas tik signalas, kuris yra įvedamas kaip norimas rezultatas. Taigi, siūlomos modifikacijos gali būti taikomos dar neegzistuojančioms modeliavimo sistemoms. Svarbi problema, kaip modeliavimo metu valdyti triukšmų kaitą (variance). Dėl to tolimesniame skyriuje bus apžvelgiama tiukšmų įtaka momentiniam gradientui. II.4 Norimo triukšmingo signalo gradiente analizė. Svorinio vektoriuas pritaikymo statiniame BP algoritme formulė, tiklui atvaizduotame 1 pav. norimame signale be triukšmų yra [Hertz et al.,1991] Svoriams tarp paslėpto sluoksnio ir išorinio sluoksnio, ir Svoriams tarp iėjimo sluoksnio ir paslėpto sluoksnio, kur ŋ yra žingsnio dydis. Su triukšmingu norimu signalu, ISE (3) lygties tampa: Lygtyse (7) ir (8) įrašant naują reikšmę εnoisy(t), gausime lygtis Palyginus lygtis (7) su (10) ir (8) su (11) daroma išvada, kad triukšmo pridėjimo prie norimo signalo poveikis, tai extra stochastinio periodo svoriniame vektorių taikyme įtraukimas, kas gali būti modeliuojama kaip pridėtinis momentinio gradiento triukšmas (pertirbation) betriukšminiam atvejui. Stochastinio periodo bendra forma Kur N(t) yra veiksmo funkcija gauta pakeitus originalią klaidą d(t)-y(t) įvestu triukšmu n(t). Panagrinėkime papildomų periodų (extra terms) statistines savybes (10) ir (11) lygtyse ir pastebėkime kaip jos veikia svorinių vektorių statistiką. Bet pirmiausia, apibrėžkime atsitiktinius kintamuosius: Jeigu atsitiktiniai kintamieji ir nepriklauso vienas nuo kito, ir g ir f funkcijos yra Borelo funkcijos, tuomet f ir gtaip pat yra nepriklausomos [Feller, 1966]. Realiausios funkcijos f(x) įskaitant sigmoido funkciją, plačiai naudojamą neuroniniuose tinkluose yra Borelio funkcijos. Taigi, galima daryti išvadą, kad (10) ir (11) lygtyse triukšmas n(t) nepriklauso nuo O taip pat ir nuo Todėl gali būti užrašytos papildomų periodų (extra terms) tikimybės ir Jų kitimas (variance) ir Iš (14) ir (15) lygties galima daryti išvadą, kad nulinės reikšmės atsitiktinis triukšmas norimame signale nedaro įtakos svoriniams vektoriams, taigi pagrindinė papildomo stochastinio periodo (extra stochastic term) svorio atnaujinimo reikšmė yra nulis. Iš (16) ir (17) lygties daromos dvi svarbios išvados: triukšmas pridėtas prie norimo signalo veikia svorio atnaujinimo kitimą proporciškai kiekvieno svorio jautrumui. Tai reiškia, kad atskiro triukšmo šaltinis išėjime yra išverčiamas į skirtingus triukšmų stiprumus kiekvienam svoriui. Antra, žingsnio dydis arba išorinio triukšmo šaltinio kitimas valdys papildomų periodų (variance of the extra terms) svoriniuose vektorių prisitaikymo formulėse kitimą, gaunamą pridedant triukšmą prie norimo signalo. Pastebima, kad, kai = 0 arba triukšmo kitimas yra nulis, tuomet stochastinis periodas (stochastic terms) išnyksta - lieka tik originalus svorio atnaujinimas (t.y. sprendžiama originali optimizacijos problema). Šie aspektai ir idėjos gautos iš globalios optimizacijos pateikia empirines taisykles išorinių triukšmų šaltinių valdymui, gaunat reikšmingus rezultatus. Modeliavimo pradžioje norėtųsi svoriams uždėti atsitiktinius trikdžius (perturbation), tam, kad būtų leista algoritmui pabėgti iš vietinio minimumo (local minima.). Tačiau artėjant prie adaptacijos pabaigos trikdžių (perturbation) kitimas turi būti sumažintas iki nulio taip, kad svoriai galėtų pasiekti reikšmes duotas originalios optimizacijos problemos. Toliau bus naudojamas (annealing) tvarkaraštis, pasiūlytas Moody [Darken, Chang, and Moody, 1992] Kur o yra inicijuojamo žingsnio dydis, c yra paieškos laiko konstanta, ir NI - iteracijos numeris. Šių konstantų reikšmės turės būti apibrėžtos eksperimentiškai, kadangi jos priklauso nuo problemų. III Modeliavimo rezultatai Patvirtinant anksčiau atliktą analizę, modeliavimo rezultatai bus pateikiami dviem pavyzdžiais. Vienas jų naudoja dviejų-lygių MLP, taip vadinamos lygiškumo problemos (parity problem), kuri buvo pademonstruota vietinio minimumo (local minima) atveju, pažinimui [Rumelhart et al, 1986]. Iš modeliavimo rezultatų bus matyti, kad mokymasis labiausiai gali būti pagerintas naudojant numatytą metodą (proposed approach) ir globalų minimumą, pasiektą statistikine prasme. Kitas pavyzdys naudoja dinaminį neuroninį tinklą TDNN [Waibel et al., 1989] laiko signalų modeliavimui. Antro modeliavimo rezultatai taip pat patvirtina ankstesnę analizę. III.I Eksperimentai su MLP Spresime 3 bitų lygiškumo problemą. Tinklo dydis 3-3-1, t.y. 3 įėjimo neuronai, 3 paslėpti neuronai, ir 1 išėjimo neuronas. Netiesiškumas (nonlinearity) yra logistinė funkcija. Tiesioginis atgalinis mokymas (backpropagation) yra naudojamas abiem atvejais. Buvo pridėtas Gauso (Gaussian) triukšmas su  =0.001 prie norimo signalo ir parinkti atitinkami parametrai 18 Lygtyje: c= 500 ir o= 0.3. Rezultatai parodyti 2 paveikslėlyje. Stora linija vaizduoja mokymasi su triukšmingu norimu signalu, o punktyrinė linija - su originaliu norimu signalu. Šis pavyzdys rodo, kad mokymasis artėja prie lokalaus minimumo, (local minimum) kuomet naudojamas originalus norimas signalas, bet naudojant triukšmingą norimą signalą mokymasis pasiekia globalų minimumą (global minimum) . Svarbu pabrėžti, kad mokymasis su originaliu signalu, naudoja pastovų žingsnio dydį, kai tuo tarpu signalo su triukšmais žingsnis yra gaunamas iš (18) lygties. Naudojant skirtingus žingsnio dydžius ir skirtingus pradinius (initial) svorius, buvo pasiekti panašūs rezultatai. Tam, kad patvirtinti šio algoritmo konvergavimo galimybę, buvo remiamasi Monte Carlo modeliavimai su 100 bandymu. Rezultatai pavaizduoti 3 Paveiksle, kur punktyrinė linija yra 100 veiksmų rezulatai originaliam signalui, o stora linija - 100 veiksmų rezulatatai triukšmingam signalui. Šiame eksperimente, svoriai yra parenkami atsistiktinai, o žingsnio dydis o atsitiktinai parenkamas iš intervalo [0,1, 0,7]. Kuomet globalus minimumas (global minimum) yra 0, tuomet yra lengva paskaičiuoti reikšmę ir pokytį (mean and variance) 100 galutinių klaidų, kurios pateiktos 1 Lentelėje. Dar daugiau, mokymasisi su triukšmingu signalu laike 99% priartėjo prie globalaus minimumo, o su originaliu signalu tik 26%. Iš 1 Lentelės, galima daryti išvadą, kad su triukšmingu signalu, mokymasis konverguoja į globalų minimumą; bet su originaliu signalu, mokymasis statistikine prasme nekonverguoja. Taigi, iš šių modeliavimo rezultatų galima daryti išvadą, kad triukšmingas signalas leidžia mokymosi algoritmui išeiti iš lokalaus minimumo (local minima). III. II Eksperimentai su dinaminiu neuroniniu tinklu III.I dalyje buvo pademonstruoti statinio neuroninio tinklo modeliavimo rezultatai. Tam, kad patvirtinti, jog aprašytas metodas taip pat veikia ir dinaminiuose neuroniniuose tinkluose. Dinaminės sistemos modeliavimui bus naudojamas TDNN [Waibel et al.,1989]. Bus nagrinėjama tokia sistema, Kur ',' žymi diferencijavimo operatorių. Sistemos įėjimai yra sinusoidžių aibė, Su atsitiktine faze l.. 4 ir 5 Paveikslėliuose vaizduojami sistemos įėjimo ir atitinkamai normalizuoti išėjimo signalai. TDNN tinklo struktūra demonstruojama 6 Paveikslėlyje, kur naudojamas keturių lygių vėlinimas. Šiame tinkle, kaip įėjimai į paslėptą lygį naudojami tik du įėjimo signalai x(t) ir x(t-4). Netiesiškumas (nonlinearity) paslėptuose neuronuose - tai logistinė funkciją. Išėjimas turi vieną tiesinį neuroną. Mokymosi algoritmas - tai BP, kur pavyzdžių klaidos surandamos atimant tinklo išėjimą y(t) iš sistemos d(t) išėjimo. Mokymosi kreivės pavaizduotos 7 Paveikslėlyje, kur triukšmo signalo žingsnio dydis surandamas naudojant (18) Lygtį, kur c=10, 0.01, ir triukšmo kitimas 2 =0.001 . Aiškiai matyti, kad su triukšmingu signalu konvergavimas yra greitesnis ir pasiekiamas žemesnis MSE. Naudojant skirtingus žingsnio dydžius ir svorius, pasiekiami panašūs rezultatai su MSE minimumu lygiu 0,0091. Paveikslėlyje 8 pavaizduotos mokymosi kreivės 100 mokymosi veiksmų, kur žingsnio dydis atsitiktinai parenkamas iš intervalo [0,1, 0,01], o svoriai taip pat yra atsitiktiniai skaičiai. Kadangi nėra žinomas šios problemos globalus minimumas, todėl naudojamas MSE minimali reikšmė 0,0091 kaip globalaus minimumo įvertis. 2 Lentelė atspindi statistikinius rezultatus 100 eksperimentų. 