InoteveraGaidhi rinotevera
Data Flywheels: Turning Production Data into Training Data
Tekinoroji
Nhungamiro yehunyanzvi
Kufurira Mabasa eKudzidzisa Data Attribution
Kufurira mabasa anofungidzira kuti yakawanda sei muenzaniso wega wega wekudzidzisa wakagadzira kufanotaura kwemuenzaniso, zvichiita kuti uteedzere zvakabuda kumashure kune data rakazvikonzera.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
They matter because they turn an opaque model into something auditable for copyright, debugging, and trust.
Kudzika Kwakadzika
Influence functions come from robust statistics and were adapted to deep learning by Koh and Liang in 2017. The core question is counterfactual: how would the model's loss on a test point change if a particular training example were removed or up-weighted? Rather than actually retraining (which is hopelessly expensive), influence functions approximate that change using calculus. They compute the gradient of the loss for the training point and the test point, then connect them through the inverse Hessian of the loss, which captures the curvature of the model's parameter space. Under Koh and Liang’s standard upweighting-on-test-loss convention, I_up,loss measures how infinitesimally increasing a training point’s weight changes the loss on a chosen test point. A positive value means upweighting raises test loss (a harmful influence); a negative value means it lowers test loss (a helpful influence). Removing a point reverses the sign of the first-order estimated change. Other influence scores may target predictions or use a different sign convention, so define the quantity before interpreting its sign. The result is a ranked list of the most responsible training examples.
Kuitwa muNyika yeChokwadi
Kutsvaga kuti ndeapi mabhuku ane kodzero dzekodzero zvakanyanya kupesvedzera ndima yemhando yemutauro yakagadzirwa, yekuongorora zviri pamutemo nerezenisi
Kugadzirisa kusarudzika nekuisa mifananidzo yakanyorwa zvisirizvo iyo yakasundira modhi kune mhinduro isiri iyo.
Kuona mienzaniso ine muchetura kana isinganzwisisike yekudzidziswa inopa simba repamusoro pane zvakafanotaurwa
Kuongorora kiredhiti kana modhi yekuhaya kuratidza kuti ndeapi marekodhi akatungamira sarudzo inopokana
Ramangwana reKufurira Mabasa eKudzidzisa Data Attribution
Tarisira pesvedzero-yakavakirwa hunhu kuti ive zvivakwa zveAI kuzvidavirira. Vatongi nematare edzimhosva ari kuongorora kana mavara ane copyrighted akaumbwa zvinobuda achada muenzaniso-level provenance, uye vanogadzira vanozoishandisa kuburitsa data isina kunyorwa kana ine chepfu. Mafungiro akachipa senge TRAK uye gradient-sketching ari kusundira kupihwa kunguva chaiyo, uye kuisanganisa nekusadzidza kunogona kuita kuti zvikwata zvibvise pesvedzero yegwaro pasina kudzidziswazve kuzere.
Kuzvishandisa
Nei zvichikosha
- Mutengo uye bhajeti. Zvisarudzo zvezvivakwa zvinotyaira kuita uye mutengo wekushandisa kwemakore.
- Sarudzo dzakajeka. Dzidzo yehunyanzvi inobatsira zvikwata kusarudza murwi wakakodzera, kwete iwo mutsva chete.
- Kudzora kwemhando yepamusoro. Sarudzo dzeinjiniya dziri nani dzinoderedza zviitiko zvekuvimbika mukugadzira.
Chenjera
- Kugadzirisa imwe bhenji kunogona kuvanza yakafara system kushaya simba.
- Infrastructure uye mari yekugadzirisa inowanzotarisirwa pasi.
- Chengetedzo uye kucherechedzwa mapundu anogona kukura sezvo masisitimu anowedzera kuoma.
Matanho anotevera
- Tsanangura latency, mhando, uye mutengo zvinangwa usati waitwa.
- Benchmark pasi pechokwadi mutoro uye data mamiriro.
- Chishandiso chekutarisa zvikanganiso, kudonha, uye mushandisi maitiro.
- Gadzirira nzira dzekudzosera kumashure uye dzezviitiko usati wawedzera.
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 Influence Functions for Training Data Attribution 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
What is Influence Functions for Training Data Attribution?
Kufurira mabasa anofungidzira kuti yakawanda sei muenzaniso wega wega wekudzidzisa wakagadzira kufanotaura kwemuenzaniso, zvichiita kuti uteedzere zvakabuda kumashure kune data rakazvikonzera. Ivo vane basa nekuti vanoshandura opaque modhi kuita chimwe chinhu chinotarisika kune copyright, debugging, uye kuvimba.
Ndeupi mubvunzo wekunyepedzera unopesvedzera mabasa akaenzana?
Kufurira mabasa anofungidzira mhedzisiro yekuvhiringa uremu hwemuenzaniso wekudzidziswa pakurasikirwa panzvimbo yekuyedzwa, kufungidzira kusiya-imwe-kunze kudzidzirazve pasina kuzviita.
Ndechipi chinhu chemasvomhu chinoita kuti pesvedzero chaiyoiyo ikodzerwe kusabatika kumamodheru makuru?
Iyo yemhando yepamusoro pesvedzero fomula inoda inverting iyo Hessian pamusoro pese paramita, izvo zvisingagoneke kune mamodheru ane mabhiriyoni emaparamita.
Ndiani akagadziridza pesvedzero mabasa kune yazvino kudzidza kwakadzama mune inozivikanwa 2017 bepa?
Pang Wei Koh naPercy Liang's 2017 bepa 'Kunzwisisa Black-bhokisi Predictions kuburikidza neKufurira Mabasa' yakaunza nzira iyi mukudzidza kwakadzama.
Under Koh and Liang’s standard I_up,loss convention, what does a negative value indicate?
For I_up,loss, a negative derivative means increasing the training point’s weight decreases the selected test loss; removing it has the opposite first-order effect.
Ndeipi fungidziro yakashandiswa ne Anthropic kupesvedzera mabasa kumhando dzemitauro mikuru?
Anthropic's 2023 ongororo yakashandisa EK-FAC kuenzanisa iyo inverse Hessian, zvichiita kuti pesvedzero yekuongorora pamhando dzemitauro mikuru.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi