Puede medir la precisión de su query de MongoDB Vector Search evaluando cuán estrechamente los resultados de una ANN búsqueda coinciden con los resultados de una ENN búsqueda contra vectores cuantizados utilizando los mismos criterios de query. Es decir, puede comparar los resultados de la búsqueda ANN con los resultados de la búsqueda ENN y medir con qué frecuencia los resultados de la búsqueda ANN incluyen a los vecinos más cercanos en los resultados de la búsqueda ENN.
Casos de uso
Es posible que quieras medir la precisión de los resultados si tienes alguna de las siguientes condiciones:
Vectores cuantizados
Gran cantidad de vectores
Vectores de baja dimensión
Requisitos previos
Para probar los ejemplos de esta página, requiere lo siguiente:
Un clúster con, opcionalmente, el conjunto de datos de muestra.
➤ Utilice el menú desplegable Seleccione su idioma para seleccionar la interfaz que desea utilizar para crear su índice.
Procedimientos
Para evaluar la precisión de su $vectorSearch query results, debes hacer lo siguiente:
Cree un índice de búsqueda vectorial de MongoDB en el campo vectorial y en cualquier otro campo por el cual desee filtrar previamente los datos.
Recomendamos usar vectores cuantizados para mejorar el almacenamiento de los vectores y la velocidad de las queries. Si no tienes vectores cuantizados, puedes habilitar la cuantización automática al realizar la indexación de tu campo tipo
vector.Construya y ejecute la consulta ENN seguida de la consulta ANN.
Compara los resultados de la consulta ANN con los resultados de la consulta ENN para evaluar las semejanzas y diferencias en los resultados.
Esta sección demuestra cómo realizar los pasos 3 anteriores sobre datos en la colección sample_mflix.embedded_movies. Si no desea usar el conjunto de datos de muestra, puede realizar los procedimientos con sus propios datos.
Cree el índice de búsqueda vectorial de MongoDB
Esta sección demuestra cómo crear un índice de Búsqueda Vectorial de MongoDB para ejecutar ANN y ENN MongoDB Vector Search queries.
En Atlas, ve a Search & Vector Search página para su clúster.
Puedes ir a la página de búsqueda de MongoDB desde la opción Search & Vector Search o desde el Data Explorer.
Si aún no aparece, se debe seleccionar la organización que contiene el proyecto en el menú Organizations de la barra de navegación.
Si aún no se muestra, seleccione su proyecto en el menú Projects de la barra de navegación.
En la barra lateral, haz clic en Search & Vector Search en la sección Database.
Si no tienes clústeres, haz clic en Create cluster para crear uno. Para obtener más información, consulta Crear un clúster.
Si el proyecto tiene varios clústeres, se debe seleccionar el clúster que se desea usar en el menú desplegable Select cluster y luego se debe hacer clic en Go to Search.
Aparece la página de Búsqueda y Búsqueda Vectorial.
Si aún no aparece, se debe seleccionar la organización que contiene el proyecto en el menú Organizations de la barra de navegación.
Si aún no se muestra, seleccione su proyecto en el menú Projects de la barra de navegación.
En la barra lateral, haz clic en Data Explorer en la sección Database.
Expanda la base de datos y seleccione la colección.
Haga clic en la pestaña Indexes para la colección.
Haga clic en el enlace Search and Vector Search en el banner.
Aparece la página de Búsqueda y Búsqueda Vectorial.
Se debe iniciar la configuración del índice.
Realiza las siguientes selecciones en la página y luego haz clic en Next.
Search Type | Seleccione el tipo de índice Vector Search. |
Index Name and Data Source | Especifique la siguiente información:
|
Configuration Method | For a guided experience, select Visual Editor. To edit the raw index definition, select JSON Editor. |
IMPORTANTE:
El índice de MongoDB Search se llama default por defecto. Si se mantiene este nombre, el índice será el índice de búsqueda por defecto para cualquier query de MongoDB Search que no especifique una opción de index diferente en sus operadores. Si se crean varios índices, recomendamos mantener una convención de nomenclatura coherente y descriptiva en todos los índices.
Especifique la definición del índice.
Ejemplo
Esta definición de índice indexa el campo plot_embedding_voyage_3_large como el tipo vector con quantization binario automático habilitado y el campo genres como el tipo filter en un índice de MongoDB Vector Search. El campo plot_embedding_voyage_3_large contiene embebidos creados usando el modelo de embebidos de Voyage IA, voyage-3-large. La definición del índice especifica 2048 dimensiones vectoriales y mide la distancia utilizando la función de similitud dotProduct.
Atlas detecta automáticamente los campos que contienen embeddings vectoriales, así como sus dimensiones correspondientes. Para la colección sample_mflix.embedded_movies, selecciona el campo plot_embedding_voyage_3_large.
Para configurar el índice, haga lo siguiente:
Selecciona Dot Product del menú desplegable Similarity Method.
Haz clic en Advanced, luego selecciona la cuantificación Binary en el menú desplegable.
En la sección Filter Field, especifica el campo
genrespor el que deseas filtrar los datos.
Pegue la siguiente definición de índice en el editor JSON:
1 { 2 "fields": [ 3 { 4 "numDimensions": 2048, 5 "path": "plot_embedding_voyage_3_large", 6 "similarity": "dotProduct", 7 "type": "vector", 8 "quantization": "binary" 9 }, 10 { 11 "path": "genres", 12 "type": "filter" 13 } 14 ] 15 }
Verifique el estado.
El índice recién creado aparece en la pestaña Atlas Search. Mientras se construye el índice, el campo Status muestra Build in Progress. Cuando se termina de construir el índice, el campo Status muestra Active.
Nota
Las colecciones más grandes tardan más en indexarse. Se recibirá una notificación por correo electrónico cuando el índice haya terminado de construirse.
Conéctate a la implementación mediante mongosh.
En la terminal, conectarse a la implementación alojada en la nube de Atlas o a una implementación local desde mongosh. Para obtener instrucciones detalladas sobre cómo conectarse, consultar Conectar a una implementación.