72% (trails) su triukšmingu signalu pasiekė globalų minimumą, ir tik 8% (trails) pasiekė globalų rezultatą su originaliu signalu. Daroma išvada, kad mokymasisi su triukšmingais signalais yra mažiau nepastovus ir mokymosi kreivės taip pat yra daug lygesnės. (smoother) IV.Discussion Eksperimentiškai buvo pademonstruota, kad mokymasis su triukšmingais signalais padidina pastovaus žingsnio dydžio BP algoritmo paieškos galimybes. Tai yra pasiekiama be papildomos kainos algoritmų realizavimo perioduose (This is accomplished at no extra cost in terms of algorithm implementation,), kadangi naudojamas tiesioginis atgalinis mokymas (straight backpropagation.) Papildomos savybės yra gaunamos įvedant nulinės reikšmės valdomo kitimo Gauso triukšmą ir žingsnio dydžio nustatymui pasinaudojant (18) Lygtimi. Buvo pademonstruota, kad triukšmo pridėjimas prie norimo signalo svorių atnaujinimo formulėse prideda nulinės reikšmės stochastinį periodą (that adding noise to the desired signal adds a zero mean stochastic term in the weightupdate formulas.). Nors atskiras triukšmo signalas ir yra įvedamas į norimą signalą, tačiau stochastinio periodo kitimas kiekvienam tinklo svoriui skiriasi (proporcingai kiekvieno svorio jautrumui). Dar daugiau, šio periodo kitimas tiesiogiai valdomas žingsnio dydžio arba išorinio triukšmo šaltinio kitimo. Tai reiškia, kad triukšmo pridėjimas prie norimo signalo yra labai paprasta ir efektyvi procedūra mokymosi proceso ištraukimo iš lokalaus minimumo. Kitimas arba žingsnio dydis turi būti parinktas (anealing) pritaikymo metu. Parinkimo (anealing) realizavimui buvo panaudota Moodžio paieška ir konvergavimo procedūra, tačiau kiekvienai problemai spręsti parametrai turi būti surandami eksperimentiškai. Žingsnio dydžio planavimas, toks, kad būtų įveiktas lokalus minimumas, išlieka atviras klausimas ne tik šiame metode, bet taip pat ir kituose stochastiniuose algoritmuose tokiuose kaip sumodeliuotas parinkimas (simulated annealing) [Kirkpatrick et al., 1983]. Mokymosi algoritmų lankstumo padidinimui yra siūlomi du skirtingi žingsnių dydžiai, vienas gradientui ir kitas - triukšmui. Ši procedūra turi neišvengiamą jungtį su globaliu optimizavimo metodu, vadinamu stochastiniu funkciniu nesklandumų šalinimu (stochastic functional smoothing) [Rubinstein, 1981 and 1986]. Priede aiškinama, kad tiesioginė stochastinio funkcinio nesklandumų šalinimo versija sutrikdo gradientą kartu su triukšmo periodu proporcingai Hesano paviršiui. ( an on-line ver-sion of stochastic functional smoothing perturbs the true gradient with a noise term proportional to the Hessian of the performance surface.) Kuomet signalas pridedamas prie norimo signalo, tikrinis (true) gradientas taip pat yra paveikiamas triukšmų periodo. Šiuo atveju poveikis yra proporcingas naujos veikimo funkcijos gradientui, kuris gaunamas iš originalaus skirtumo tarp d(t) ir y(t), kartu su įvestu triukšmu. Šis paviršius yra susijęs su originaliu, bet gali ir žymiai skirtis. Taigi, kuomet triukšmų šaltinis paprastai yra nustatomas į nulinę Gauso reikšmę, galima tikėtis mažiau optimalių rezultatų, lyginant su stochastinės funkcijos lyginimu (stochastic functional smoothing.). Tačiau algoritmo paprastumas ir geras veikimas gautas eksperimentuose skatina toliau dirbti prie šio metodo. Priedas Šio priedo tikslas yra susieti triukšmo pridėjimą prie norimo signalo naudojant stochastinį funkcinį lyginimo metodą, kuris yra globali optimizacijos procedūra. A.I Stochastinio funkcinio lyginimo optimizacijos apžvalga Stochastiniame funkciniame lyginime, originali neišgaubta funkcija yra perkeliama pagalbinės lyginimo funkcijos, kuri turi kai kurias optimizavimo savybes (t.y. atskiras minimumas). Dirbant su lyginimo funkcija, gali būti atsrastas optimalios problemos globalus minimumas . Lyginimo kainos funkcijos klasė parametrizuota ß yra apibrėžiama kaip [Rubinstein, 1981 and 1986] Kur ß yra valdymo parametras, o y yra atsitiktinis dydis. Dėl J ˆ (wtam, kad būti naudingam originaliai optimizacijai, h ˆ (v impulso atsakymas turi tenkinti keleta sąlygų [žr. Rubinstein, 1981 ir 1986 detaliau], taip, kad parametras apsprendžia lyginimo taikomo J(w) laipsnį. Dideliam lyginimo poveikis yra didelis ir atvirkščiai. Kuomet  0 J ˆ () = J , tuomet nėra lyginimo. Intuityviai aišku, kad norint išvengti lokalaus minimumo, optimizacijos pradžioje  turi būti pakankamai didelis. Tačiau siekiant optimumo lyginimo efektyvumas turi būti mažinamas leidžiant ß artėti prie nulio. Taigi minimumo taške w* laukiamas sutapimas tarp J(w) ir J ˆ ( Atitinkamai, konstruojant iteratyvią w* paieškos procedūrą, yra reikalinga lyginimo funkcijų aibė J ˆ(ß s=1,2,...... Jei signalo atsakymo dalis yra išrenkama kaip daugianormalinė funkcija su dydžiu n ir kitimu ß, tai Lyginimo kainos funkcijos gradientas gali būti įvertintas taip [Styblinski and Tang, 1990] Kur N yra pavyzdžių su daugybe kintamųjų iš (23) Lygties skaičius. Taigi, lyginimo kainos funkcijos gradientas gali būti randamas iš originalios kainos funkcijos. A.II Tiesioginė stochastinės lyginimo funkcijos Optimizavimo realizacija Susiejami triukšmo norimame signale poveikis ((10) ir (11) lygtys) su tiesiogine lyginimo funkcinių gradientų realizacija ((24) Lygtis)). Bus taikoma stochastinės aproksimacijos savybė [Robbins and Monro, 1951]. Kuomet ISE aproksimuoja (stochastine prasme) į MSE ir gradiento operatorius yra tiesinis operatorius, lyginimo kainos funkcijos gradientas εˆ gali būti įvertintas pagal analogiją su Lygtimi (25), taip Dėl supaprastinimo, čia yra ignoruojamas diskretinio laiko indeksas t. Reiktų pabrėžti, kad iš L.(24), kuri atspindi originalios stochastinės lyginimo funkcijos optimizacijos artėjimą prie L.(26), kuri yra tiesioginis L.(24) įvertis, tik vienintelė stochastinės aproksimacijos savybė buvo taikoma taip, kad būtų garantuotas tiesioginio įvertinimo stabilumas [Robbins and Monro, 1951; Kusher and Calrk, 1978; Wang and Principe, 1995]. Tiesioginis vienpusisi įvertis naudojamas L.(26) yra pagrindas gradiento įverčio naudojamo LMS ir BP algoritmuose. L.(26) išreiškia įvertinimą ε(w) gradiento, kuomet w yra paveikiamas atsitiktinio kintamojo βv j . Šis metodas praktiniam realizavimui yra per brangus, kadangi svoriai turi būti veikiami (gradiento skaičiavimui pageidaujamas antras tinklas). Taigi, šis metodas tiesiogiai nėra įgyvendinamas. Realizacijos supaprastinimui siūloma atlikti Teiloro seriją ekspansijų apie w, ir antrame etape jį suskaidyti. Literatūra 1. Darken, C., Chang, J., and Moody, J., “Learning Rate Schedules for Faster Stochastic Gradient Search,” IEEE Neural Networks for Signal Processing, 1992. 2. Fahlman, S., “Fast-Learning Variations on Back-Propagation: An Empirical Study,” In Proc. Of 1988 Conn. Model Summer School. 3. Feller, W, An Introduction to Probability Theory and Its Applications, Vol. 1, 2rd ed. Wiley, NewYork, 1966. 4. Haykin, S, Neural Networks---A Comprehensive Foundation, Macmillan College Publishing Company, New York, 1994. 5. Hertz J., Krogh A., Palmer R. G., “Introduction to the theory of neural computation,” Addison-Wesley,1991. 6. Hinton G. E., “Connectionist learning procedure,” In machine learning: Paradigms and methods, J. G. Carbonell, ed., pp. 185-234. MIT Press, Cambridge, MA, 1989. 7. Holmstrom L., and Koistinen, P., “Using Additive Noise in Back-Propagation Training,” IEEE Trans. on Neural Networks, Vol. 3, No.1, 1992. 8. Kirkpatrick, S., et. al., “Optimization by simulated annealing,” Science 220, 671-680. 9. Krogh, A. and Hertz, J., “Generalization in a Linear Perceptron in the Present of Noise,” J. Phys. A: Math. Gen. 25(1992) 1135-1147. 10. Kushner, H., “Asymptotic Global Behavior for Stochastic Approximation and Diffusions with Slowly Decreasing Noise Effects: Global Minimization via Monte Carlo,” SIAM J. APPL.MATH. Vol. 47, No. 1 Feb., 1987. 11. Kushner, H, and Clark, D. S., Stochastic Approximation Methods for Constrained and Uncon-strained Systems, Springer-Verlag, New York, 1978. 12. Matsuoka, K., “Noise Injection into Inputs in Back-Propagation Learning,” IEEE Trans. Systems, Man, and Cybernetics, Vol. 22, No. 3, 1992. 13. Richard M., Lippmann R. P., “Neural network classifiers estimate Bayesian a posteriori probabil-ity,” Neural Computation, 3, 461-483, 1991. 14. Robbins, H., and S. Monroe, “A stochastic approximation method,” Annals of Mathematical Sta-tistics 22, 1951. 15. Rognvaldsson, T., “On Langevin Updating in Multilayer Perceptrons,” Neural Computation, 6.916-926, 1994. 16. Rubinstein, R., Simulation and the Monte Carlo Method, Wiley,1981. 17. Rubinstein, R., Monte Carlo Optimization, Simulation and Sensitivity of the Queueing Networks,Wiley, 1986. 18. Rumelhart et al, Parallel Distributed Processing, Vol.1, MIT Press, 1986. 19. Styblinski, M.A., and Tang, T.-S, “Experiments in Nonconvex Optimization: Stochastic Approxi-mation with Function Smoothing and Simulated Annealing,” Neural Networks, Vol.3, 1990. 20. Szu, H., “Fast simulated annealing,” AIP conf. Proc. 151:Neural Networks for Computing, Snow-bird, UT, 1986. 21. Waibel, A., T. Hanazawa, G. Hinton, K. Shikano, K. J. Lang, “Phoneme recognition using time-delay neural networks,” IEEEE Trams. ASSP-37, 1989. 22. Wang, C., and J. C. Principe, “On-line stochastic functional smoothing optimization for neural network training, submitted to Neural Networks, 1995. 23. Werbos, p., “Generalization of backpropagation with application to a recurrent gas market model,” Neural Networks, 1, 339-356. 24. Widrow, B., and Hoff, M., “Adaptive switching circuits,” IRE WESCON Convention Record, pp.96-104, 1960.

