InoteveraGaidhi rinotevera
Distance Metrics in Machine Learning
Basics
Basics GUIDE
Kurasikirwa Kwetatu uye Metric Kudzidza
Kurasika katatu kunodzidzisa neural network kuisa zvinhu zvakafanana padyo pamwe chete uye zvinhu zvakasiyana zviri kure munzvimbo yekumisikidza.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
Ndiwo hwaro kuseri kwekuzivikanwa kwechiso, kutsvaga mifananidzo, uye masisitimu ekurudziro anoda kuenzanisa zvinhu pane kungozvironga.
Kudzika Kwakadzika
Metric yekudzidza inodzidzisa modhi kugadzira embeddings, mavekita uko kureba kunoratidza kufanana. Triplet kurasikirwa kunoita izvi uchishandisa matatu ekuisa panguva: anchor, yakanaka (yakafanana kirasi seiyo anchor), uye isina kunaka (yakasiyana kirasi). Chinangwa chinosundidzira chibatiso pedyo nechakanaka pane chakashata neinenge yakatemwa muganho. Pakare, kurasikirwa kuri max(0, d(a,p) - d(a,n) + margin), uko d inowanzonzi Euclidean chinhambwe. Google's 2015 FaceNet yakasimudzira nzira iyi, kudzidza 128-dimensional yekumisikidza kumeso zvakananga. Kana uchinge wadzidziswa, unoenzanisa chero zvinhu zviviri nekombuta chinhambwe, hapana kudzidziswazve kunodiwa kune zvitsva kuzivikanwa. Uku kugona kwakavhurika-seti ndiko kusaka metric yekudzidza masimba ekuongorora uye kudzoreredza mabasa emhando isingagone kubata zviri nyore.
Kuitwa muNyika yeChokwadi
Kuongororwa kwechiso cheFaceNet: nhare uye magedhi epasipoti anosimbisa chitupa nekutarisa kana maviri ekumisikidza kumeso achiwira mukati mechikumbaridzo.
Kutsvaga kwechigadzirwa chinooneka: e-commerce saiti rega vatengi vaise pikicha uye vatore zvinooneka zvinhu zvakafanana nepedyo-yepedyo-embedding yekutarisa.
Ongororo yeMutauriri: vabatsiri vezwi vanoisa sampuli yezwi uye voienzanisa neprofile yakanyoreswa kuratidza kuti ndiani ari kutaura.
Siginicha nerunyoro rwekusimbisa: mabhangi anomisikidza referensi uye masiginecha emibvunzo uye mureza wefogeries kana chinhambwe chadarika chikamu chakadzidziswa.
Ramangwana reKurasika Kwetatu uye Metric Kudzidza
Kurasika kwakachena katatu kuri kuramba kuchitsiviwa nezvinangwa zvebatch-wide senge-multi-kufanana, proxy-anchor, uye kurasikirwa kwakasiyana (InfoNCE) inofananidza mapairi mazhinji padanho uye kuchinjika nekukurumidza. Nzira dzekuzvitarisira dzakaita seSimCLR dzinoratidza metric kudzidza inogona kushanda isina mavara nekubata akawedzera maonero seakanaka. Sezvo vector dhatabhesi uye kudzoreredza-yakawedzera chizvarwa kuwedzera, yakadzidzwa embeddings inotsigisa semantic yekutsvaga pabhiriyoni-chinhu chikero, saka pfungwa yepakati-sekufanana-kufanana iri kuwedzera pakati, kunyangwe iyo chaiyo triplet kuumbwa kunopera.
Kuzvishandisa
Nei zvichikosha
- Sarudzo dzakajeka. Inokubatsira kuparadzanisa zvakajeka zvichemo zvehunyanzvi kubva mumutauro wekushambadzira.
- Mutengo uye bhajeti. Iwe unogona kubvunza zvirinani kuita mibvunzo usati washandisa mari kana nguva.
- Team uye workflow. Zvikwata zvine nzwisiso yakagovaniswa inoita zvirinani chigadzirwa, mutemo, uye sarudzo dzekudzidza.
Chenjera
- Zvikwata zvakasiyana zvinogona kushandisa izwi rimwechete zvakasiyana, saka tsanangura nzvimbo nekukurumidza.
- Benchmarks inogona kutaridzika yakasimba nepo chaiyo-yenyika kuita isina kuenzana.
- Kuregeredza mhando yedata uye zvirongwa zvekuongorora zvinowanzogadzira mhedzisiro isina kusimba.
Matanho anotevera
- Tanga netsanangudzo yemutauro wakajeka yemhedzisiro yaunoda.
- Sarudza metric imwe yekubudirira uye imwe yekutadza mamiriro usati waedzwa.
- Mhanya mutyairi mudiki ane data remumiriri, kwete demo rakakwenenzverwa.
- Gwaro uko Kurasika Kwetatu uye Metric Kudzidza kunobatsira uye uko nzira dzakareruka dziri nani.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Take the Triplet Loss and Metric Learning quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
Chii chinonzi Triplet Kurasika uye Metric Kudzidza?
Kurasika katatu kunodzidzisa neural network kuisa zvinhu zvakafanana padyo pamwe chete uye zvinhu zvakasiyana zviri kure munzvimbo yekumisikidza. Ndiwo hwaro kuseri kwekuzivikanwa kwechiso, kutsvaga mifananidzo, uye masisitimu ekurudziro anoda kuenzanisa zvinhu pane kungozvironga.
Ndezvipi zvinhu zvitatu zvinoumba hutatu mukurasikirwa katatu?
Katatu kane anchor, yakanaka inogovera kirasi yeancho, uye isina kunaka kubva kune imwe kirasi.
Sei kurasikirwa katatu kuchisanganisira izwi remuganho?
Pasina muganho, modhi yaigona mepu zvese kusvika panzvimbo imwechete, zvichiita kuti madaro ese ave zero uye zvishoma kugutsa kurasikirwa. Muganho unomanikidza kuparadzana chaiko.
Ndeipi yakakurumbira 2015 system yakakurumbira kurasikirwa katatu kwekuzivikanwa kwechiso?
Google's FaceNet yakashandisa kurasikirwa katatu kudzidza kusungirirwa kwechiso uye kuseta benchmark mukusimbisa kumeso.
Chii chakadzidziswa metric-yekudzidza modhi inobuda pane yega yega?
Iyo modhi inoisa mamepu ekumisikidza mavheji; iwe wobva waenzanisa zvinhu nekuyera kureba pakati pekumisikidzwa kwavo.
Nei kudzidza metric kuchikodzera kuvhura-seti matambudziko sekuwedzera zviso zvitsva?
Nekuti kucherechedzwa kunoenderana nekumisikidza chinhambwe, unogona kunyoresa chitupa chitsva nekungochengeta chakadzikwa, hapana modhi yekudzidzira inodiwa.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi