Theyidentify a statisticiansoreconomistsorbiologistsandtheinterfaceswebuildtomodelingandstatisticsfromourtendtobereallyfluent, highlevelinterfacesthatkindofreallywellmarriedtothevocabularyofthedomainthatpeoplearedoingworking.
Andsotensoflowbringsthisawesomecapabilityforneuralnetworksandevenjustas a generalpurposenumericalcomputingframework.
And I thought, Wow, wecoulddoincrediblethingsfromourtomakeallthispoweravailabletoallthesepeoplesolving, solvingheartproblemsthen, particularlythecaressinterfacewhich I hopewegettotalkabout a littlebitmoreis a thereallynicehighlevelvocabularyfordoingdeeplearningthatfitsreallywellwiththewayourusersworkandthinkon.
So I wasexcitedtobothcanmakethecorecapabilitiesoftensorflowavailable, butthenalsotomakethecaress a P I available.
Wecreated a wholebunchofexamples, probably 25 orsoexamplesofusingcaressandmany, manyotherexamples.
Wecreated a galleryofkindoflongerformblawgpoststhatdescribedlikereallyindepthworkedexamplesindifferentdomains.
Sowehaveinvested a hugeamountandeducationresourceis, andwe'regonnacontinuetodothatandthisbookisexcellent.
Thebookisawesome.
Yes, I readthepythonversioncovertocover, and I taught a classwithit.
Studentslovedit.
I stronglyrecommendittoallthedevelopers a meet.
What's remarkableaboutitisthatitcoverstheconcepts, theconceptualterrainofdeeplearningin a waythat I thinkwasreallyintuitiveforpeopletounderstandwasnotnot a hugenumberofprerequisitesbutalsohas a wealthofpracticalinformationabouthowtoactuallydodeeplearning.
Soit's justit's a rarebookthatcombineskindofconceptualmaterialandpracticalmaterial.
So I I recommendthatisthefirstthingthatpeoplewhoareintheourcommunityandtheysay, I wanttolearnmoreaboutthis.
I recommendthattheygetthatbookandreaditfirst.
It's probably a lotofworkportingthatoverourYeah, itwas, butmostofthebookspredominatelyconceptual.
Andthentherearetheseexamples, andso I reallyjustchangedtheexamples, butreallymostofthebookisendsupbeingconceptual, thatkindofat a higherlevelthanspecificlanguages.
There's ourpackagecalledtensorflow, whichis a lowlevelinterfacetothefulltensorflowgraphinthefullpotentialFBI.
There's a caress a poetrybeentalkingabout, andwehavetensorflowestimators, whichisanotherhighlevelframeworkthatGooglehasfordoingmodelslikeclassifications, regressionmodels, classificationandregressionusing D N ends.
Andsothere's a really, reallyhighlevelfunctionsthatcomewith a niceframeworkforfordata, preprocessingandthingsand a niceframeworkfordeployingthemodels.
TheestimatorsFBIdeploymentisreallyimportanttohave a swellitis.
Andthat's oneofthethingsthat I'm mostexcitedaboutbecausetraditionallywithourwhenyoudomodelingandthenyouwanttodeploythemodelinsomefashion, youhavetobringthe R runtimealongwithyou, andthatcouldbe a challenge.
SowhattheThedesignofTensorflowisthatyou'rewriting a program, butthatprogramiscreating a graph, andthatgraphisexecutedby a runtime a C plusplusruntime.
There's a librarycardcaress J askedyoutocaress, modelandruninyourbrowser.
Sowe'vegotlotsoftoolsforthat.
Andthat's actuallyoneofthemostexcitingthingsabouttensorflowisthatisthisdeploymentmodel, sotensorflowdotJsokay, supports a caresscompatibleAPIook.
And I believewhatthismeansisthatpeoplewillbeabletoauthormodelsinourThat's right, usingthecaressinterfacethat's rightandthendeployment.
Deploythemrightinto a browser.
Yeah, that's gonnabehuge.
It's phenomenal.
Yeah, So I think a lotofpeopleintheargumentthat I'vetaught I meanallcommunities I thinkthisissuch a fundamentallygoodideatotakemodelsandserializethemintothisruntimeformat.
A websitecalledTensorflowthatourstudiodotcomithasdocumentationforallthepackages I talkedaboutParisandestimatorsandandthecourttensorflowFBI, asinformationaboutdeploymenthas a galleryofindepthexampleskindofblawgposts.