Informatika  Kursiniai darbai   (164,25 kB)

'I never saw a dog with a highfalutin' name that ever was worth a rap,' he said, as he concluded his task and shoved her aside. 'They just fade away and die under the responsibility. Did ye ever see one go wrong with a sensible name like Cassiar, Siwash, or Husky? No, sir! Take a look at Shookum here, he's Snap! The lean brute flashed up, the white teeth just missing Mason's throat. 'Ye will, will ye?' A shrewd clout behind the ear with the butt of the dog whip stretched the animal in the snow, quivering softly, a yellow slaver dripping from its fangs. 'As I was saying, just look at Shookum here- he's got the spirit. Bet ye he eats Carmen before the week's out.' 'I'll bank another proposition against that,' replied Malemute Kid, reversing the frozen bread placed before the fire to thaw. 'We'll eat Shookum before the trip is over. What d'ye say, Ruth?' The Indian woman settled the coffee with a piece of ice, glanced from Malemute Kid to her husband, then at the dogs, but vouchsafed no reply. It was such a palpable truism that none was necessary. Two hundred miles of unbroken trail in prospect, with a scant six days' grub for themselves and none for the dogs, could admit no other alternative. The two men and the woman grouped about the fire and began their meager meal. The dogs lay in their harnesses for it was a midday halt, and watched each mouthful enviously. 'No more lunches after today,' said Malemute Kid. 'And we've got to keep a close eye on the dogs- they're getting vicious. They'd just as soon pull a fellow down as not, if they get a chance.' 'And I was president of an Epworth once, and taught in the Sunday school.' Having irrelevantly delivered himself of this, Mason fell into a dreamy contemplation of his steaming moccasins, but was aroused by Ruth filling his cup. 'Thank God, we've got slathers of tea! I've seen it growing, down in Tennessee. What wouldn't I give for a hot corn pone just now! Never mind, Ruth; you won't starve much longer, nor wear moccasins either.' The woman threw off her gloom at this, and in her eyes welled up a great love for her white lord- the first white man she had ever seen- the first man whom she had known to treat a woman as something better than a mere animal or beast of burden. 'Yes, Ruth,' continued her husband, having recourse to the macaronic jargon in which it was alone possible for them to understand each other; 'wait till we clean up and pull for the Outside. We'll take the White Man's canoe and go to the Salt Water. Yes, bad water, rough water- great mountains dance up and down all the time. And so big, so far, so far away- you travel ten sleep, twenty sleep, forty sleep'- he graphically enumerated the days on his fingers- 'all the time water, bad water. Then you come to great village, plenty people, just the same mosquitoes next summer. Wigwams oh, so high- ten, twenty pines. Hi-yu skookum!'