Ejecuta el método db.collection.createSearchIndex() para crear el índice.
db.embedded_movies.createSearchIndex( "vector_index", "vectorSearch", { "fields": [ { "numDimensions": 2048, "path": "plot_embedding_voyage_3_large", "similarity": "dotProduct", "type": "vector", "quantization": "binary" }, { "path": "genres", "type": "filter" } ] } } )
vector_index
Ejecuta las consultas
No puedes ejecutar consultas de MongoDB Vector Search en la Interfaz de Usuario de Atlas Search Tester. Use mongosh o un controlador compatible para ejecutar la consulta.
Esta sección demuestra cómo ejecutar las consultas ENN y ANN en la colección indexada.
Prepara las incrustaciones del query.
Guarde las siguientes incrustaciones en un archivo llamado query-embeddings.js:
TIME_TRAVEL_EMBEDDING=[-0.034731735,0.008558298,-0.0153717,-0.029912498,0.011549547,0.010261648,-0.011964999,-0.023265276,0.010303194,-0.006896493,-0.00054528,0.003926015,-0.025757983,0.027419789,0.001199616,-0.036227357,-0.005297005,0.021935832,0.010303194,-0.019193852,0.025093261,-0.040049512,-0.033900831,-0.011466458,-0.01827986,-0.0153717,0.023265276,0.007727395,0.000114249,0.005317777,-0.043871664,-0.02127111,-0.019609304,0.016368784,-0.004756918,0.003552109,0.006522586,-0.005400868,-0.015620971,-0.034565553,-0.018695312,-0.023099095,0.050851244,-0.034731735,0.004819236,0.022268193,-0.095719993,0.05517194,-0.046198189,-0.036393538,0.007187308,-0.02459472,-0.036725901,0.009472291,0.019027673,0.020938748,-0.011051006,0.027087428,0.04586583,-0.022600554,-0.05517194,0.044204023,0.01213118,0.047859997,-0.03938479,0.002928932,0.002056484,0.019443123,-0.028583053,0.013543714,0.022932915,0.011632638,0.004923099,0.000389486,0.020024756,-0.024096178,-0.022766734,0.011217186,-0.003198975,0.007104218,-0.047195274,-0.013377533,0.013294443,0.024096178,-0.056501385,-0.026755067,-0.008433662,-0.001911076,0.007976666,-0.008101301,-0.014042255,0.008641388,-0.02176965,0.010012378,-0.000607598,-0.024927082,0.024927082,-0.018612221,-0.001184036,0.005567048,0.001324251,-0.019526213,-0.023597637,0.060489718,-0.010178559,-0.019609304,0.004112968,-0.011217186,-0.031574301,-0.008766023,0.005483958,-0.061819162,-0.023431456,-0.040714234,0.015039339,0.026422706,0.016202603,0.004653055,0.041046593,-0.018030589,0.040381871,-0.002638116,0.013045172,0.004216831,0.005650138,0.027419789,0.003926015,-0.028749233,0.004798463,-0.030244859,0.063813329,0.007145763,-0.017448956,0.025591804,-0.045201108,0.010718645,0.002804297,0.014291527,0.04586583,-0.015205519,-0.021603471,-0.035230275,0.00760276,0.033236109,0.016534964,-0.043206941,-0.003115885,-0.026256526,0.005940954,0.016534964,0.024262359,-0.001630647,0.028084511,-0.012795902,0.007270399,0.001381376,-0.009763107,-0.006896493,0.008433662,-0.019360034,0.000386889,0.030411039,0.025591804,0.010469374,0.037722982,-0.001147684,-0.005400868,0.052845411,-0.052513052,0.00768585,-0.004299921,0.00922302,0.011881908,0.012962082,-0.068798743,0.003593654,0.020938748,-0.013792985,-0.034565553,-0.007519669,-0.04021569,-0.020689478,0.006273315,0.046862911,0.006107135,0.002638116,-0.013792985,-0.005400868,-0.020274026,0.007644305,-0.010801735,0.026422706,0.043871664,0.003780607,0.010261648,-0.064145692,0.011881908,-0.009056839,0.009347656,-0.02459472,0.026422706,0.033236109,0.041212775,0.019027673,-0.00315743,0.004424557,0.020689478,-0.0153717,-0.015205519,-0.034897912,0.020274026,0.016867325,0.040714234,-0.022766734,-0.010967916,0.026256526,0.007062673,-0.015953332,-0.007727395,0.031574301,-0.002887387,-0.00614868,0.004569965,0.019027673,0.012878992,0.011798819,0.004258377,-0.019193852,-0.021437289,-0.021603471,0.000301202,-0.051183607,-0.004985416,-0.030078677,0.012629721,0.065142773,-0.031740483,-0.021104928,-0.03938479,-0.003365156,-0.016036423,0.036393538,0.009804652,-0.018612221,0.060489718,-0.003697517,0.000547876,0.063480966,0.02758597,0.010053922,-0.003655972,-0.001485239,0.018362951,0.021104928,-0.003905243,0.019443123,-0.002658889,-0.00380138,-0.013626805,0.035894997,0.035396457,-0.005691683,0.002762751,0.012878992,-0.009596926,-0.009970833,-0.015953332,0.022434372,0.00614868,-0.021188019,0.001557943,-0.020190936,0.009763107,0.017448956,0.006730312,0.005567048,0.019692395,-0.00218112,-0.016867325,0.006854947,0.007976666,0.019193852,0.040880412,0.007353489,-0.02127111,-0.031906664,-0.026755067,-0.017947499,0.040381871,0.042209856,0.00913993,-0.0307434,-0.017781317,-0.015039339,0.03057722,0.017532047,0.0187784,-0.060822077,0.002928932,-0.026422706,-0.005899409,0.039717149,0.026588887,-0.000971118,0.004923099,-0.013626805,0.0187784,-0.031408124,-0.000695881,0.050851244,-0.014457707,-0.007311944,-0.001293092,-0.002139574,-0.019276943,0.00290816,0.019360034,-0.017781317,0.002160347,0.016618054,-0.006522586,0.011798819,0.029247776,-0.02775215,0.010344739,-0.018362951,-0.036725901,-0.015870241,0.015704062,-0.012463541,0.02459472,-0.024096178,0.001152877,-0.031408124,0.025425622,0.027087428,0.00922302,0.034565553,0.015704062,-0.020689478,