{"id":20690,"date":"2022-08-16T14:36:55","date_gmt":"2022-08-16T12:36:55","guid":{"rendered":"https:\/\/kinit.sk\/components-and-properties-of-good-explanations-of-artificial-intelligence-models-decisions\/"},"modified":"2023-10-12T14:28:43","modified_gmt":"2023-10-12T12:28:43","slug":"komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie","status":"publish","type":"post","link":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/","title":{"rendered":"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie"},"content":{"rendered":"
\n

Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom<\/h2>\n<\/div>\n\n
\n
Z poh\u013eadu strojov\u00e9ho u\u010denia (angl. Machine learning, ML) \u017eijeme \u0161\u0165astn\u00e9 \u010dasy. Pre mnoh\u00e9 \u00falohy, ktor\u00e9 sme doteraz nevedeli s pomocou umelej inteligencie uspokojivo vyrie\u0161i\u0165, pozn\u00e1me nie jednu, ale mnoho r\u00f4znych ML met\u00f3d alebo modelov, z ktor\u00fdch si m\u00f4\u017eeme vybra\u0165 a dosiahnu\u0165 aspo\u0148 akceptovate\u013en\u00e9 v\u00fdsledky. Toto mno\u017estvo modelov, sp\u00f4sobov spracovania d\u00e1t, ich vari\u00e1ci\u00ed a mo\u017en\u00fdch kombin\u00e1ci\u00ed predstavuje v\u00fdzvu \u2013 mus\u00edme n\u00e1js\u0165 tak\u00fa konfigur\u00e1ciu, ktor\u00e1 vyhovuje na\u0161ej \u00falohe a d\u00e1tam.<\/h5>\n<\/div>\n\n
\n

Aby sme na\u0161li spr\u00e1vny model, mus\u00edme definova\u0165 krit\u00e9ri\u00e1, ktor\u00e9 meraj\u00fa, ako dobre konkr\u00e9tny model a jeho parametre a hyperparametre adresuj\u00fa dan\u00fd probl\u00e9m. \u010casto pou\u017e\u00edvan\u00fdmi krit\u00e9riami s\u00fa (v z\u00e1vislosti od \u00falohy) napr\u00edklad presnos\u0165, spr\u00e1vnos\u0165, \u00faplnos\u0165, F1 sk\u00f3re alebo kvadratick\u00e1 chyba. <\/p>\n<\/div>\n\n

\n

V \u010fal\u0161om kroku zvy\u010dajne vykon\u00e1me optimaliz\u00e1ciu tzv. hyperparametrov. Tu sa sna\u017e\u00edme n\u00e1js\u0165 dobr\u00fa konfigur\u00e1ciu modelu (napr\u00edklad architekt\u00faru neur\u00f3novej siete) a postupnos\u0165 krokov predspracovania a n\u00e1sledn\u00e9ho spracovania d\u00e1t. Ak m\u00e1me k dispoz\u00edcii dostato\u010dn\u00e9 prostriedky, m\u00f4\u017eeme dokonca vyu\u017ei\u0165 automatizovan\u00e9 strojov\u00e9 u\u010denie (angl. Automated ML alebo AutoML<\/em>), prostredn\u00edctvom ktor\u00e9ho m\u00f4\u017eeme n\u00e1js\u0165 vhodn\u00fa konfigur\u00e1ciu v priestore v\u0161etk\u00fdch dostupn\u00fdch konfigur\u00e1ci\u00ed plne automaticky.<\/p>\n<\/div>\n\n

\n

V\u010faka pozornosti, ktor\u00e1 je v posledn\u00fdch rokoch venovan\u00e1 vysvetlite\u013en\u00e9mu strojov\u00e9mu u\u010deniu, sa stal po\u010det met\u00f3d vysvetlite\u013enosti podobne ohromuj\u00faci ako po\u010det r\u00f4znych met\u00f3d strojov\u00e9ho u\u010denia. Ak chceme napr\u00edklad poskytn\u00fa\u0165 pou\u017e\u00edvate\u013eom dobr\u00e9 vysvetlenia rozhodnut\u00ed komplexn\u00e9ho modelu, mus\u00edme n\u00e1js\u0165 nielen spr\u00e1vnu met\u00f3du vysvetlite\u013enosti ako tak\u00fa, ale aj jej konkr\u00e9tnu konfigur\u00e1ciu. <\/p>\n<\/div>\n\n