Anglų kalba  Pagalbinė medžiaga   (140,32 kB)

Last year teenagers committed about 535 crimes. During one-year period delinquency raised 16.6%. Biggest part of crimes was committed by teenagers aged from 13 to 19. Thefts from cars are 42.3% and burgalyries-31.5% off all committed crimes. Every 6th crime is burglary. Films of violence, detailed crime stories in the press have a big influence for crimes increasing. In 1998 investigated 47 teenagers’ burglaries in Siauliai, this year, after 4 months - 28. 22 of them were investigated. Comparing with last year Siauliai has 46.7% increases. Dogging adult’s steps teenagers begin extort wealth, cheat, make drugs, use guns, process money, resell burglaries things. Statistic shows that drunk or intoxicated teenagers made many crimes. From 615 criminals 249 are pupil from secondary school. 53% guilty juveniles don’t study or work. We can group teenager criminals into two groups. One group of them become criminals, because those teenagers are weakling persons, their friends make great influence on them on their way of thinking or by these friends help they do a crime for fun. Other group of teenager’s criminals does crimes for their bad social status. How a teenager can become a criminal? Teenager can become a criminal when: • This teenager’s friends make great influence on him on his way of thinking. • This teenager is a weakling person and he can’t resist the temptation to alcohol, drugs, so he does a crime, because at that moment he did not understand what he was doing, because he was drunk. • This teenager does not have what to do in his spare time, so he does a crime just for having fun. • This teenager’s social status is bad, so he does a crime for having money. What kind of teenager criminals are in Lithuania? A teenager criminal can be: • vandal (a person who likes to draw on the cars, walls, houses, who likes to brake something); • filches (some kind of stealer); • pilferer (some kind of stealer); • pugnacious person (a person who likes to fight against somebody); • burglar (a person who steals from the houses); • rapist (a person who likes to rape women); • racketeer (a person who orders another person to give all his money); As we all know the bigger part of teenager criminals are of male sex. And we also know that a teenager criminal is not so dangerous like a professional criminal, who has got lots of experience in that sphere. And that a teenager criminal’s way of life could be easily changed to another way of life, normal way of life, just you have to show that there is another way of living. Police account Why do youngsters become criminals? It’s the question, which bothers a lot of people. Here are some reasons why that happen: Youngsters don’t have interesting facilities and hobbies These are the main things why youngsters become criminals. Now we want to tell some ideas how to solve this problem. Should be some educational centers where young people could find a professional psychologist that would help a lot. Schools should try to help solve that problem and organize some lectures for students about crimes, drugs, how drugs can make people do very bad things. We were explaining how to solve that problem, but we forgot to tell what kinds of crimes are most popular. There are a lot of hooligans, but it isn’t the biggest problem in our country. They have a lot of problems with muggers, because they are getting money like that for drugs and then they start feeling bad and start robbing (old ladies), stealing or even burgling. That makes a lot of problems for police officers. And the other kind of crimes is shoplifting (that is the most popular kind of crimes) Very many shops loose a lot of money, because of that. And the main thing with shoplifters is that they get used to it and become addict. We think you want to ask why police isn’t doing anything about that. But they do. They try to organize some summer caps for youngsters try to take them to psychologist or to talk with them; some times they organize shows for pupil. So I think you can’t say that police is doing nothing.

Anglų kalba  Kalbėjimo temos   (6,05 kB)

To begin with, all people use various services: postal services, telephone, telegraph, bank, police, diplomatic services, car maintenance services, petrol stations and personal services. Personal service establishments are very useful and many people visit them every day. I`m not an exception . When I am going to go to a party I go to hairdresser’s establishment. If I need my shoes to get mended, I go to the repair shop. When they are rather worn, I take them to the shoemaker’s. Occasionally I need to have my photos taken. Then I go to the photographer’s. In addition I have a camera at home so I can take photos myself. Unfortunately I can’t make photographs, so I have to go to the photographer’s and have them made. As far as I can see the telephone is widespread nowadays. The telephone is of great importance in business, in administrative relations and in the national economy. I have the telephone at home too and I like to talk over it. When necessary, the police, the fire brigade, the first aid and ambulance service can be called by anybody from any telephone. It is easy of access. You have to dial only two digits. We use the fire brigade. For example, once there was a strange smell and smoke from our neighbours door. Such being the case, we call a fire brigade. Fortunately for us it was only burnt stewpot [‹stju:pƒt]. What comes to postal service, I use it too. I like to communicate with people. I have some friends in Lithuania and Denmark, and I communicate with them. I send letters to Denmark by air mail. It costs more but it is faster. At the post office you can send telegrams too. I can send parcels with gifts to my friends when they celebrate their birthdays. A few years ago I used to go to café-internet. There you can Relax with a cup of tea or Espresso while you Surf the Net. Now I give preference to my computer at home. Finally sometimes I use medical services. If I am not ill very seriously I treat myself. However, a few years ago I broke my hand so for this reason I had to use services of medical professions. Another reason why I go to see the doctors is medical certificate for school. Once a half-year I go to the dentist. On the whole I havn`t a fear of various medical institutions like dentist and so on. I like listening comments and discussions on current affairs, interviews with various people on the radio. The programs are followed by music it`s the main reason why it interests me. I like to watch something on TV in the evenings. When my radio or TV has broken down I have to repaired it in the service shop. And when I was going to go abroad I used the diplomatic services. In that case I went to Vilnius embassy for my visa. In conclusion, all services are in your interest. Service industry provide a variety of facilities to help in every situation. Moreover it`s easy of access. And to my mind it make our life easier.