-0.00517237,-0.014706978,-0.001589101,0.026090344,0.014956249,0.011715728,0.004299921,-0.00913993,0.022434372,-0.03705826,0.048524719,-0.030411039,0.008433662,0.017033506,-0.000511525,-0.031408124,0.005940954,-0.012962082,-0.031574301,0.017448956,0.010178559,-0.011383367,-0.020107845,-0.005151597,0.006647222,0.013128263,0.007145763,0.008059756,-0.045201108,-0.004943871,0.015787151,-0.045201108,-0.020772567,-0.020274026,0.028250692,-0.024262359,-0.004424557,0.009804652,0.000472576,-0.005691683,0.001443693,-0.013294443,0.001412535,0.013211353,-0.01213118,-0.002118802,0.017781317,-0.007353489,-0.031075761,-0.004923099,0.011383367,-0.004486875,-0.010178559,0.016618054,0.014457707,0.023763817,-0.02459472,-0.00388447,0.012546631,-0.007519669,0.015704062,-0.014291527,0.009680017,-0.035562634,0.023763817,0.053510133,-0.0555043,-0.003572882,0.022102011,0.021603471,-0.017282777,-0.001474852,-0.043539301,0.007810486,-0.025757983,-0.005400868,0.029912498,-0.00760276,0.014125346,0.030909581,-0.03340229,-0.009680017,0.018030589,0.008849114,0.03057722,0.019775484,0.014125346,0.031906664,-0.03057722,-0.027087428,-0.023597637,-0.022434372,-0.012878992,0.016285693,-0.021603471,-0.029746316,0.029746316,0.020357117,0.006314861,-0.001158071,0.028749233,-0.045201108,0.011383367,0.011134096,-0.021437289,-0.035728816,0.001827986,0.008267482,-0.057498466,0.01213118,-0.01213118,-0.040548053,0.010718645,0.004798463,-0.004881553,-0.019526213,-0.008558298,0.0059825,-0.000262254,-0.017615138,0.005193142,0.019692395,-0.00198378,-0.002845842,0.012546631,0.006107135,-0.008225936,-0.008890659,0.015870241,0.00517237,0.002596571,-0.010427829,-0.019110762,0.024262359,0.012048089,-0.032405205,0.006522586,0.013211353,0.013211353,-0.038221523,-0.007727395,-0.008267482,-0.019276943,0.001474852,0.031408124,-0.035562634,0.017532047,-0.023431456,-0.015454791,-0.011383367,0.016534964,-0.02176965,0.008682934,0.027253609,0.020190936,-0.0247609,-0.007311944,0.009555381,-0.01852913,-0.011632638,0.011549547,0.027419789,-0.034067012,-0.01229736,0.0307434,0.003946788,0.0046946,0.037722982,0.03057722,-0.010427829,0.002284982,0.033236109,0.030078677,-0.013377533,0.007395034,-0.012048089,0.040714234,-0.028749233,-0.000102565,-0.0059825,-0.041046593,0.017698228,-0.006356406,0.003178203,0.009056839,0.023099095,0.00606559,0.011881908,-0.02127111,-0.001126912,-0.027087428,0.011134096,0.001204809,-0.017033506,0.011051006,-0.014374617,0.017864408,0.023431456,-0.002077257,-0.026755067,-0.043871664,0.025757983,-0.006190225,0.001152877,0.011798819,-0.024262359,0.006564131,-0.070128188,-0.004362239,0.012962082,-0.013626805,-0.001402148,-0.012214269,0.011217186,-0.015953332,0.015787151,0.011134096,0.027253609,0.024262359,-0.048192356,0.009970833,0.018944582,-0.00517237,0.021935832,0.02775215,0.003406701,-0.010884825,0.075113602,-0.015953332,0.007727395,0.026755067,-0.006190225,-0.012712811,0.013377533,0.005940954,-0.008309027,0.02459472,0.002316141,-0.022434372,-0.012712811,0.03057722,-0.015787151,0.026755067,-0.001069787,0.03988333,-0.003697517,0.039550968,-0.019027673,-0.0059825,-0.00031029,-0.012546631,-0.003614427,0.007478124,0.005525503,0.032571387,-0.011798819,-0.011466458,-0.00606559,-0.011798819,0.018446039,0.007976666,0.018944582,-0.02176965,0.026588887,-0.006273315,-0.012463541,-0.007395034,0.012048089,-0.029247776,0.015454791,-0.007145763,0.006481041,-0.015620971,-0.00388447,-0.025757983,-0.001651419,-0.032903746,-0.005068507,0.03938479,0.003926015,0.004715373,0.022600554,-0.012546631,0.022932915,0.007810486,0.040714234,0.019941665,0.013543714,0.003406701,0.010884825,-0.03988333,0.042209856,-0.022766734,0.027419789,-0.029580137,0.043206941,0.022932915,0.021104928,-0.056833744,0.005193142,0.036061179,-0.012878992,0.008516753,-0.02758597,-0.030244859,-0.011798819,0.001111332,-0.014125346,-0.014125346,0.019027673,0.029081594,0.018861491,0.013626805,0.06846638,0.023099095,0.041378956,0.001599488,-0.028749233,0.017781317,0.016285693,0.021603471,-0.018113678,0.011300277,-0.032239024,0.022434372,-0.02459472,-0.013626805,0.005483958,0.013460624,-0.031574301,-0.015620971,0.016451873,0.014790068,-0.008849114,0.011134096,0.00461151,0.015122429,0.036227357,0.00206687,0.000877641,0.022102011,-0.028250692,0.022600554,-0.026422706,0.004029878,-0.032072846,0.017116595,0.010884825,0.019609304,0.00614868,0.005733229,0.016119512,0.002866614,-0.014540797,0.012463541,-0.003905243,0.003759835,-0.000485559,-0.022766734,-0.016285693,0.037722982,0.009513836,0.001506011,0.011964999,0.004029878,0.019941665,-0.000965924,0.002129188,0.015205519,0.071125269,0.022932915,0.005940954,-0.00044661,0.010220103,-0.03423319,-0.016285693,-0.016867325,-0.000659529,-0.008018211,-0.011383367,0.000016634,0.004071423,-0.029413955,0.019941665,-0.00913993,-0.024096178,0.010635555,0.010594009,0.001547556,0.036227357,-0.030078677,0.020772567,0.022268193,-0.014125346,0.008766023,-0.012962082,-0.007187308,0.017033506,-0.007187308,-0.015205519,-0.005608593,0.044536386,-0.001235968,0.007852031