\n

Aby sme tak\u00fato konfigur\u00e1ciu na\u0161li, m\u00f4\u017eeme si po\u017ei\u010da\u0165 niektor\u00e9 koncepty z AutoML. Podobne ako v AutoML, na vstupe m\u00e1me priestor dostupn\u00fdch met\u00f3d a ich konfigur\u00e1ci\u00ed. Na v\u00fdstupe chceme dosta\u0165 tak\u00fa kombin\u00e1ciu met\u00f3dy a konfigur\u00e1cie, ktor\u00e1 n\u00e1m bude poskytova\u0165 dobr\u00e9 vysvetlenia pre na\u0161u kombin\u00e1ciu \u00falohy, modelu a d\u00e1t. V takomto pr\u00edpade m\u00f4\u017eeme hovori\u0165 o automatizovanej vysvetlite\u013enej umelej inteligencii alebo AutoXAI<\/strong>.<\/p>\n<\/div>\n\n

\n
\"\"
Obr\u00e1zok 1: <\/strong>Dnes u\u017e existuje pomerne ve\u013ek\u00e9 mno\u017estvo r\u00f4znych met\u00f3d vysvetlite\u013enosti a mnoh\u00e9 z nich s\u00fa dokonca vysoko konfigurovate\u013en\u00e9. Ak chceme identifikova\u0165 tak\u00fa met\u00f3du, ktor\u00e1 poskytne dobr\u00e9 vysvetlenia pre konkr\u00e9tnu kombin\u00e1ciu \u00falohy, modelu a d\u00e1t, m\u00f4\u017eeme aplikova\u0165 pr\u00edstup podobn\u00fd automatizovan\u00e9mu strojov\u00e9mu u\u010deniu (AutoML). M\u00f4\u017eeme teda hovori\u0165 o automatizovanej vysvetlite\u013enej umelej inteligencii – AutoXAI. Najv\u00e4\u010d\u0161ou v\u00fdzvou je zadefinova\u0165 krit\u00e9ri\u00e1, ktor\u00e9 zmeraj\u00fa kvalitu vysvetlen\u00ed produkovan\u00fdch r\u00f4znymi met\u00f3dami  vysvetlite\u013enosti a ich konfigur\u00e1ciami.<\/em><\/figcaption><\/figure>\n<\/div>\n\n
\n

N\u00e1ro\u010dnou \u010das\u0165ou AutoXAI je zadefinova\u0165, ako porovn\u00e1va\u0165 r\u00f4zne met\u00f3dy vysvetlite\u013enosti. In\u00fdmi slovami \u2013 \u010do to znamen\u00e1 \u201cdobr\u00e9 vysvetlenie\u201d?<\/p>\n<\/div>\n\n

\n

Dobr\u00e9 vysvetlenie vyva\u017euje dva komponenty – zrozumite\u013enos\u0165 a vernos\u0165 <\/h3>\n<\/div>\n\n
\n

Pod\u013ea viacer\u00fdch autorov (napr. [1, 2]) by dobr\u00e9 vysvetlenie malo by\u0165 v rovnov\u00e1he medzi dvoma komponentmi. Malo by verne opisova\u0165 spr\u00e1vanie modelu (vernos\u0165, angl. fidelity<\/em>) a malo by by\u0165 zrozumite\u013en\u00e9 pre \u013eud\u00ed (angl. understandability<\/em>).<\/p>\n<\/div>\n\n

\n

Ak je vysvetlenie zrozumite\u013en\u00e9, ale ch\u00fdba mu vernos\u0165, mohli by sme skon\u010di\u0165 s vysvetleniami, ktor\u00e9 vyzeraj\u00fa \u201cdobre\u201d alebo tak, ako by si to \u010dlovek predstavoval, ale nemusia spr\u00e1vne opisova\u0165 rozhodovac\u00ed proces modelu. Znamen\u00e1 to, \u017ee vysvetlenie napr\u00edklad nemus\u00ed by\u0165 \u00fapln\u00e9. M\u00f4\u017ee v \u0148om ch\u00fdba\u0165 inform\u00e1cia o tom, \u017ee nejak\u00e1 \u010das\u0165 vstupn\u00fdch d\u00e1t v\u00fdznamne ovplyvnila rozhodnutie modelu strojov\u00e9ho u\u010denia. Pr\u00edklad m\u00f4\u017eete vidie\u0165 na Obr\u00e1zku 2.<\/p>\n<\/div>\n\n