Anglų kalba  Pagalbinė medžiaga   (108,32 kB)

To begin with, all people use various services: postal services, telephone, telegraph, bank, police, diplomatic services, car maintenance services, petrol stations and personal services. Personal service establishments are very useful and many people visit them every day. I`m not an exception . When I am going to go to a party I go to hairdresser’s establishment. If I need my shoes to get mended, I go to the repair shop. When they are rather worn, I take them to the shoemaker’s. Occasionally I need to have my photos taken. Then I go to the photographer’s. In addition I have a camera at home so I can take photos myself. Unfortunately I can’t make photographs, so I have to go to the photographer’s and have them made. As far as I can see the telephone is widespread nowadays. The telephone is of great importance in business, in administrative relations and in the national economy. I have the telephone at home too and I like to talk over it. When necessary, the police, the fire brigade, the first aid and ambulance service can be called by anybody from any telephone. It is easy of access. You have to dial only two digits. We use the fire brigade. For example, once there was a strange smell and smoke from our neighbours door. Such being the case, we call a fire brigade. Fortunately for us it was only burnt stewpot [‹stju:pƒt]. What comes to postal service, I use it too. I like to communicate with people. I have some friends in Lithuania and Denmark, and I communicate with them. I send letters to Denmark by air mail. It costs more but it is faster. At the post office you can send telegrams too. I can send parcels with gifts to my friends when they celebrate their birthdays. A few years ago I used to go to café-internet. There you can Relax with a cup of tea or Espresso while you Surf the Net. Now I give preference to my computer at home. Finally sometimes I use medical services. If I am not ill very seriously I treat myself. However, a few years ago I broke my hand so for this reason I had to use services of medical professions. Another reason why I go to see the doctors is medical certificate for school. Once a half-year I go to the dentist. On the whole I havn`t a fear of various medical institutions like dentist and so on. I like listening comments and discussions on current affairs, interviews with various people on the radio. The programs are followed by music it`s the main reason why it interests me. I like to watch something on TV in the evenings. When my radio or TV has broken down I have to repaired it in the service shop. And when I was going to go abroad I used the diplomatic services. In that case I went to Vilnius embassy for my visa. In conclusion, all services are in your interest. Service industry provide a variety of facilities to help in every situation. Moreover it`s easy of access. And to my mind it make our life easier.

Anglų kalba  Rašiniai   (4,74 kB)

Golden Rose str. 15, P.O. Box 2312, London Dear Carla, Thanks a lot for your letter and sorry for not writing before – I was soooooo busy! You know, school, music, friends and so on.:) So, you decided to come to my country – it’s so great! I can’t wait for you to come! You said that you have no ideas what places to visit and where to go. Well, I have some. First, I recommend you to stay in capital – Vilnius – or in Kaunas. It is the second largest city in Lithuania and it’s where I live. :P Next – if you want to make an acquaintance with our culture you’d better stay in Vilnius. There you can visit lot great places. For example – Gediminas Castle, or a Television Tower with a valiant (and expensive, of course) cafe “The Milky Way”, from where you can see a view of town. Also the Royal Palace of Lithuania and amazing St. Ona church which is inflexible fascinating. And because of that Napoleon wanted to take it on him palm to France! If you’ll decide to stay in Kaunas, you should visit Liberty Avenue which is a symbol of town. But because of that, that in it is now very much shoe shops it’s sometimes called “Shoes Avenue”. We also have Kaunas Castle, a lot of museums, Nemunas and Neris confluence, which is very beautiful place, and ctr. So, I promise – you won’t be bored. Can’t wait for you to come, Love, Emily

Anglų kalba  Pagalbinė medžiaga   (2,79 kB)

They serve as a useful means of communication. How could we understand or talk to people from other countries, if we couldn’t speak their language? I think that some people like learning foreign languages, various people have different reasons. I think that some people like learning languages in general, the others learn because they have to do that at school. Besides, nowadays pupils realize quite well that they might need foreign languages in the nearest future or it might be a credit for them to get a better job.

Anglų kalba  Kalbėjimo temos   (2,5 kB)

Touching some other bases, I must say that there are many reasons of having a pet. The most important is loneliness. However I can’t deny that some people keep pets for fun. There are strong arguments both for and against keeping a pet. I would like begin with advantages of owning a pet. The first reason of having a pet is that their benefits for children are undeniable. Keeping a pet is an important part of every child’s childhood and plays a big role in both their physical and mental development. It teaches children how to play, look after them, while at the same time aiding tolerance and understanding of each other. From a social standpoint, pets play a large role in many lonely people’s lives. Nowadays more than ever lonely people are fond of keeping pets, because it helps to raise their mood, when they are in low spirits, and not to feel so extremely lonely. The final advantage of having a pet is that it makes fun for people who are crazy about attending various animals’ exhibitions and showing the best points of their pets to others. However, there are many disadvantages of keeping a pet as well. Perhaps the most obvious one is that pets cause a lot of damage. In fact, a lot of people face with this problem. To avoid that they always have to look after their pet and try not to keep them in rooms which are full of valuable things. Secondly, some people can’t keep pets because of their health. Usually they are allergies and can not live with a pet in the same place. Furthermore, some pets, especially dogs, dislike little children and they are likely to bite them. So it is advisable to think twice before buying a pet and consider both advantages and disadvantages. Talking about pets, I am able to say that I have a pet too. I have big, fat cat. He is called Meilas. He is Persian race. All my family’s life changed when we got this cat. This cat brought to our lives more fun. I feel very happy when Meilas meets me coming back from school. He is my best friend. To sum it up, I must say that if you love your pet, he makes you life as fun as you can imagine.

Anglų kalba  Kalbėjimo temos   (3,48 kB)

Regulations for checking and preventing fires were developed. In the reindustrialize era most cities had watchmen who sounded an alarm at signs of fire. Fire fighting tool was simple water. Another important fire-fighting tool was the axe, used to remove the fuel and prevent the spread of fire as well as to make openings that would allow heat and smoke to escape a burning building. Following the Great Fire of London in 1666, fire brigades were formed by insurance companies. The government was not involved until 1865, when these brigades became London's Metropolitan Fire Brigade. The first modern standards for the operation of a fire department were not established until 1830, in Edinburgh, Scotland. These standards set out, for the first time, what was expected of a good fire department. In all industrial countries fire fighters undergo training, beginning with probationary fire fighters' school and continuing throughout a fire fighter's career. Great Britain has several fire training centers. In Russia, fire schools are in Moscow and St Petersburg; Sweden and Denmark have similar schools. Most fire fighting consists of applying water to the burning material to cool it. Fires involving flammable liquids, certain chemicals, and combustible metals often require special extinguishing agents and techniques. With some fuels the use of water may actually be dangerous. Now fire fighters use special technique. Their uniform is made from special material that protects from heat and poisonous gas they have masks and if there is people in the burning house they have oxygen masks for them. This work is very important because fire can make a lot of damage.

Anglų kalba  Kalbėjimo temos   (2,55 kB)

Regulations for checking and preventing fires were developed. In the reindustrialize era most cities had watchmen who sounded an alarm at signs of fire. Fire fighting tool was simple water. Another important fire-fighting tool was the axe, used to remove the fuel and prevent the spread of fire as well as to make openings that would allow heat and smoke to escape a burning building. Following the Great Fire of London in 1666, fire brigades were formed by insurance companies. The government was not involved until 1865, when these brigades became London's Metropolitan Fire Brigade. The first modern standards for the operation of a fire department were not established until 1830, in Edinburgh, Scotland. These standards set out, for the first time, what was expected of a good fire department. In all industrial countries fire fighters undergo training, beginning with probationary fire fighters' school and continuing throughout a fire fighter's career. Great Britain has several fire training centers. In Russia, fire schools are in Moscow and St Petersburg; Sweden and Denmark have similar schools. Most fire fighting consists of applying water to the burning material to cool it. Fires involving flammable liquids, certain chemicals, and combustible metals often require special extinguishing agents and techniques. With some fuels the use of water may actually be dangerous. Now fire fighters use special technique. Their uniform is made from special material that protects from heat and poisonous gas they have masks and if there is people in the burning house they have oxygen masks for them. This work is very important because fire can make a lot of damage.

