InoteveraGaidhi rinotevera
Building Local LLM Apps with Ollama
Tekinoroji
Nhungamiro yehunyanzvi
Local Outlier Factor (LOF) inocherekedza zvakacherechedzwa nekuenzanisa density yayo yemunharaunda nekuwanda kwevavakidzani vepedyo.
Teerera kugwaro iriAI-yakarondedzerwa · ~3 min
Inogona kuona mapoinzi asina kujairika mukati menzvimbo yakaomesesa kana yakakora, asi kuverenga kwevavakidzani, kuyera chinhambwe uye mutsauko uripo pakati pekuonekwa kwekunze nekuonekwa kwezvitsva zvinokanganisa kushandiswa kwayo.
LOF ndeye density-based anomaly mamakisi. Pakutarisa kwega kwega, inowana nharaunda ye k padyo nzvimbo uye inofungidzira kusvikika kwenzvimbo, izvo zvinoratidza kuti nzvimbo yacho yakarongedzerwa zvakadii maererano nevavakidzani vayo. LOF inoenzanisa avhareji yekusvikirwa kwenzvimbo yevavakidzani nekuwanda kwenzvimbo yacho. Chibodzwa chiri pedyo nechimwe chinoratidza kuenzana kwenzvimbo; zvibodzwa zvakati kurei zvinoratidza kuti poindi yacho haina kukora pane zviri pedyo. Iyo ienzaniso yenzvimbo kwete yepasi rose chinhambwe-kubva-pakati chiyero. Iyi nharaunda inogona kuburitsa anomalies iyo nzira dzepasi rose dzinopotsa. Imwe poindi inogona kuve isina kujairika munharaunda yakaomeswa ichiri munzvimbo ine dataset yakazara. Sezvineiwo, poindi iri kure kubva mugore guru inogona kunge iri yeboka rakaomesesa uye inogamuchira zvakajairwa munharaunda mamaki. Kuti boka diki iroro rinofanira kuverengeka serisinganzwisisike zvinoenderana nebasa racho. Chiverengero chemuvakidzani chinodzora chiyero chekuenzanisa. Zvivakidzani zviduku zvinonzwa kune chimiro chakanaka uye ruzha; makuru anogona kuenzanisa mumapoka akasiyana. Distance metric and feature units zvakakosha nekuti LOF inoshandisa vavakidzani vepedyo. Cheresa nhamba dzakasiyana-siyana zvakakodzera uye encode zvikamu pasina kugadzira inorasisa nhamba. Sarudza marongero zvichienderana nehukuru hweboka hunotarisirwa uye ongorora maitiro ezvibodzwa, zvine machesi akaongororwa. Scikit-Learn inosiyanisa kuonekwa kwekunze, uko LOF inotaridza zvionerero zvisina kujairika mudhatabheti rakaiswa, kubva pakuona zvinyowani, apo modhi inoiswa pane referensi data uye inoshandiswa pamapoinzi matsva. Iyo yekupedzisira inoda hutsva = Chokwadi, uye zvinyorwa zvinoyambira kusashandisa nzira dzezvibodzwa padanho rekudzidzisa mune iyoyo modhi. Zvibodzwa zveLOF hazvina kuverengerwa zvingangoitika uye zvinogona kushandisa masaini akapidigurwa mumaAPI. Ongorora kana uwandu hwepamusoro kana hudiki hunoratidza kusaenzana kukuru. Nharaunda inomaly humbowo hunoenderana nereferensi yakasarudzwa uye metric, kwete mutongo wakajeka wemhando yedata kana njodzi.
Ipoinzi inogara muhomwe yakati wandei padivi penzvimbo yakamanikana. Density yayo yemunharaunda yakadzikira pane yevavakidzani vayo, saka chiyero cheLOF chinokwira kunyangwe nzvimbo yayo yakakwana isiri kure kure nedata rese.
Maonero ekufungidzira ari kure negore guru asi mukati meboka diki rakadzikira. LOF inogona kuiona seyakajairwa munharaunda nekuti ine vavakidzani vane hunyoro hwakafanana, ichiratidza musiyano uripo pakati pekushaikwa kwepasirese uye chimiro chekunze.
Chikwata chinoenzanisa n_neighbours values uye zvikero zvenhamba pamberi pekuverenga chinhambwe. Nevavakidzani vashomanene, zvivako zviduku zvingatonga; vakawandisa vanogona kudzima mapoka emunharaunda.
Kuti zvionekwe zvitsva, muongorori anoenderana neLOF pane referenzi data ine novelty=Chokwadi uye anowana gare gare mapoinzi asingaonekwe. Ivo vanodzivirira kushandisa hutsva hwezvibodzwa nzira kukudzidziswa kwakaiswa pasi peiyi modhi, vachitevera raibhurari yakanyorwa musiyano.
LOF inogona kutsigira kunyatsoongorora kana mishumo isinganzwisisike ichiratidza chiyero chevavakidzani chakasarudzwa, preprocessing uye gwara renzvimbo. Zvikwata zvinofanirwa kuenzanisa zviitiko zvisina kujairika nemapoka avo chaiwo evezera kuti vaone kana kutsauka kwenzvimbo kwakakosha. Sezvo kugoverwa kwedata kuchichinja, gadziridza referensi seti nekufunga uye tarisa mamakisi ekuchinja nechikamu. Kana kucherechedzwa kwechitsva kuchishandiswa, chengetedza kudzidziswa uye nzira dzemagoro dzeramangwana dzakasiyana. Zvirinani zvinosanganisirwa zvinogona kutsanangura kuti ndevapi vavakidzani vakapa zvibodzwa, ukuwo zvichijekesa kuti kuenzanisa kwevavakidzani hakusi iko kukonzeresa kana kuongororwa kwenjodzi.
Tsamba yemahara
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
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
Local Outlier Factor (LOF) inocherekedza zvakacherechedzwa nekuenzanisa density yayo yemunharaunda nekuwanda kwevavakidzani vepedyo. Inogona kuona mapoinzi asina kujairika mukati menzvimbo yakaomesesa kana yakakora, asi kuverenga kwevavakidzani, kuyera chinhambwe uye mutsauko uripo pakati pekuonekwa kwekunze nekuonekwa kwezvitsva zvinokanganisa kushandiswa kwayo.
LOF inofananidza hukama hwevavakidzani nehuwandu hwepoindi; reshiyo hombe inoratidza hushoma.
LOF inzvimbo; poindi munharaunda yayo yakaomesesa inogona kuve yakajairika munharaunda kunyangwe kuparadzana kwepasirese.
k inotara saizi yenharaunda uye nokudaro chiyero chenzvimbo yekuenzanisa kwehuwandu.
Scale misiyano inogona kutonga chinhambwe metric uye shandura izvo zvinoonekwa vavakidzani.
Novelty modhi inoenderana nereferensi data uye inoongorora zvisingaonekwe zvakaonekwa, zvakasiyana nekuzivisa kunze mukati memuenzaniso wakakodzera.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
Building Local LLM Apps with Ollama
Tekinoroji