\n
\"\"
Obr\u00e1zok 2: <\/strong>Teplotn\u00e1 mapa obsahuje ve\u013emi jasn\u00e9 a dobre ohrani\u010den\u00e9 zv\u00fdraznen\u00e9 oblasti, ktor\u00e9 pod\u013ea pou\u017eitej met\u00f3dy vysvetlite\u013enosti prispeli k tomu, \u017ee obr\u00e1zok bol klasifikovan\u00fd ako trieda \u201cpapag\u00e1j\u201d. Tak\u00e9to vysvetlenie je jednoducho zrozumite\u013en\u00e9 pre \u010dloveka, preto\u017ee zv\u00fdraz\u0148uje iba mal\u00e9 mno\u017estvo ve\u013emi dobre ohrani\u010den\u00fdch \u010dast\u00ed obr\u00e1zku. Na druhej strane, toto vysvetlenie nie je \u00fapln\u00e9. V skuto\u010dnosti ku klasifik\u00e1cii obr\u00e1zku do triedy \u201cpapag\u00e1j\u201d prispela aj oblas\u0165, v ktorej sa nach\u00e1dza hlava papag\u00e1ja, ktor\u00fd je na obr\u00e1zku v\u013eavo.<\/em><\/figcaption><\/figure>\n<\/div>\n\n
\n

Na druhej strane, ak vysvetleniu ch\u00fdba zrozumite\u013enos\u0165, m\u00f4\u017ee pou\u017e\u00edvate\u013ea zahlti\u0165 nadbyto\u010dn\u00fdmi inform\u00e1ciami, alebo m\u00f4\u017ee by\u0165 tak\u00e9 chaotick\u00e9, \u017ee mu pou\u017e\u00edvate\u013e (\u010dlovek) nebude rozumie\u0165. Napr\u00edklad, ak niektor\u00e9 \u010dasti vstupu mali zanedbate\u013en\u00fd vplyv na v\u00fdsledn\u00fa predikciu modelu, m\u00f4\u017ee by\u0165 kontraprodukt\u00edvne zahrn\u00fa\u0165 tak\u00fato inform\u00e1ciu do vysvetlenia, ktor\u00e9 je poskytnut\u00e9 \u010dloveku. Pr\u00edklad vysvetlenia vo forme teplotnej mapy, ktor\u00e1 zv\u00fdraz\u0148uje aj pixely obr\u00e1zku, ktor\u00e9 mali iba minim\u00e1lny vplyv na predikciu, m\u00f4\u017eete vidie\u0165 na Obr\u00e1zku 3.<\/p>\n<\/div>\n\n

\n
\"\"
Obr\u00e1zok 3: <\/strong>T\u00e1to teplotn\u00e1 mapa poskytuje ve\u013emi detailn\u00fa inform\u00e1ciu o tom, ktor\u00e9 oblasti obr\u00e1zku prispeli k predikcii pozit\u00edvne (\u017elt\u00e1 a \u010derven\u00e1 farba) a ktor\u00e9, naopak, negat\u00edvne (modr\u00e1 farba). Tak\u00e9to vysvetlenie poskytuje s\u00edce pomerne vern\u00fa inform\u00e1ciu o skuto\u010dnom spr\u00e1van\u00ed modelu, no z poh\u013eadu pou\u017e\u00edvate\u013ea obsahuje ve\u013ek\u00e9 mno\u017estvo bezv\u00fdznamn\u00fdch detailov, ktor\u00e9 ho m\u00f4\u017eu zahlti\u0165. <\/em><\/figcaption><\/figure>\n<\/div>\n\n
\n

Aby to nebolo tak\u00e9 jednoduch\u00e9, nie v\u017edy plat\u00ed, \u017ee medzi zrozumite\u013enos\u0165ou a vernos\u0165ou mus\u00ed by\u0165 rovnov\u00e1ha. Zv\u00fd\u0161enie zrozumite\u013enosti m\u00f4\u017ee v skuto\u010dnosti vies\u0165 k zn\u00ed\u017eeniu vernosti a naopak. Ak chceme napr\u00edklad pou\u017eiti\u0165 vysvetlenie na automatizovan\u00e9 ladenie modelu (angl. debugging<\/em>) a identifik\u00e1ciu ch\u00fdb v \u0148om, vernos\u0165 vysvetlen\u00ed bude d\u00f4le\u017eitej\u0161ia ako zrozumite\u013enos\u0165.<\/p>\n<\/div>\n\n

\n

Bli\u017e\u0161\u00ed poh\u013ead na zrozumite\u013enos\u0165 a vernos\u0165<\/h3>\n<\/div>\n\n
\n