Anglų kalba  Kalbėjimo temos   (3,58 kB)

Emotions cause not only general reactions, but specific ones as well. We may laugh when happy, withdraw when frightened, get aggressive when angry, and so forth. Among these typical emotional reactions, psychologists have singled out one in particular for extensive study: aggression. The components of an emotion include autonomic arousal (sužadinimas), cognitive appraisal (įvertinimas), and emotional expression. Intense emotions usually involve physiological arousal caused by activation of the autonomic nervous system. People who have spinal cord injuries, report experiencing less intense emotions. Most people report getting angry at least several times a week often at loved ones. Though they commonly feel like aggressing physically when angered most control these impulses. Frustration and pain (mental and physical) arouse anger and can stimulate aggression. Incentives may also trigger aggression. Aggression is a typical reaction to anger (though it can occur for other reasons as well). According to early psychoanalytic theory, aggression is a frustration-produced drive; according to social-learning theory, aggression is a learned response. Biology gives animals the capacity to hurt one another. The threshold levels of numerous aggressive brain systems are thought to be influenced by heredity other neutral circuits, blood chemistry, neurotransmitters, and experience. Cultures that sanction aggression have high rates of it. Families teach aggression directly and indirectly. When treated harshly, children pick up the same habits. Other contributers to aggression include school failures and frustrations, anonymity, poverty, and the availability of weapons. Several techniques for the control of human aggression exist. These include punishment, catharsis, exposure to nonaggressive models, and training in basic social skills. In addition, aggression can often be reduced though the induction of responses or emotional states incompatible with such behavior.

Anglų kalba  Pagalbinė medžiaga   (33,22 kB)

Emotions cause not only general reactions, but specific ones as well. We may laugh when happy, withdraw when frightened, get aggressive when angry, and so forth. Among these typical emotional reactions, psychologists have singled out one in particular for extensive study: aggression. The components of an emotion include autonomic arousal (sužadinimas), cognitive appraisal (įvertinimas), and emotional expression. Intense emotions usually involve physiological arousal caused by activation of the autonomic nervous system. People who have spinal cord injuries, report experiencing less intense emotions. Most people report getting angry at least several times a week often at loved ones. Though they commonly feel like aggressing physically when angered most control these impulses. Frustration and pain (mental and physical) arouse anger and can stimulate aggression. Incentives may also trigger aggression. Aggression is a typical reaction to anger (though it can occur for other reasons as well). According to early psychoanalytic theory, aggression is a frustration-produced drive; according to social-learning theory, aggression is a learned response. Biology gives animals the capacity to hurt one another. The threshold levels of numerous aggressive brain systems are thought to be influenced by heredity other neutral circuits, blood chemistry, neurotransmitters, and experience. Cultures that sanction aggression have high rates of it. Families teach aggression directly and indirectly. When treated harshly, children pick up the same habits. Other contributers to aggression include school failures and frustrations, anonymity, poverty, and the availability of weapons. Several techniques for the control of human aggression exist. These include punishment, catharsis, exposure to nonaggressive models, and training in basic social skills. In addition, aggression can often be reduced though the induction of responses or emotional states incompatible with such behavior.

Anglų kalba  Kalbėjimo temos   (3,45 kB)

Computer – it`s a machine for carrying out calculations and performing specified transformations on information, such as storing, sorting, correlating, retrieving and processing data. The main disadvantage of computers is that staring at a screen for long periods of time can be damaging to the eyes, and sitting on a chair for hours at a time is certainly not healthy. Secondly computers distract from social interactions such as conversation. Also, people can be inclined to become anti-social and stay at home and use their computer. Finally, the most persuasive argument against the use of computers is that the more jobs are done by computers the less are done by people. It isn`t very good for increasing people population. However, the advantages of computers are numerous, such as undeniable educational benefits, especially to children. School subjects become more interesting when presented on a computer screen. Moreover, computers can be fun with a seemingly endless variety of games which can be played on them. In addition, computers are valuable to any business, making life easier and saving time by being capable of storing and retrieving vast amounts of information at he touch of the button. I think computer has a lot of disadvantages and advantages. Yet, despite the health problems, risk to jobs and personal interaction, it seems that as long as the use of computers is regulated the benefits computers provide to education and businesses are invaluable. Computers make our life easier.

Anglų kalba  Kalbėjimo temos   (3,6 kB)

I take a very difficult exam. I have to study for about one week. But I do not. I read a bit every day and I feel calm, because I know I will be able to cheat. Everybody knows that, but some of my group mates study a lot. They study for themselves, for the future or because they are really interested in that subject. I am not interested at all. After the exam we find out that all students got good marks. But somebody is not satisfied or jealous, so he goes and tells the lecturer the names of those people who cheated. He tells my name too. I am very upset- not because I have cheated, but because I can trust nobody in my group. How could I be so stupid and tell that man such a simple thing… Cheating is a very popular phenomenon in our country. In other countries the prevalence of cheating differs. And it is a problem of educational system, not of students. When the disciplines are interesting and useful, students are fond of learning, not of cheating. But the worst thing is that nobody can change this system rapidly: it requires much time, money and hard work. Firstly, honors code like Groveton’s ruins the relationships between students. A student can not trust anybody and tell that he is going to cheat. That is, he has to lie to his mates that he knows a lot, that he has studied for weeks. And it is very difficult (I would say impossible) to find real friends from the university environment. For example, I would not call a friend such a man, to whom I have to lie, who can lodge a complaint against me after every test or exam. The most important thing in friendship is trust and freedom to talk, share impressions. Sometimes it is so great to tell a real friend: “That was the best cheating in my life…” In our society lots of students’ best friends are from their course. If students would have to sign an honor code, the relationships between them changed a lot. There could be no talk about marriages between group mates, because what is a marriage without trust and supporting a spouse? Or there could be another way out of such situation: to break an oath and to lodge complaints only against some students. But on the other hand, if one breaks an oath, there is no need to report about cheating at all. So what should a poor student do? Another difficulty is that the information can be not reliable. If one student hates somebody, it is natural that he tries to do harm to the object of hate. And he can say that he is cheating after every exam and test. Karen Horney says that some people are inclined to competitiveness. Such people try to damage a competitor in order to enhance their own position or glory or to keep down a potential rival. For such a man it is more important to see others defeated than to succeed himself. In universities it is natural that they would try to defeat their mates (competitors) in the easiest way- they will tell that others have cheated. And it is very difficult to check such information. I think that people can not be forced to sign such documents as honor codes, because nobody can make disgraceful people become honorable. The ones who are honorable will not cheat without signing any paper, and the ones who want to cheat, will pay no attention to what they have signed. I think that forcing to sign an honor code would insult me a bit, because I can be honorable without a code, and it is nobody’s business. It is up to me and my conscience to decide whether I will cheat or not. University is not a secondary school, where students have to learn lots of needless subjects. Almost everything that is taught at universities is necessary for the future job. Students choose a profession to study, which interests them, and so they are fond of learning and knowing a lot.

Anglų kalba  Kalbėjimo temos   (4,38 kB)

