What's New in ML.NET Model Builder

hey everyone welcome back and in this video we're going to go over what is new and the latest ml.net model builder and so Marty in the main screen here selecting the scenario I had a couple of new things here a couple new scenarios and down here in the natural language processing we had a text classification and in a certain similarity so we'll do both of these first is text classification he had the option to do local CPU and then local GPU and I don't have a GPU online so I can't do that so I just do the CPU I do a file I'm going to do the Yelp review data set here which is basically some sentiment analysis here so predict the column one it's going to be the sentiments and then column zero is going to be the text it's going to start training notice we don't have the option to choose how long we want to train for so it's going to do that for us automatically and this is going to take a few minutes so we'll come back when it's done all right so that's finished and we have a micro accuracy of about 70 so not too bad there and we can keep going we can try the model here this crust is not good the sentiment is going to be probably negative 88 negative and we can consume it console app whatever API all right so let's go back to this scenario here go to since it's similarity same thing local CPU or GPU got a file now I have a just kind of a custom file here that that I did so what we need here is we need a column with the similarity and that is what it says here one to five Fabian the closest similarity one being the one furthest apart and then the both sentences and the rows here I'm going to start training this probably will take a few minutes as well so we'll come back all right so that finished and we see we have a Pearson correlation of about negative 26 now this is most likely due to the data that I use which is mostly kind of just some custom data that I created to help show this so that's probably why that doesn't look good and we can evaluate it let's see alabama.net and I'm hungry it should give maybe a one and we got two so not not too bad of course if you want to find out more about what's going on in the text classification there's this blog post by Louise here which I just noticed is almost a year ago so I'm a little late doing this this video here but that's okay so yeah a bit more technical details of what's going on here if you're interested in that and of course you can do the API of course you can do this with the a model.net APR which will do a video soon on that right so that's it for the scenarios now let me do aggression here and I'm going to do a SQL Server item here it's got a server here this is just a SQL instance on Azure that I have a load of tables here all right so let's switch the table the seller data table here all right get the data preview all right put it into salary next up and you may have seen this in the previous scenarios but now we have this Advanced Training options there's a couple of things we can change here we can change what optimizing metric that we want to do so because it's a regression we can choose r squared which is the default or root mean squared error or mean squared error and also we can tell out what algorithms that we wanted to use for our training so if we know we don't want to use light gradient boosted machine we could just unclick it there and it won't use that with an auto ML and as one of the trainers to try all right so that's a bit of a preview of some of the the big things that's been new in model builder I hope you enjoyed the video and I'll see y'all next time

Channel: Jon Wood

Views: 1,490

Keywords: ml.net model builder, ml.net model builder text classification, ml.net model builder sentence similarity, what's new in ml.net model builder, what's new in model builder, model builder

Id: ShQe4AmtgwU

Length: 4min 22sec (262 seconds)

Published: Mon May 22 2023