Pre jednoduchos\u0165 sme sa doteraz obmedzili na dva komponenty vysvetlite\u013enosti – zrozumite\u013enos\u0165 a vernos\u0165. Po\u010fme sa teraz pozrie\u0165 na jednu z mo\u017en\u00fdch taxon\u00f3mi\u00ed, in\u0161pirovan\u00fa pr\u00e1cami [1, 2], ktor\u00e1 poskytuje podrobnej\u0161\u00ed poh\u013ead na vysvetlite\u013enos\u0165. E\u0161te detailnej\u0161ia taxon\u00f3mia po\u017eiadaviek na vysvetlite\u013en\u00e9 met\u00f3dy a syst\u00e9my a na r\u00e1mce na ich vyhodnocovanie bola publikovan\u00e1 autormi Sokol a kol. [3].<\/p>\n<\/div>\n\n

\n
\"\"
Obr\u00e1zok 4: <\/strong>Komponenty a vlastnosti vysvetlite\u013enosti. In\u0161pirovan\u00e9 [1, 2].<\/em><\/figcaption><\/figure>\n<\/div>\n\n
\n

Z poh\u013eadu spr\u00e1vnosti <\/strong>(angl. soundness<\/em>) sa p\u00fdtame, do akej miery je vysvetlenie pravdiv\u00e9 vzh\u013eadom na model (a d\u00e1ta), na ktor\u00e9 sa via\u017ee. Explicitne sa zameriava na to, \u010di vysvetlenie skuto\u010dne reflektuje iba spr\u00e1vaniu modelu. Vysvetlenie by teda nemalo zav\u00e1dza\u0165, malo by \u010dloveku hovori\u0165 \u201cni\u010d in\u00e9, ne\u017e pravdu\u201d. Niekedy sa pri implement\u00e1cii met\u00f3d vysvetlite\u013enosti, ako napr\u00edklad okl\u00fazna anal\u00fdza, vyu\u017e\u00edvaj\u00fa triky a optimaliz\u00e1cie, v\u010faka ktor\u00fdm s\u00fa v\u00fdpo\u010dtovo menej n\u00e1ro\u010dn\u00e9. Pri tejto met\u00f3de, namiesto toho, aby sme postupne \u201cmazali\u201d \u010dasti vstupu jednu po druhej a sledovali dopad na predikciu modelu, zma\u017eeme v\u00e4\u010d\u0161iu \u010das\u0165 vstupu naraz. M\u00f4\u017ee sa teda sta\u0165, \u017ee ak naraz vyma\u017eeme \u010das\u0165, ktor\u00e1 mala vplyv na predikciu modelu a aj tak\u00fa, ktor\u00e1 vplyv nemala, fin\u00e1lne vysvetlenie o oboch t\u00fdchto \u010dastiach povie, \u017ee v\u00fdznamne ovplyvnili predikciu.<\/p>\n<\/div>\n\n

\n

Kompletnos\u0165 <\/strong>(angl. completeness<\/em>) opisuje kvalitu vysvetlen\u00ed z h\u013eadiska toho, ak\u00e1 \u010das\u0165 cel\u00e9ho rozhodovacieho procesu a dynamiky modelu je zachyten\u00e1 dan\u00fdm vysvetlen\u00edm. P\u00fdtame sa, do akej miery vysvetlenie opisuje cel\u00fd model (alebo dokonca syst\u00e9m). Chceme, aby n\u00e1m vysvetlenie hovorilo nielen pravdu, ale \u201ccel\u00fa pravdu\u201d.<\/p>\n<\/div>\n\n

\n

Jasnos\u0165 <\/strong>(angl. clarity<\/em>) hovor\u00ed o tom, \u017ee vysvetlenie by nemalo by\u0165 zm\u00e4to\u010dn\u00e9 a malo by by\u0165 jednozna\u010dn\u00e9. Okrem toho by pre podobn\u00e9 vstupy, pri ktor\u00fdch sa aj model samotn\u00fd rozhodoval podobn\u00fdm sp\u00f4sobom, mali by\u0165 aj vysvetlenia podobn\u00e9. V opa\u010dnom pr\u00edpade tak\u00e9to vysvetlenia nebude \u010dlovek vn\u00edma\u0165 ako konzistentn\u00e9.<\/p>\n<\/div>\n\n