PRESENT SIMPLE PRESENT CONTINOUS PAST SIMPLE PAST CONTINOUS SUDARYMAS Veiksmažodžio 1 forma Veiksmažodžio be forma ir veiksmažodis su galūne -ing Veiksmažodžio 2 forma arba veiksmažodis su –ed galūne Was/Were ir veiksmažodis su galūne -ing Pvz. Teigiamam sakiny; I play; You play; He plays; I am playing; You are playing; He is playing; I/You/he/etc walked I was studying; You were studying; He was studying. Neigiamam sakiny; I/You don‘t play; he doesn‘t play I am not playing; You are not playing; He is not playing; I/You/he/etc did not walk I wasn‘t studying; You weren‘t studying; He wasn‘t studying. Klausime; Do I/You play? Does he play? Am I playing? Are You playing? Is he playing? Did I/You/he/etc walk? Was I studying? Were you studying? Was he studying? NAUDOJIMAS • Kasdienei rutinai; (I catch the bus to work at 7 o‘clock every day ) • Pasikartojantiems veiksmams (he cleans the house every week ) • Įpročiams, pomėgiams (She usually stays at home on Sundays) • Pastoviai būsenai (He works in a bank) • Dieneotvarkei, tvarkaraščiams (The train leaves at 7 o‘clock) • Bendros, neginčijamos tiesos, gamtos dėsniai (water boils to 100C degrees) • Sporto, knygos komentarams, peržiūrai, pasakojimui. • Veiksmui, kuris vyksta dabar, tuo metu kai kalbama. (She is talking on the phone now) • Veiksmui, kuris vyksta šiomis dienomis (Peter is staying at his friend‘s these days) • Netolimoje ateityje suplanuotiems veiksmams; (I am flying to Paris tomorrow) • Situacijoms, kurios nėra pastovios ir keičiasi; (The climate is becoming warmer and warmer) • Per dažnai pasikartojantiems veiksmams, kurie erzina. Su žodeliais allways, constantly, continuously. (She is always talking in the lessons) • Veiksmui, kuris vyko nustatytu laiku praeityje. (Bob bought a car last month) • Veiksmams, kurie įvyko vienas po kito praeityje. (First he took a shower and then he went to bed.) • Praeities įpročiams Tam galima naudoti ir used to. (My grandmother read/used to read me bedtime stories when I was little) • Veiksmui, kuris vyko žinomu laiku praeityje. Mes nežinome kada jis prasidėjo ir kada baigėsi. (At eight o‘clock yesterday evening Tom and Mark were doing their homework) • Praeities veiksmui, kuris vyko kai kitas veiksmas jį pertraukė. Veiksmui, kuris vyko naudojame Past Continous, o veiksmui, kuris pertraukė- Past Simple. (He was exercising when he felt a sharp pain in his lower back) • Dviems ar daugiau veiksmų, kurie vyko tuo pačiu metu praeityje. (sarah was watching TV while I was reading a novel) • Suteikti pasakojimui „foną“, prieš aprašant pagrindinius faktus. (The band was playing and the people were singing and dancing when we arrived in the concert.) ŽODELIAI Every; at noon; at night; always; usually; often; sometimes; seldom; never. Now; at the moment; at present; these days; tomorrow; today; tonight; etc. Yesterday; yesterday morning/evening/etc. last; ago, in 1984. While; when; as; all day/night/morning; all day yesterday; etc. PRESENT PERFECT PRESENT PERFECT CONTINOUS PAST PERFECT PAST PERFECT CONTINOUS SUDARYMAS Have ir veiksmažodžio 3 forma arba veiksmažodis su galūne -ed Have/has been ir veiksmažodis su galūne -ing Had ir veiksmažodžio 3 forma, arba veiksmažodis su –ed galūne. Had been ir veiksmažodis su –ing galūne. Pvz. Teigiamam sakiny; I/you have booked; He has booked; I/You have been cycling; He has been cycling. I/You/had started I/You/he had been listening. Neigiamam sakiny; I/you haven‘t booked; He hasn‘t book. I/You haven‘t been cycling; He hasn‘t been cycling. I/You/he hadn‘t started I/You/he hadn‘t been listening. Klausime; Have I/you booked? Has he booked? Have I/You been cycling? Has he been cycling? Etc. Had I/You/he started? Had I/You/he been listening? NAUDOJIMAS • Veiksmams, kurie prasidėjo praeityje, ir tęsiasi dabar. (I have been techer for five years) • Praeities veiksmams, kurių rezultatus matome dabartyje.(He has washed his car) • Veiksmams, kurie įvyko praeityje, nenustatytu laiku. Veiksmas yra svarbesnis už laiką. (she has sold her house) • Petyrimams (have you ever slept in a tent?) • Veiksmams, kurie vyko tam tikru laiko periodu ir šis periodas dar nėra pasibaigęs tuo metu kai kalbama. (He has visited three museumsthis morning. Morning is not over yet.) • Veiksmams, kurie prasidėjo, kurįlaiką tęsėsi ir baigėsi praeityje. Rezultatas matomas dabartyje. (They have been sunbathing) • Veiksmams, kurie prasidėjo praeityje ir tęsiasi dar ir šiuo metu. (She has been cooking for two hours.) • Išreikšti pasipiktinimui, susierzinimui (Who has been using my computer?) • Veiksmui, kuris įvyko prieš kitą veiksmą praeityje. (She had painted the room by 5 o‘clock. (She finished painting before 5 o‘clock)) • Veiksmui, kuris įvyko prieš kitą būtąjį laiką. (She had booked a hotel before she arrived in Paris) • Veiksmui, kuris prasidėjo ir pasibaigė praeityje, ir rezultatas buvo matomas praeityje. (He had broken his arm, so he couldn‘t write) • Pabrėžti tąsą veiksmo, kuris prasidėjo ir pasibaigė praeityje.(She had been playing tennisfor ten years by the time she was eighteen.) • Pabrėžti tąsą veiksmo, kuris pasibaigė prieš kitą praeities veiksmą. (He had been studying for seven years before he got his degree.) • Veiksmui, kuris kurį laiką tęsėsi praeityje, ir jo rezultatas buvo matomas praeityje. (She was tired because she had been cleaning the house all day.) ŽODELIAI Ever, never, already, yet, just, so far, for, since, recently, lately, before. Ever, never, already yet, just, so far, for, since, recently, lately, before. Before; after; already;just; till/untill; when;by...; by the time; never; etc. For; since; how long; before; until; etc. FUTURE SIMPLE FUTURE CONTINOUS FUTURE PERFECT FUTURE PERFECT CONTINOUS BE GOING TO SUDARYMAS Will ir bendratis (be to) Will be (am, is, are) ir veiksmažodis su –ing galūne. Will have ir veiksmažodžio 3 forma arba veiksmažodis su –ed galūne. Will have been ir veiksmažodžio 3 forma arba veiksmažodis su galūne –ed. Be going to NAUDOJIMAS • Ateities spėjimams, kai remiamės tuo, ką manome ar ko tikimės. (I believe Bill will get a job) • „Greitiems sprendimams“. (We‘ve run out of milk. I will go to the supermarket and get some.) • Pažadams, grąsinimams, įspėjimams, reikalavimams, viltims, pasiūlimams. • Veiksmams, atsitikimams, situacijoms, kurios būtinai įvyks ateityje, ir kurių negalima kontroliuoti. (Ben will be five years old in August.) • Veiksmams, kurie vyks nustatytu laiku ateityje. (I am going to Spain. This time next week I will be lying in the sun.) • Veiksmams, kurie tikrai įvyks ateityje kaip rezultatas kokios nors rutinos arba pasirengimų. • Kai klausiama mandagiai apie netolimos ateities planus, dažniausiai siekiant sau naudos. (Will you be driving to the party tonight? Would you be able to give me a lift?) • Veiksmui, kuris bus pasibaigęs iki tam tikro ateities momento. (They will have arrived in London by 5 o‘clock) • Pabrėžti tąsą veiksmo, kuris tęsėsi iki tam tikro momento ateityje. (By June, he will have been teaching in this school for fifteen years) • Ketinimams, planams, ambicijoms ateičiai. (I am going to go to India one day.) • Veiksmams, kuriuos mes jau nusprendėme atlikti netolimoje ateityje. (Nicky is going to visit her friends in London next week.) • Sprendimams (spėjimams), kurie remiasi tuo, ką matome ar žinome, ypač kai yra įrodymai, kad kažkas tikrai nutiks. (There isn‘t a cloud in the sky. It is going to be a beautiful day.) ŽODELIAI Think, believe, probably, certainly, parhaps, expect... etc. Before; by; by then; by the time;until/till By ... for Tomorrow; the day after tomorrow; tonight; soon; next week/month/year/sommer/etc; in a week/month/etc.