,0.001599488,0.005857864,-0.005940954,-0.010510919,-0.005567048,0.006730312,0.016285693,-0.010801735,-0.024428539,0.015122429,-0.02176965,0.01528861,-0.007436579,0.00226421,-0.004715373,0.004507647,0.004341467,0.005525503,-0.031075761,-0.005899409,0.037556801,0.014873158,-0.000342747,0.009970833,-0.019443123,0.023597637,-0.012048089,-0.025259443,0.006024044,-0.01827986,0.010012378,0.016784234,0.013211353,-0.005400868,-0.024428539,-0.02176965,-0.035230275,0.009347656,0.028583053,-0.015704062,-0.017781317,0.00226421,0.001199616,-0.003385928,0.008267482,0.002326528,0.022434372,-0.020190936,-0.015787151,0.000789358,0.031241942,0.011300277,0.001506011,-0.023265276,-0.010967916,0.009056839,0.011300277,-0.030244859,0.007478124,0.001111332,-0.035894997,0.0153717,0.002700434,0.021104928,0.010884825,-0.003344383,0.00768585,0.010386284,0.00452842,-0.014706978,0.028084511,0.013377533,0.014873158,0.046862911,-0.015454791,0.021188019,0.013959166,0.012629721,0.025924165,-0.018695312,-0.00922302,-0.0093892,0.007727395,0.036892079,0.007228854,-0.01229736,0.029247776,-0.004943871,-0.027253609,-0.008433662,0.043206941,0.002825069,0.028583053,-0.023431456,0.034897912,-0.041545134,-0.016534964,0.003053567,-0.012712811,0.002741979,-0.007187308,-0.025093261,-0.045201108,-0.004424557,-0.016618054,-0.008890659,0.008018211,-0.05184833,-0.019526213,-0.013377533,-0.010469374,0.030244859,-0.005068507,0.051183607,0.005483958,-0.006024044,0.035064094,-0.011134096,0.014956249,0.002284982,0.001724123,-0.01229736,0.012629721,0.010261648,0.014540797,0.048857078,-0.029580137,-0.024927082,-0.008350573,-0.03988333,0.000939959,0.013543714,0.013626805,-0.021437289,-0.012962082,0.006771857,0.013709894,-0.0059825,0.035396457,-0.006439496,-0.029580137,0.0046946,0.019609304,-0.007270399,0.014291527,-0.015620971,0.00118923,-0.00760276,-0.017199686,0.023265276,0.026588887,-0.030078677,-0.016701145,-0.025757983,0.004964644,0.026588887,0.043206941,0.011051006,-0.009846197,0.028915415,0.031574301,0.023763817,0.009264565,-0.008433662,-0.035064094,-0.000579035,-0.0247609,0.014125346,0.016618054,0.028749233,-0.052513052,-0.016867325,-0.01238045,0.002741979,0.013709894,0.010718645,0.013626805,0.009596926,-0.004403784,-0.02758597,-0.000945152,0.000420645,0.003759835,0.012546631,-0.011881908,0.008392117,0.012795902,0.005483958,-0.009763107,0.006397951,-0.010801735,0.012795902,-0.03938479,0.005733229,0.005733229,-0.000433627,0.015454791,0.002357686,-0.006564131,0.030244859,-0.024428539,0.016036423,0.014291527,-0.004964644,0.029413955,0.040381871,0.012629721,-0.033568468,-0.026422706,-0.037889164,-0.034399372,-0.03423319,0.021935832,0.004133741,-0.014623888,-0.013543714,-0.05517194,0.004736145,0.006314861,0.00006037,0.006356406,0.003323611,-0.010344739,0.007062673,0.005899409,-0.00623177,-0.001973394,-0.0555043,0.011881908,0.001350217,-0.033069927,-0.026921248,0.022268193,0.028583053,-0.021021837,0.010884825,0.019692395,-0.005442413,0.031574301,-0.014956249,0.01238045,-0.006356406,0.006273315,-0.003095113,-0.014540797,-0.02176965,0.005006189,-0.002658889,0.042542219,-0.02176965,0.017199686,-0.016701145,-0.001599488,0.016950415,-0.021188019,0.017864408,0.023763817,-0.000669915,0.025093261,0.021104928,0.008807569,0.037390623,-0.025591804,-0.003178203,-0.001319058,0.020523297,0.005255459,0.019276943,-0.00226421,0.00760276,-0.057166107,-0.006896493,-0.034067012,0.043871664,0.038221523,0.008101301,0.03988333,0.015870241,0.000955538,-0.004299921,-0.002928932,-0.002118802,-0.020523297,-0.001168457,-0.011134096,-0.000685495,0.003323611,0.011549547,0.034565553,0.029247776,-0.029746316,0.005213914,0.019110762,-0.003302838,0.026422706,0.028915415,-0.036227357,0.033236109,0.038387705,-0.035230275,0.004071423,-0.021935832,0.002928932,0.000976311,0.000527104,-0.006854947,-0.003822153,-0.001199616,0.019858574,-0.002762751,0.039052427,-0.008641388,0.032239024,-0.002295369,0.035396457,0.044536386,-0.029413955,0.025093261,-0.03423319,-0.016867325,-0.008849114,0.008433662,-0.004486875,0.017033506,0.006730312,-0.008599843,-0.008225936,-0.024428539,0.006564131,-0.007561215,-0.032072846,-0.019941665,0.035396457,0.019276943,0.010261648,0.005857864,0.032239024,-0.044204023,-0.018944582,0.002409618,0.032903746,0.05517194,-0.03655972,0.007976666,0.030909581,-0.023929998,0.016368784,0.01528861,-0.00768585,0.02176965,0.013626805,-0.02459472,0.04021569,-0.032737568,0.006854947,-0.011383367,0.014873158,-0.02176965,0.00243039,0.0093892,0.0093892,-0.029580137,0.019858574,0.01827986,0.024428539,0.017864408,-0.028250692,-0.001111332,0.056169022,0.007478124,-0.010718645,0.041046593,-0.015704062,0.034731735,0.002523867,-0.032571387,0.004341467,-0.023597637,-0.011881908,-0.035562634,0.006688767,0.007810486,-0.012712811,0.022600554,0.03057722,0.022600554,0.010552464,0.0307434,-0.009638472,0.02176965,-0.018030589,0.024262359,-0.036227357,-0.020772567,0.001641033,-0.022932915,-0.014623888,0.018362951,0.002575798,0.006190225,-0.011051006,0.021021837,0.019110762,0.02127111,-0.028583