\n

\u0160\u00edrka <\/strong>(angl. broadness<\/em>) opisuje, na ak\u00fa \u010das\u0165 \u00falohy a d\u00e1t je vysvetlenie mo\u017en\u00e9 aplikova\u0165. Ako pr\u00edklad si uvedieme hlbok\u00fa konvolu\u010dn\u00fa neur\u00f3nov\u00fa sie\u0165, ktorej \u00falohou je klasifikova\u0165 ma\u013eby ma\u013eovan\u00e9 v r\u00f4znych \u0161t\u00fdloch. Od met\u00f3dy vysvetlite\u013enosti chceme, aby n\u00e1m objasnila, pre\u010do konkr\u00e9tny obraz neur\u00f3nov\u00e1 sie\u0165 zaradila do konkr\u00e9tneho \u0161t\u00fdlu. M\u00f4\u017ee sa sta\u0165, \u017ee jedna a t\u00e1 ist\u00e1 met\u00f3da vysvetlite\u013enosti bude schopn\u00e1 poskytn\u00fa\u0165 zmyslupln\u00e9 vysvetlenie len pre ur\u010dit\u00fa skupinu obrazov. Napr\u00edklad, pri modern\u00fdch obrazoch, ktor\u00e9 sa skladaj\u00fa len zo z\u00e1kladn\u00fdch geometrick\u00fdch tvarov, bude zd\u00f4vodnenie tak\u00e9, \u017ee sa na obraze nach\u00e1dza ve\u013ek\u00e9 mno\u017estvo trojuholn\u00edkov, \u0161tvorcov \u010di kruhov a \u017eiadne in\u00e9 tvary. Naopak, pri ve\u013emi abstraktn\u00fdch dielach (napr\u00edklad na \u0161t\u00fdl Jacksona Pollocka) nebude met\u00f3da schopn\u00e1 poskytn\u00fa\u0165 \u017eiadne zmyslupln\u00e9 vysvetlenie.<\/p>\n<\/div>\n\n

\n

Na \u00faspornos\u0165 <\/strong>(angl. parsimony<\/em>) sa vieme pozera\u0165 ako na aplik\u00e1ciu princ\u00edpu Occamovej britvy. Ako sme uk\u00e1zali na pr\u00edklade vy\u0161\u0161ie, vysvetlenie by z poh\u013eadu zrozumite\u013enosti malo obsahova\u0165 len tie detaily, ktor\u00e9 prin\u00e1\u0161aj\u00fa adres\u00e1tovi (\u010dloveku) nejak\u00fa hodnotu. Malo by by\u0165 tak\u00e9 jednoduch\u00e9, ako je to len mo\u017en\u00e9. Ak m\u00e1 vysvetlenie formu vety v prirodzenom jazyku, t\u00e1to veta by nemala obsahova\u0165 zbyto\u010dn\u00fa vatu. Napr\u00edklad, lep\u0161\u00edm vysvetlen\u00edm toho, pre\u010do model klasifikoval aktivitu na obr\u00e1zku ako \u201cvarenie\u201d, je veta \u201cna obr\u00e1zku je mu\u017e v z\u00e1stere a dr\u017e\u00ed v ruke varechu\u201d a nie veta \u201cna obr\u00e1zku je star\u0161\u00ed mu\u017e so \u0161ediv\u00fdmi vlasmi, v ruke dr\u017e\u00ed vyrez\u00e1van\u00fa varechu hnedej farby, hodiny na stene ukazuj\u00fa 11 hod\u00edn a na stole s\u00fa kvety\u201d.<\/p>\n<\/div>\n\n

\n

Z\u00e1ver<\/h3>\n<\/div>\n\n
\n

V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sme sa pozreli, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie. <\/p>\n<\/div>\n\n

\n

Uk\u00e1zali sme si, \u017ee dobr\u00e9 vysvetlenie by malo vyva\u017eova\u0165 zrozumite\u013enos\u0165 a vernos\u0165. Mus\u00ed by\u0165 pochopite\u013en\u00e9 pre \u010dloveka a z\u00e1rove\u0148 mus\u00ed verne opisova\u0165 skuto\u010dn\u00e9 spr\u00e1vanie modelu, ktor\u00e9ho predikcie vysvet\u013euje. <\/p>\n<\/div>\n\n

\n

Dosiahnu\u0165 rovnov\u00e1hu medzi zrozumite\u013enos\u0165ou a vernos\u0165ou pritom nie je jednoduch\u00e1 \u00faloha, preto\u017ee zv\u00fd\u0161enie niektorej z vlastnost\u00ed zrozumite\u013enosti m\u00f4\u017ee z\u00e1rove\u0148 zn\u00ed\u017ei\u0165 vernos\u0165 a naopak. Z\u00e1rove\u0148 plat\u00ed, \u017ee z\u00e1le\u017e\u00ed od konkr\u00e9tnej \u00falohy, \u010di chceme ma\u0165 vysvetlenia viac zrozumite\u013en\u00e9, alebo \u010di, chceme, aby boli ve\u013emi detailn\u00e9 a pokr\u00fdvali v\u0161etky detaily rozhodovacieho procesu modelu strojov\u00e9ho u\u010denia.<\/p>\n<\/div>\n\n