Anglų kalba  Paruoštukės   (10,67 kB)

American English speaking without a lot of effort How To Learn American English Speaking Easily? Everyone learns to speak their native language. Is there some things that kids are missing when they are emerging? Dr. J. Marvin Brown says that “both adults and children can do it right, but only adults can do it wrong.” So why we can not use the same approach with a second language? Why we can’t learn American English speaking like children? Easily and without a lot of effort. In fact we can. There are a lot of researches which shows that adults can learn languages even more easily and fast. But here‘s the thing. We need forget all that we were learning in school. Because it is the most important reason why we can not speak American English at all. That’s the reason why we don’t understand anything when American people are talking to each other. Yes, we can read American and English newspapers, but this language is not the same. The people are not speaking that way. Their speech is different. If we want to understand native American speakers and speak with them, we must find the different way. In 1984, the American University Language Center in Bangkok started using a new approach to teach the Thai language. The method is known as “Automatic Language Growth”. The first mistake they noticed is that adults are trying to speak in new language to early. They don’t have enough exposure to language yet and they are struggling and feel uncomfortable. They are bringing patterns from their first language and later they are speaking with a strong accent. Children in the beginning are just listening to the new language. They starts to speak when they are prepared. That’s the big difference. The second mistake adults make is that they are thinking, that they must study grammar. That’s a huge mistake. A vital mistake. Children do not studying grammar and that’s the key. If we want to learn American English speaking we must learn speaking. Not grammar. Learning grammar does not make you fluent. You must train your ears and you must train your mouth. So how we can learn speaking without grammar. Simply. First of all we must listen. Every day one or two hours. The more is the best. We must listen to easy and understandible content. That builds our confidence. And we must begin to speak when we are prepared only. When we can do this without struggling. It commes authomatically. Believe me. I have done this with Original Effortless English Lessons. It is totally new approach in learning English. It is real language. You can learn English speaking naturally. You can do the same. You can learn American English speaking just like I did.

Anglų kalba  Rašiniai   (4,61 kB)

Z. Oh, this party was pretty cool, but I and My friends were only in official part. I come home at 5 o’clock in the morning and what do you think about this party? M. Oh, this party was not bad but I was only in official part too and after this I with my class-mates went to the garden and celebrated this wonderful festival very well. But in my opinion to do the official part is a serious job. Z. Of course. Planning the 100-Day party was very hard. I and two my friends were responsible for the party. I had to invite primary school teachers. My friends were responsible for making speeches. M. Whose idea was to plan the funny part? Z. I think that this idea belongs to my friend Dainius. M. And who advised him? Z. I don’t now, but I think this part was original and fun. M. And who suggested doing the party decoration of the hall? It was nice. Z. Oh, it was Ramunas idea. But I think that it were a bit boring, because here were no disco. M. Oh, don’t worry, the party was wonderful.

Anglų kalba  Pagalbinė medžiaga   (3,08 kB)

The subjects of this presentation are the origins of American lexicography which will be represented through main problems, bilingual dictionaries, historical development of British and American lexicography, the great works of Noah Webster, also by the lexicographic influences of encyclopaedic dictionary by John Ogilvie and Funk and Wagnall’s standard dictionaries.

Anglų kalba  Referatai   (9 psl., 17,83 kB)

Paslaugos samprata. Produkto savybių ir vartotojų pasitenkinimo ryšys. Paslaugų ir prekių marketingo skirtumai. Paslaugų marketingo kompleksas. Vartotojų tipai ir pagrindiniai vartotojų elgsenos bruožai. Elektroninės bankininkystės situacija . Elektroninės bankininkystės teikiamas produktas. Telebanko paslaugos vartotojai ir teikėjai . Telebanko teikimo procesas ir fizinė paslaugos teikimo aplinka. Telebanko rėmimo kompleksas. Telebanko paslaugos teikimo vieta. Telebanko paslaugos kainodara. Telebanko paslaugos kokybės matavimo kriterijai.

Administravimas  Kursiniai darbai   (36 psl., 78,76 kB)

Cosmetics are substances used to enhance or protect the apperance or oder of the human body. Cosmetics include skin-care creams, lotions, powder, perfumes, lipstick, fingernail and toe nail polish, eye and facial make up, permanent waves, colored contact lenses, hair color, hair sprays and gels, deodorant, baby product, bath oils, bubble baths, bath salts, butters and many othertypes of products.

Anglų kalba  Pateiktys   (14 psl., 70,25 kB)

Dėsningumas tai ar atsitiktinumas, bet Lietuva tarpukario metais beveik neturėjo didžių mokslininkų bei kultūros veikėjų, galėjusių pradėti ir baigti darbus savo Tėvynėje, perduoti patirtį ir žinias ateinančiai kartai. Taip atsitiko ir daugeliui geografų. Užtenka čia prisiminti Joną Čerskį, Igną Domeiką, Steponą Kolupailą, kaip ir daugelį menininkų, rašytojų, politikų. Savo gyvenimus jie padalijo tarp Tėvynės ir svečių šalių.

Geografija  Referatai   (11 psl., 20,12 kB)

Tai recenzija Collen McCullough knygos "The Thorn Birds". Turinys atitinka visus reikalavimus. na manau toki pateikus mokytojai tikrai gausite gerą įvertinimą.

Anglų kalba  Pateiktys   (1 psl., 3,63 kB)

Vocabulary. Definitions of key vocabulary. The means of marketing. Advertising. Advantages and disadvantages of major advertising media. Formulating of the advertising message. The role of the advertising in the promotional mix. Advertising as persuasion. Marketing influences and often actually controls almost every part of company's activities. Everyone who works for the company must “think marketing”. To think marketing we must understand in the right way what is the marketing concept.

Rinkodara  Konspektai   (32 psl., 212,25 kB)

Here are facts showing the dangers of eating meat because of the large amounts of antibiotics fed to livestock to control staphylococci (commonly called staph infections), which are becoming immune to these drugs at an alarming rate. The animals that are being raised for meat in the United States are diseased. The livestock industry attempts to control this disease by feeding the animals antibiotics. Huge quantities of drugs go for this purpose. Of all antibiotics used in the U.S., 55% are fed to livestock.

Maistas, sveikata, higiena  Pagalbinė medžiaga   (4 psl., 8,18 kB)

Cocaine's recent notoriety belies the fact that the drug has been used as a stimulant by people for thousands of years. Its properties as a stimulant have led people in the past to use it in a number of patent medicines and even in soft drinks. But cocaine's highly addictive nature and addicts' willingness to pay a high price for the drug have propelled it into the public eye. The crime and violence associated with its transportation and sale, and the celebrity nature of some of its victims has kept cocaine in the news.

Chemija  Pagalbinė medžiaga   (1 psl., 3,9 kB)

Fire fighting, techniques used to extinguish fires and limit the damage caused by them. Fire fighting consists of removing one or more of the three elements essential to combustion — fuel, heat, and oxygen — or of interrupting the combustion chain reaction. The Roman emperor Augustus is credited with instituting a corps of fire - fighting watchmen in 24 hours. Regulations for checking and preventing fires were developed. In the preindustrial era most cities had watchmen

Anglų kalba  Rašiniai   (1 psl., 3,59 kB)

One is not born, but rather becomes woman. No biological, physiological, or economic fate determines the figure that the human female presents in society; it is civilization as a whole that produces this creature, intermediate between male and eunuch, which is described as feminine. But what was really position of woman in 1950s – 1960s. I overlook different women positions in western and communist nations.

Anglų kalba  Referatai   (5,41 kB)

Basketball is a ball sport in which two teams of five players each try to score points by throwing the ball through a basket. Basketball is highly suited to viewing by spectators, as it is primarily an indoor sport, played in a relatively small playing area, or "court," with only ten players, and using a large ball which is easy to follow. Additionally, the lack of protective gear makes it easy to see the reactions of the players. It is one of the most popular sports in the United States, and is also popular in other parts of the world, including South America, southern Europe, and the former Soviet Union, especially Lithuania.

Anglų kalba  Referatai   (4,88 kB)