053,-0.052180689,-0.014291527,-0.010552464,0.036393538,0.042542219,-0.04586583,-0.001869531,0.008350573,-0.008516753,-0.020772567,0.000294711,0.015704062,-0.014457707,-0.020772567,0.008766023,-0.026588887,-0.004736145,-0.028084511,-0.007519669,0.010552464,-0.016534964,0.006190225,0.012962082,-0.016618054,0.012546631,0.02459472,0.022932915,0.020440206,-0.027918331,-0.008059756,0.020689478,-0.014623888,-0.011466458,-0.006896493,-0.020024756,-0.031408124,0.021603471,0.007270399,-0.03057722,0.008350573,-0.021437289,0.00072704,-0.043871664,0.006314861,-0.017199686,0.02176965,0.024262359,-0.020357117,-0.000542683,-0.005213914,0.001963008,-0.00064395,-0.022434372,0.022102011,-0.006688767,-0.028583053,0.002191506,-0.005047734,0.002368073,0.014956249,0.023929998,-0.003302838,-0.032239024,0.022268193,-0.013377533,-0.010801735,0.003676744,0.009015295,-0.039550968,0.010884825,-0.033568468,0.013709894,-0.029413955,-0.006356406,-0.020274026,0.023597637,0.030909581,0.02176965,0.016285693,0.045533467,-0.024096178,-0.030909581,-0.026422706,0.002783524,-0.010594009,0.004362239,-0.070792913,0.009472291,-0.022102011,0.011134096,-0.017448956,-0.011549547,-0.056833744,0.00082571,0.026588887,-0.013709894,0.002575798,0.02176965,-0.000568649,-0.007270399,0.004279149,-0.042874578,-0.026588887,0.016784234,0.036725901,-0.028915415,-0.009513836,0.017448956,0.002035712,-0.007228854,0.011383367,0.011134096,0.028915415,0.0153717,-0.027087428,0.043871664,-0.005089279,0.006314861,0.014291527,-0.003240521,0.025924165,-0.001230775,-0.015454791,-0.012629721,0.031740483,-0.039717149,-0.031075761,0.006605676,-0.008641388,-0.032239024,0.037722982,-0.03705826,-0.024096178,0.001911076,0.018196769,-0.007353489,-0.011300277,-0.029081594,0.004590738,-0.018030589,-0.026588887,0.010261648,0.038221523,0.008392117,-0.01213118,0.018362951,-0.034731735,-0.017781317,-0.011632638,0.005255459,0.000851675,0.014208436,-0.000039922,-0.000228498,0.014790068,0.00913993,0.0004544,-0.011798819,-0.020440206,0.005899409,0.008350573,0.006314861,0.040548053,0.003427474,-0.010801735,0.008599843,0.002586185,-0.041212775,-0.016368784,0.020024756,0.000965924,-0.021021837,-0.008475208,0.0307434,0.00760276,0.003614427,0.003489791,-0.025924165,0.000799744,0.013460624,-0.020440206,0.048857078,0.004320694,-0.048857078,0.015039339,-0.029580137,0.025924165,0.018861491,-0.014706978,0.000576439,-0.031241942,0.0307434,0.0153717,0.014706978,0.028084511,-0.01238045,-0.031241942,0.018196769,-0.034897912,0.008142847,0.010718645,0.00922302,0.047859997,-0.00072704,-0.010427829,0.007104218,0.026256526,0.012214269,-0.013377533,-0.05184833,0.005276232,0.021935832,-0.007021128,0.009804652,0.007893575,0.024096178,-0.002357686,0.033900831,-0.031740483,0.034565553,-0.036892079,-0.015454791,0.030411039,0.010552464,-0.022268193,-0.001391762,-0.008184392,-0.008558298,0.008475208,-0.009929287,0.010427829,0.041378956,-0.009555381,-0.008724478,-0.039052427,0.034731735,-0.014291527,0.023099095,0.029081594,0.007519669,0.010967916,-0.008142847,0.006190225,-0.031075761,0.033734649,-0.001672192,0.047859997,-0.022434372,-0.007395034,0.01213118,0.056169022,0.002762751,-0.029413955,-0.000763392,-0.015787151,0.010801735,0.008142847,0.029912498,-0.0018176,0.033236109,-0.046198189,-0.002492708,-0.006730312,0.008807569,-0.03655972,0.009430746,-0.053842496,-0.060489718,0.046862911,0.002783524,-0.0187784,0.000571246,0.00760276,0.002482322,0.001319058,-0.014291527,0.001464466,-0.011632638,-0.012463541,-0.004902326,0.000841289,0.006688767,0.030244859,0.018944582,0.000532297,-0.015620971,0.007104218,0.005608593,0.002035712,-0.023763817,0.003032795,0.010594009,-0.023597637,-0.042376038,-0.005255459,0.001199616,-0.0247609,-0.007893575,-0.011632638,0.013045172,-0.005691683,-0.007104218,0.027419789,-0.004320694,-0.005525503,-0.026090344,0.031408124,-0.012795902,-0.007062673,0.000939959,0.000030185,0.004175286,0.014291527,0.033236109,-0.038720068,0.074116521,-0.019692395,0.001589101,0.013792985,-0.056169022,-0.028749233,-0.001599488,0.004175286,0.014790068,0.00162026,-0.007519669,-0.041378956,0.016534964,-0.003572882,-0.002575798,-0.019526213,-0.00922302,-0.033900831,-0.042043675,-0.014208436,0.010178559,0.017698228,0.032239024,0.00913993,0.009264565,-0.012463541,-0.005857864,-0.015870241,0.004486875,0.018861491,-0.000176567,-0.029912498,0.016784234,0.012546631,0.051183607,0.023597637,0.032903746,0.0153717,-0.013377533,-0.000016634,-0.061486799,-0.034565553,0.016119512,0.00380138,-0.003863698,0.004362239,-0.017532047,-0.002762751,0.000102565,-0.021437289,0.029247776,-0.010718645,-0.015870241,-0.016285693,0.010220103,-0.000373906,0.012962082,0.010137013,-0.007228854,0.02127111,-0.029247776,0.018113678,0.009181475,0.002233051,0.014374617,-0.00396756,0.010801735,0.007644305,0.020855658,0.014790068,0.032737568,-0.037390623,0.003032795,0.010801735,-0.01553788,-0.014790068,0.019526213,-0.017947499,-0.007893575,-0.011964999,-0.00614868,-0.005857864,-0.032072846,-0.025924165,0.001163264,-0.013294443,-0.01553788,