\n

V \u010fal\u0161ej \u010dasti seri\u00e1lu si uk\u00e1\u017eeme dve rodiny pr\u00edstupov a pr\u00edklady konkr\u00e9tnych metr\u00edk, ktor\u00fdmi m\u00f4\u017eeme mera\u0165 kvalitu vysvetlen\u00ed z poh\u013eadu vlastnost\u00ed, ktor\u00e9 sme si v tomto \u010dl\u00e1nku pop\u00edsali.<\/p>\n<\/div>\n\n

\n
\n
\n

<\/p>\n<\/div>\n\n

\n

<\/p>\n<\/div>\n\n

\n

Projekt podporil Nada\u010dn\u00fd fond PricewaterhouseCoopers v Nad\u00e1cii Pontis.<\/em><\/p>\n<\/div><\/div>\n<\/div>\n\n

\n
\n
\"\"<\/figure>\n<\/div><\/div>\n<\/div><\/div>\n<\/div>\n\n
\n

Referencie<\/h4>\n<\/div>\n\n
\n

[1] ZHOU, Jianlong, et al. Evaluating the quality of machine learning explanations: A survey on methods and metrics. Electronics, 2021.<\/p>\n<\/div>\n\n

\n

[2] MARKUS, Aniek F.; KORS, Jan A.; RIJNBEEK, Peter R. The role of explainability in creating trustworthy artificial intelligence for health care: a comprehensive survey of the terminology, design choices, and evaluation strategies. Journal of Biomedical Informatics, 2021.<\/p>\n<\/div>\n\n

\n

[3] SOKOL, Kacper; FLACH, Peter. Explainability fact sheets: a framework for systematic assessment of explainable approaches. In: Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency. 2020.<\/p>\n<\/div>\n\n

\n\t
\n\t\t
\n\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom<\/h2><\/div>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t

Ako rozm\u00fd\u0161\u013ea umel\u00e1 inteligencia?<\/h3>\n \t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tapr 28. 2022\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom M\u00f4\u017ee umel\u00e1 inteligencia rozhodova\u0165 o tom, kto bude prepusten\u00fd z v\u00e4zenia, alebo kto post\u00fapi do \u010fal\u0161ieho kola pracovn\u00fdch pohovorov? M\u00f4\u017ee odpor\u00fa\u010da\u0165…<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/a>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/article>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t

Tri pr\u00ednosy vysvetlite\u013enej umelej inteligencie<\/h3>\n \t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tm\u00e1j 25. 2022\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom Umel\u00e1 inteligencia bude \u010doraz v\u00e4\u010d\u0161mi zasahova\u0165 do na\u0161ich \u017eivotov. Ve\u013ea \u013eud\u00ed v\u0161ak umelej inteligencii ned\u00f4veruje a maj\u00fa probl\u00e9m prenecha\u0165 jej aj…<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/a>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/article>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t

Transparentn\u00e9 modely vs. modely typu \u010dierna skrinka<\/h3>\n \t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tj\u00fan 21. 2022\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom M\u00f4\u017ee umel\u00e1 inteligencia rozhodova\u0165 o tom, kto bude prepusten\u00fd z v\u00e4zenia, alebo kto post\u00fapi do \u010fal\u0161ieho kola pracovn\u00fdch pohovorov? M\u00f4\u017ee odpor\u00fa\u010da\u0165…<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/a>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/article>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t

Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie<\/h3>\n \t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\taug 16. 2022\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom Z poh\u013eadu strojov\u00e9ho u\u010denia (angl. Machine learning, ML) \u017eijeme \u0161\u0165astn\u00e9 \u010dasy. Pre mnoh\u00e9 \u00falohy, ktor\u00e9 sme doteraz nevedeli s pomocou umelej…<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/a>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/article>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t

Ako mera\u0165 kvalitu vysvetlenia predpovede umelej inteligencie?<\/h3>\n \t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tsep 26. 2022\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom V minulej \u010dasti seri\u00e1lu sme sa pozreli na to, ak\u00e9 komponenty a vlastnosti by mali ma\u0165 dobr\u00e9 vysvetlenia. Prv\u00fd komponent, zrozumite\u013enos\u0165,…<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/a>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/article>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t

S\u00fa vysvetlenia, o ktor\u00fdch si mysl\u00edme, \u017ee s\u00fa dobr\u00e9, naozaj dobr\u00e9?<\/h3>\n \t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\tsep 27. 2023\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t
Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom V predo\u0161l\u00fdch \u010dl\u00e1nkoch sme si predstavili tri ve\u013emi zauj\u00edmav\u00e9 koncepty z oblasti vysvetlite\u013enej umelej inteligencie (vysvetlite\u013en\u00e1 AI alebo XAI): V tomto…<\/div>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t<\/a>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/article>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\t<\/div>\n<\/section>\n<\/div>","protected":false},"excerpt":{"rendered":"

