Episode 1.2: Building an inference for the machine learning framework

for the inference part what you want to do is you need to train on the pose and so and then you need to make predictions on the test dataset so what I'm going to do is stop from here so what do we have I don't need this one yeah I don't need no not sure this one's country Alice and what else do I have okay so I got this no okay this one okay now the problem with this data is you have some new IDs in the asset that are not in the train set so we need to make some changes to our water so just gonna have a D it just there that's data and these that II's and uh just it to set up this way and Jessica so so she just ago and now what we want to change it's only a very simple thing we're just going to extend this to support test data all the IDS are and I said to so I'm going to do now is go to predict ok the EF which is the face SPL meter there's this all this doesn't exist okay probably missing something so I'm going to make one more time say train hit rock bottom - or columns let's check out that and here I'll meet all the same our label on colors so they wait for this to be over so just sending to our models and the name of our chocolate pie was this one [Applause] come on let's go as far so each of them so we need to change drink a little bit add whole roll of them saw the Folger I need to make the pollutions okay so label and Colour is basically going to be the same for all I'm the same same I copies you don't need to say by potties so what I'm going to do it as a pipe falls [Music] that's horrible a this one here sure okay this is also going to be the same for all poles but it's just a check that you can probably end up using this somewhat complicated phonons see you have this one and this one I love that color here I will change the liquid could encode a little bit more so this is kind to be because sure then here we need and colors see I just need to transform it so take your frame okay you know maybe settle so we go to mount model you need operations and that's should work okay so we're running it today to change things a little bit what I do let's like the starter tie downs this is Armando say is zero okay so I have let's try to run it now Jacques that's working by the time I try in slits and as the function a little bit from here so okay yeah yeah let's finish it's finished I love moved stuff around so let's see savers like this and I just did this so I had a precious nun 4-0 that precious threads as was I can just do this I suppose that's you have that and I'm a operations predictions / wife is the silver gold when you share it you can also get just index let's say you've got the desenex all lighting up forgot your why there's a frame as big balls okay just see the samples you should buy what you have here sample solution if ideal target okay so over to us and we have [Music] [Music] [Music] remove it to CSU 5 / or the name name here is okay let's see if this works did you make this work we can just modify this file knowledge back already we don't need these see see take that okay train DF this clock defined as oh yes so just just be or I can just like read this here and I can say okay let's try that okay I think it's like Asami okay so I guess it's pretty unseen levels and that shouldn't happen okay so this problem occurs because my data frame is changing so what's happening now oh yeah justice so yeah just not a good practice but can she walk because the Gators it is small I can really dance but it has I want so - that's training I was sure the covered so this was a problem I was changing later frame so it was having your values all the time but we don't want to do that so I'm just reading it and that's it that's all I'm doing I think it's almost done oh it's okay and stuff so let's see how do I have a solution files no solution part hi I'm Parris I got the sonship out here at the IDM target arms just go back I start a summit for stops you say you'd probably get some love school let's see Oh No so you can see I have some issues on what I call them okay so what I'm going to do so I'm just going to do some kind of Sonic is change the chair my assumption is what should be done so that's right so let's go again go to late submission or excuse me hopefully text and some okay nice so I got a score of 0.75 and my CB score if I can remember each night before it was also at the same range so it's quite good no Saints for the main idea being drained all the cysts like now you have all the faults saved here you can do the poll which you can do stacking awesome on top of the chicken two different kinds of blending so everything is saved and the auto training file encryption file compact you don't need to care about the model in the training path you can dispose of other dispatcher or something else so I think I'll be ending the video today here at this point and all this code that we have done today will be available on my github in this link here for the next video and for the next part so today I kept everything quite basic for the next parts tell me what you want to learn how would it happen I was thinking more of digging into categorical variables in different ways of handling categorical data but we can do whatever you want I have just let me know in the comments or originally out and for LinkedIn and thank you very much see you

Channel: Abhishek Thakur

Views: 10,642

Rating: 4.92278 out of 5

Keywords: machine learning, python, vscode, inference, data science, kaggle, submit on kaggle

Id: zcqgj-Udcqs

Length: 19min 17sec (1157 seconds)

Published: Wed Dec 25 2019