0.016701145,-0.013460624,-0.001111332,0.00760276,0.01553788,-0.033734649,0.048192356,-0.003282066,0.031906664,0.002845842,0.003240521,0.017116595,-0.01827986,0.006896493,-0.00760276,-0.009680017,-0.02459472,-0.020689478,-0.053510133,0.00614868,-0.010552464,-0.032405205,-0.0307434,0.025093261,0.003635199,-0.008101301,-0.00606559,-0.007436579,0.00606559,-0.012962082,0.026921248,0.009098385,0.046530552,-0.011632638,0.032571387,-0.033900831,0.009846197,0.002866614,0.032903746,0.008973749,0.012712811,0.040049512,0.013626805,-0.026256526,-0.031408124,0.036227357,0.011964999,-0.006024044,-0.001848759,0.015704062,-0.021188019,-0.035064094,-0.013377533,-0.009721561,-0.01553788,0.008766023,0.005400868,0.004507647,-0.018362951,-0.026588887,-0.00913993,-0.025591804,0.035894997,0.021935832,-0.031906664,-0.000602404,0.026422706,-0.006397951,0.006647222,0.0093892,0.020606387,0.00913993,0.015620971,-0.024096178,0.00063616,-0.006564131,0.01238045,-0.013709894,0.000563456,-0.009887742,0.016618054,-0.003323611,0.000451803,0.001609874,0.008682934,0.025259443,0.020024756,-0.027253609,0.010884825,0.028250692,-0.054839578,0.033568468,-0.004902326,0.003053567,0.020274026,-0.015704062,-0.00614868,-0.063813329,0.002482322,0.009763107,-0.001609874,-0.012214269,0.020107845,0.001921462,0.018695312,-0.004923099,0.007270399,-0.023763817,0.005234687,0.003406701,0.002565412,0.007104218,0.000841289,0.016202603,0.01827986,-0.031075761,-0.035562634,-0.025259443,-0.007021128,0.000641353,-0.033069927,0.010718645,0.005650138,0.024927082,-0.002658889,0.00380138,0.009929287,-0.004258377,-0.039717149,-0.022434372,0.025425622,0.00198378,0.006356406,0.017615138,-0.032072846,0.046862911,-0.026921248,0.005940954,0.021603471,-0.002253824,0.002825069,-0.030411039,-0.003115885,0.023597637,-0.004320694,-0.007852031,0.018030589,-0.008724478,-0.005733229,0.032903746,0.013876075,0.015454791,-0.023597637,0.005151597,-0.035396457,0.02176965,-0.012463541,0.025591804,0.014540797,-0.027918331,0.004154514,0.008724478,0.016036423,-0.015870241,0.005400868,-0.017365867,-0.044868745,-0.000485559,0.020357117,-0.00760276,-0.023265276,-0.012048089,0.008433662,0.018362951,-0.006979583,0.0307434,0.008392117,0.027087428,-0.019360034,0.016119512,0.02127111,0.010801735,0.00299125,0.002949705,0.012463541,-0.000025966,0.015953332,0.029413955,0.020024756,0.003780607,0.022102011,-0.031740483,0.01553788,0.010386284,0.028749233,-0.010884825,0.008682934,-0.003531337,-0.05517194,-0.019360034,-0.009347656,-0.002025325,0.003261293,-0.025425622,-0.01553788,-0.000251867,0.014291527,0.012546631,0.035728816,-0.007062673,-0.006605676,0.000384293,-0.005047734,-0.032571387,-0.021188019,-0.02127111,-0.016036423,0.008475208,-0.004009106,0.014291527,-0.008101301,0.004424557,-0.038221523,-0.019360034,0.015039339,-0.015454791,-0.029580137,0.035728816,0.004466102,-0.000778971,-0.005068507,-0.017781317,0.00477769,0.001838372,0.030244859,0.01213118,-0.022932915,-0.005359322,0.037390623,0.005899409,0.002046098,0.037889164,0.016701145,0.010303194,0.02127111,-0.009513836,-0.022268193,-0.005650138,-0.00388447,0.016534964,-0.023265276,-0.00054528,0.004819236,0.004715373,-0.001178843,-0.051183607,-0.00614868,-0.010552464,-0.002741979,-0.009181475,0.023597637,0.019193852,0.017199686,-0.036393538,-0.00243039,-0.015870241,-0.014706978,-0.00145408,0.016368784,-0.011632638,-0.014623888,-0.01229736,-0.01553788,0.040880412,0.023929998,-0.014623888,0.002648502,0.031906664,-0.033734649,-0.026755067,0.002783524,0.005359322,0.009970833,0.001412535,0.016950415,0.016285693,-0.006730312,-0.02459472,0.050851244,-0.001827986,-0.020855658,0.020938748,0.004071423,-0.021603471,-0.007852031,-0.023929998,-0.029912498,-0.003365156,0.017365867,-0.010427829,-0.011715728,0.014956249,0.011383367,0.032405205,-0.028583053,-0.017448956,0.018446039,0.017615138,0.035728816,-0.010095468,-0.00254464,0.010012378,0.028250692,-0.020855658,-0.002305755,-0.001002276,-0.014125346,-0.007021128,-0.028583053,-0.045533467,-0.02758597,-0.020440206,0.001350217,0.010053922,0.020689478,-0.017615138,0.026422706,0.040880412,0.012463541,-0.010718645,-0.014706978,0.068134025,0.038720068,0.047859997,-0.012546631,0.015704062,-0.002087643,-0.010303194,0.014790068,0.018612221,0.007395034,-0.014790068,-0.017864408,-0.005068507,-0.054507218,0.004902326,-0.004050651,0.021603471,0.019775484,-0.024262359,-0.012795902,0.021935832,-0.004009106,-0.039717149,0.037556801,-0.016701145,-0.025757983,0.005483958,-0.005110051,-0.021935832,-0.003406701,0.010594009,0.015787151,-0.049854163,0.007727395,-0.008392117,-0.017199686,0.009970833,-0.008849114,-0.013876075,-0.0059825,-0.015870241,-0.007104218,0.028250692,-0.029081594,0.026921248,0.00299125,-0.017781317,0.042542219,0.018196769,0.052845411,-0.004819236,-0.014125346,0.02459472,-0.011715728,0.015787151,-0.005774774,0.004902326,-0.004964644,-0.02758597,-0.013959166,-0.033568468,-0.027918331,-0.017698228,0.003489791,-0.020024756,-0.021603471,0.019360034,0.028084511,-0.002503094,-0.018861491,-0.002295369,0.