Vysvetlite\u013en\u00e1 umel\u00e1 inteligencia: od \u010diernych skriniek k transparentn\u00fdm modelom Z poh\u013eadu strojov\u00e9ho u\u010denia (angl. Machine learning, ML) \u017eijeme \u0161\u0165astn\u00e9 \u010dasy. Pre mnoh\u00e9 \u00falohy, ktor\u00e9 sme doteraz nevedeli s pomocou umelej…<\/p>\n","protected":false},"author":26,"featured_media":20677,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[76,88],"tags":[386],"class_list":["post-20690","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-natural-language-processing-sk","category-pop-science-sk","tag-xai-sk"],"acf":[],"yoast_head":"\nKomponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie - KInIT<\/title>\n<meta name=\"description\" content=\"V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sa pozrieme, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/\" \/>\n<meta property=\"og:locale\" content=\"sk_SK\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie - KInIT\" \/>\n<meta property=\"og:description\" content=\"V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sa pozrieme, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/\" \/>\n<meta property=\"og:site_name\" content=\"KInIT\" \/>\n<meta property=\"article:published_time\" content=\"2022-08-16T12:36:55+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-10-12T12:28:43+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/kinit.sk\/wp-content\/uploads\/2022\/08\/A4.2-Feautre-Web-post-XAI-series.png\" \/>\n\t<meta property=\"og:image:width\" content=\"2401\" \/>\n\t<meta property=\"og:image:height\" content=\"1261\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Marianna Palkova\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@kinit\" \/>\n<meta name=\"twitter:site\" content=\"@kinit\" \/>\n<meta name=\"twitter:label1\" content=\"Autor\" \/>\n\t<meta name=\"twitter:data1\" content=\"Marianna Palkova\" \/>\n\t<meta name=\"twitter:label2\" content=\"Predpokladan\u00fd \u010das \u010d\u00edtania\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 min\u00fat\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/\"},\"author\":{\"name\":\"Marianna Palkova\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/#\\\/schema\\\/person\\\/8b175aaaf3267b5bbbbb97e4a6db8cea\"},\"headline\":\"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie\",\"datePublished\":\"2022-08-16T12:36:55+00:00\",\"dateModified\":\"2023-10-12T12:28:43+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/\"},\"wordCount\":1886,\"image\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/kinit.sk\\\/wp-content\\\/uploads\\\/2022\\\/08\\\/A4.2-Feautre-Web-post-XAI-series.png\",\"keywords\":[\"XAI\"],\"articleSection\":[\"Natural Language Processing\",\"Pop science\"],\"inLanguage\":\"sk-SK\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/\",\"url\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/\",\"name\":\"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie - KInIT\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/kinit.sk\\\/wp-content\\\/uploads\\\/2022\\\/08\\\/A4.2-Feautre-Web-post-XAI-series.png\",\"datePublished\":\"2022-08-16T12:36:55+00:00\",\"dateModified\":\"2023-10-12T12:28:43+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/#\\\/schema\\\/person\\\/8b175aaaf3267b5bbbbb97e4a6db8cea\"},\"description\":\"V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sa pozrieme, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#breadcrumb\"},\"inLanguage\":\"sk-SK\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"sk-SK\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#primaryimage\",\"url\":\"https:\\\/\\\/kinit.sk\\\/wp-content\\\/uploads\\\/2022\\\/08\\\/A4.2-Feautre-Web-post-XAI-series.png\",\"contentUrl\":\"https:\\\/\\\/kinit.sk\\\/wp-content\\\/uploads\\\/2022\\\/08\\\/A4.2-Feautre-Web-post-XAI-series.png\",\"width\":2401,\"height\":1261},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/uvod\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Natural Language Processing\",\"item\":\"https:\\\/\\\/kinit.sk\\\/sk\\\/category\\\/natural-language-processing-sk\\\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/#website\",\"url\":\"https:\\\/\\\/kinit.sk\\\/\",\"name\":\"KInIT\",\"description\":\"Vyu\u017e\u00edvame v\u00fdskum pre \u013eud\u00ed a priemysel\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/kinit.sk\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"sk-SK\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/kinit.sk\\\/#\\\/schema\\\/person\\\/8b175aaaf3267b5bbbbb97e4a6db8cea\",\"name\":\"Marianna Palkova\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie - KInIT","description":"V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sa pozrieme, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/","og_locale":"sk_SK","og_type":"article","og_title":"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie - KInIT","og_description":"V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sa pozrieme, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie.","og_url":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/","og_site_name":"KInIT","article_published_time":"2022-08-16T12:36:55+00:00","article_modified_time":"2023-10-12T12:28:43+00:00","og_image":[{"width":2401,"height":1261,"url":"https:\/\/kinit.sk\/wp-content\/uploads\/2022\/08\/A4.2-Feautre-Web-post-XAI-series.png","type":"image\/png"}],"author":"Marianna Palkova","twitter_card":"summary_large_image","twitter_creator":"@kinit","twitter_site":"@kinit","twitter_misc":{"Autor":"Marianna Palkova","Predpokladan\u00fd \u010das \u010d\u00edtania":"8 min\u00fat"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#article","isPartOf":{"@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/"},"author":{"name":"Marianna Palkova","@id":"https:\/\/kinit.sk\/#\/schema\/person\/8b175aaaf3267b5bbbbb97e4a6db8cea"},"headline":"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie","datePublished":"2022-08-16T12:36:55+00:00","dateModified":"2023-10-12T12:28:43+00:00","mainEntityOfPage":{"@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/"},"wordCount":1886,"image":{"@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#primaryimage"},"thumbnailUrl":"https:\/\/kinit.sk\/wp-content\/uploads\/2022\/08\/A4.2-Feautre-Web-post-XAI-series.png","keywords":["XAI"],"articleSection":["Natural Language Processing","Pop science"],"inLanguage":"sk-SK"},{"@type":"WebPage","@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/","url":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/","name":"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie - KInIT","isPartOf":{"@id":"https:\/\/kinit.sk\/#website"},"primaryImageOfPage":{"@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#primaryimage"},"image":{"@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#primaryimage"},"thumbnailUrl":"https:\/\/kinit.sk\/wp-content\/uploads\/2022\/08\/A4.2-Feautre-Web-post-XAI-series.png","datePublished":"2022-08-16T12:36:55+00:00","dateModified":"2023-10-12T12:28:43+00:00","author":{"@id":"https:\/\/kinit.sk\/#\/schema\/person\/8b175aaaf3267b5bbbbb97e4a6db8cea"},"description":"V tejto \u010dasti n\u00e1\u0161ho seri\u00e1lu o vysvetlite\u013enej umelej inteligencii sa pozrieme, ak\u00e9 komponenty a vlastnosti by malo ma\u0165 dobr\u00e9 vysvetlenie.","breadcrumb":{"@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#breadcrumb"},"inLanguage":"sk-SK","potentialAction":[{"@type":"ReadAction","target":["https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/"]}]},{"@type":"ImageObject","inLanguage":"sk-SK","@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#primaryimage","url":"https:\/\/kinit.sk\/wp-content\/uploads\/2022\/08\/A4.2-Feautre-Web-post-XAI-series.png","contentUrl":"https:\/\/kinit.sk\/wp-content\/uploads\/2022\/08\/A4.2-Feautre-Web-post-XAI-series.png","width":2401,"height":1261},{"@type":"BreadcrumbList","@id":"https:\/\/kinit.sk\/sk\/komponenty-a-vlastnosti-dobrych-vysvetleni-rozhodnuti-modelov-umelej-inteligencie\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/kinit.sk\/sk\/uvod\/"},{"@type":"ListItem","position":2,"name":"Natural Language Processing","item":"https:\/\/kinit.sk\/sk\/category\/natural-language-processing-sk\/"},{"@type":"ListItem","position":3,"name":"Komponenty a vlastnosti dobr\u00fdch vysvetlen\u00ed rozhodnut\u00ed modelov umelej inteligencie"}]},{"@type":"WebSite","@id":"https:\/\/kinit.sk\/#website","url":"https:\/\/kinit.sk\/","name":"KInIT","description":"Vyu\u017e\u00edvame v\u00fdskum pre \u013eud\u00ed a priemysel","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/kinit.sk\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"sk-SK"},{"@type":"Person","@id":"https:\/\/kinit.sk\/#\/schema\/person\/8b175aaaf3267b5bbbbb97e4a6db8cea","name":"Marianna Palkova"}]}},"_links":{"self":[{"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/posts\/20690","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/users\/26"}],"replies":[{"embeddable":true,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/comments?post=20690"}],"version-history":[{"count":7,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/posts\/20690\/revisions"}],"predecessor-version":[{"id":28795,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/posts\/20690\/revisions\/28795"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/media\/20677"}],"wp:attachment":[{"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/media?parent=20690"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/categories?post=20690"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/kinit.sk\/sk\/wp-json\/wp\/v2\/tags?post=20690"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}