050851244,-0.020689478,-0.000459593,-0.026090344,0.002783524,-0.005899409,-0.026921248,-0.0093892,-0.004112968,0.031574301,0.003926015,-0.032903746,-0.046198189,-0.019027673,-0.00913993,0.030411039,-0.019443123,0.001963008,-0.005193142,0.010884825,-0.02127111,-0.025259443,0.032737568,0.00089322,0.003282066,0.001713737,-0.006439496,0.016867325,-0.031574301,0.031075761,-0.009970833,0.022600554,-0.023597637,-0.014956249,0.004009106,0.00198378,0.026588887,-0.023431456,-0.023763817,-0.013294443,-0.029746316,0.001381376,-0.042874578,-0.00913993,0.014873158,0.016202603,0.012878992,-0.006024044,0.009638472,0.010552464,-0.017033506,-0.027087428,0.044536386,-0.038055345,0.001329444,-0.019609304,0.023597637,-0.043206941,0.040049512,0.017615138,0.046862911,0.02127111,0.013294443,-0.039550968,-0.018861491,-0.019609304,-0.033734649,0.00623177,-0.017199686,0.041212775,-0.017781317,-0.024262359,0.054507218,-0.009721561,0.005816319,-0.00206687,-0.008766023,0.017365867,-0.000737426,0.018362951,-0.023597637,-0.019110762,0.021935832,0.041545134,-0.020357117,-0.017615138,0.044868745,-0.018030589,-0.032405205,-0.050186522,-0.014540797,0.005213914,-0.006688767,0.047527634,0.040714234];
Conéctese al clúster usando mongosh.
Abre una ventana de terminal y conéctate a tu clúster utilizando mongosh. Para aprender más, consulte Conectar a un clúster a través de mongosh.
Cambie a su base de datos.
Ejemplo
Utilice la sample_mflix base de datos. Para cambiar a la sample_mflix base de datos, ejecute el siguiente comando en mongosh el indicador:
use sample_mflix
switched to db sample_mflix
Ejecuta tu query ENN ENN.
Ejemplo
Copia y pega la siguiente consulta de muestra en tu terminal y luego ejecútala con mongosh. mongosh podría presentar un ligero retraso al pegar la query debido a la cantidad de caracteres en la vectorización.
1 db.embedded_movies.aggregate([ 2 { 3 "$vectorSearch": { 4 "index": "vector_index", 5 "path": "plot_embedding_voyage_3_large", 6 "filter": { 7 "$and": [ 8 { 9 "genres": { "$eq": "Action" } 10 }, 11 { 12 "genres": { "$ne": "Comedy" } 13 } 14 ] 15 }, 16 "queryVector": TIME_TRAVEL_EMBEDDING, 17 "exact": true, 18 "limit": 10 19 } 20 }, 21 { 22 "$project": { 23 "_id": 0, 24 "plot": 1, 25 "title": 1, 26 "genres": 1, 27 "score": { $meta: "vectorSearchScore" } 28 } 29 } 30 ])
[ { plot: 'A psychiatrist makes multiple trips through time to save a woman that was murdered by her brutal husband.', genres: [ 'Action', 'Crime', 'Drama' ], title: 'Retroactive', score: 0.760047972202301 }, { plot: 'A time-travel experiment in which a robot probe is sent from the year 2073 to the year 1973 goes terribly wrong thrusting one of the project scientists, a man named Nicholas Sinclair into a...', genres: [ 'Action', 'Sci-Fi' ], title: 'A.P.E.X.', score: 0.7576861381530762 }, { plot: 'An officer for a security agency that regulates time travel, must fend for his life against a shady politician who has a tie to his past.', genres: [ 'Action', 'Crime', 'Sci-Fi' ], title: 'Timecop', score: 0.7576561570167542 }, { plot: 'A reporter, learning of time travelers visiting 20th century disasters, tries to change the history they know by averting upcoming disasters.', genres: [ 'Action', 'Sci-Fi', 'Thriller' ], title: 'Thrill Seekers', score: 0.7509932518005371 }, { plot: 'Lyle, a motorcycle champion is traveling the Mexican desert, when he find himself in the action radius of a time machine. So he find himself one century back in the past between rapists, ...', genres: [ 'Action', 'Adventure', 'Sci-Fi' ], title: 'Timerider: The Adventure of Lyle Swann', score: 0.7502642869949341 }, { plot: 'Hoping to alter the events of the past, a 19th century inventor instead travels 800,000 years into the future, where he finds humankind divided into two warring races.', genres: [ 'Sci-Fi', 'Adventure', 'Action' ], title: 'The Time Machine', score: 0.7502503395080566 }, { plot: 'A modern aircraft carrier is thrown back in time to 1941 near Hawaii, just hours before the Japanese attack on Pearl Harbor.', genres: [ 'Action', 'Sci-Fi' ], title: 'The Final Countdown', score: 0.7469133734703064 }, { plot: "Ba'al travels back in time and prevents the Stargate program from being started. SG-1 must somehow restore history.", genres: [ 'Action', 'Adventure', 'Drama' ], title: 'Stargate: Continuum', score: 0.7468316555023193 }, { plot: 'With the help of his uncle, a man travels to the future to try and bring his girlfriend back to life.', genres: [ 'Action', 'Adventure', 'Drama' ], title: 'Love Story 2050', score: 0.7420939207077026 }, { plot: "Captain Picard and his crew pursue the Borg back in time to stop them from preventing Earth's first contact with an alien species. They also make sure that Zefram Cochrane makes his famous maiden flight at warp speed.", genres: [ 'Action', 'Adventure', 'Sci-Fi' ], title: 'Star Trek: First Contact', score: 0.7356286644935608 } ]
Esta query utiliza las siguientes etapas de pipeline:
| |
|
Ejecuta tu consulta ANN.
Ejemplo
Copia y pega la siguiente consulta de muestra en tu terminal y luego ejecútala con mongosh. mongosh podría presentar un ligero retraso al pegar la query debido a la cantidad de caracteres en la vectorización.
1 db.embedded_movies.aggregate([ 2 { 3 "$vectorSearch": { 4 "index": "vector_index", 5 "path": "plot_embedding_voyage_3_large", 6 "filter": { 7 "$and": [ 8 { 9 "genres": { "$eq": "Action" } 10 }, 11 { 12 "genres": { "$ne": "Comedy" } 13 } 14 ] 15 }, 16 "queryVector": TIME_TRAVEL_EMBEDDING, 17 "numCandidates": 100, 18 "limit": 10 19 } 20 }, 21 { 22 "$project": { 23 "_id": 0, 24 "plot": 1, 25 "title": 1, 26 "genres": 1, 27 "score": { $meta: "vectorSearchScore" } 28 } 29 } 30 ])
[ { plot: 'A psychiatrist makes multiple trips through time to save a woman that was murdered by her brutal husband.', genres: [ 'Action', 'Crime', 'Drama' ], title: 'Retroactive', score: 0.760047972202301 }, { plot: 'A time-travel experiment in which a robot probe is sent from the year 2073 to the year 1973 goes terribly wrong thrusting one of the project scientists, a man named Nicholas Sinclair into a...', genres: [ 'Action', 'Sci-Fi' ], title: 'A.P.E.X.', score: 0.7576861381530762 }, { plot: 'An officer for a security agency that regulates time travel, must fend for his life against a shady politician who has a tie to his past.', genres: [ 'Action', 'Crime', 'Sci-Fi' ], title: 'Timecop', score: 0.7576561570167542 }, { plot: 'A reporter, learning of time travelers visiting 20th century disasters, tries to change the history they know by averting upcoming disasters.', genres: [ 'Action', 'Sci-Fi', 'Thriller' ], title: 'Thrill Seekers', score: 0.7509932518005371 }, { plot: 'Lyle, a motorcycle champion is traveling the Mexican desert, when he find himself in the action radius of a time machine. So he find himself one century back in the past between rapists, ...', genres: [ 'Action', 'Adventure', 'Sci-Fi' ], title: 'Timerider: The Adventure of Lyle Swann', score: 0.7502642869949341 }, { plot: 'Hoping to alter the events of the past, a 19th century inventor instead travels 800,000 years into the future, where he finds humankind divided into two warring races.', genres: [ 'Sci-Fi', 'Adventure', 'Action' ], title: 'The Time Machine', score: 0.7502503395080566 }, { plot: 'A modern aircraft carrier is thrown back in time to 1941 near Hawaii, just hours before the Japanese attack on Pearl Harbor.', genres: [ 'Action', 'Sci-Fi' ], title: 'The Final Countdown', score: 0.7469133734703064 }, { plot: "Ba'al travels back in time and prevents the Stargate program from being started. SG-1 must somehow restore history.", genres: [ 'Action', 'Adventure', 'Drama' ], title: 'Stargate: Continuum', score: 0.7468316555023193 }, { plot: 'With the help of his uncle, a man travels to the future to try and bring his girlfriend back to life.', genres: [ 'Action', 'Adventure', 'Drama' ], title: 'Love Story 2050', score: 0.7420939207077026 }, { plot: "Captain Picard and his crew pursue the Borg back in time to stop them from preventing Earth's first contact with an alien species. They also make sure that Zefram Cochrane makes his famous maiden flight at warp speed.", genres: [ 'Action', 'Adventure', 'Sci-Fi' ], title: 'Star Trek: First Contact', score: 0.7356286644935608 } ]
Esta query utiliza las siguientes etapas de pipeline:
| |
|
Comparar los Resultados
Los nueve primeros documentos en los resultados de ENN y ANN query de ejemplo son los mismos y tienen la misma puntuación. Esto indica un elevado nivel de similitud en los principales resultados de la query. Sin embargo, el décimo documento en los resultados de la query ENN y ANN es diferente, lo que refleja una ligera variación en la búsqueda exacta y aproximada del vecino más cercano.
ENN la búsqueda examina todos los candidatos posibles y devuelve la coincidencia más cercana a la query según el puntaje de similitud. ANN la búsqueda utiliza aproximaciones para acelerar la búsqueda, lo que podría alterar el puntaje de los documentos. Si incrementas el valor numCandidates en la consulta ANN, los resultados serán una coincidencia más cercana a los resultados de la consulta ENN. Sin embargo, esto consumiría recursos computacionales adicionales y podría reducir la velocidad de la query. El décimo documento en los resultados refleja el equilibrio entre precisión y velocidad.
Tras evaluar cuantitativamente los resultados con la verdad fundamental de ENN, recomendamos probar un conjunto de 100 consultas de la misma manera y calcular la "similitud Jaccard" entre los conjuntos de resultados. La similitud Jaccard se puede calcular dividiendo la intersección entre dos conjuntos (es decir, los elementos superpuestos) entre el tamaño total del conjunto. Esto proporciona una idea del rendimiento de recuperación de las consultas ANN, incluidas las realizadas con vectores cuantificados.
Si notas grandes discrepancias entre tus ENN y ANN resultados de query, te recomendamos ajustar el valor numCandidates para lograr un equilibrio ideal entre precisión y velocidad para tu aplicación.
Recomendamos utilizar listas de juicio para crear una lista estructurada de consultas con sus resultados ideales para la consulta de ANN o los valores de verdad fundamentales de ENN. Puede usar los resultados de la consulta de ENN como lista de juicio de referencia y luego compararlos con esta lista para medir la recuperación, la superposición y el rendimiento. Las listas de juicio permiten evaluar si las consultas de ANN cumplen con los umbrales de precisión o recuperación deseados en comparación con la línea base de ENN. Utilice LLM para generar las consultas de ejemplo.