I finally got a copy of "Approaching (Almost) Any Machine Learning Problem"

Key Takeaways

hello everyone how are you guys doing am i audible to everyone so if i'm audible to you please write a message on the live chat awesome so welcome to this video and uh i'm very happy today so i finally got a copy of my own book and it was sent to me from a friend so the country i live in right now doesn't have amazon so someone else ordered it for me and sent it to me so one of the copies is mine and it's quite exclusive copy so i will show you in a while and i'm also sending like uh i'm giving a talk in south korea in a couple of days um so it's it's a online talk and i'm also sending around 15 copies there so about india so um everything has been delayed and it's because of uh covet 19. so i'm waiting on a final print copy from the publisher uh in india just to check if everything is all right and that's been taking a lot of time the price you see right now on amazon india is around 1700 rupees but it's not actually 1700 it's an imported copy so be careful if you can wait for a few days i would suggest you to wait for a few days because i'm planning to launch the same book the print copy in india for 800 rupees so waiting would be more advisable so the book yeah this video is all about the book and i'm going to show you the book of the rahm you will get a copy from me for free um so this one this is a special copy and as you can see you can see the not for resale stuff there and this is the normal book that you will buy the content is the same the only difference is this one is colored and this one is black and white so when you are when you're selling it for uh such low prices the problem is i cannot i cannot afford to make uh the books colored so if i have to sell it for 800 rupees and the price will reduce further so if a lot of people will buy this book um then i'm going to slash the prices a lot more so if people like it uh it will be cheaper if people don't like it it will stay the same price so um this is the book and uh as if you if you go to the like i don't have a lot of crap in in the book so in the starting of most books so you have like 30 pages of crap that nobody gives a about so i don't have that i have i have a simple thank you for the people who have uh helped me and you can probably see their names now so these are the people who took their time and helped me with this book helped me to review the book and these are some of the most awesome people i know and um when you go to page three of the book you will see that first of all you if if you plan to buy the book do read the description of the book because don't don't read don't do it without a description read the description of the book and that would be the best and you see like uh in third page of the book i have written one sentence in the end and that says if you didn't code you didn't learn so if you if you are not a coding person don't buy the book it's not good for you the book has a lot of code and it has a lot of code and each and every line of code i is quite welcome and commented if there is something that's not commented uh let me know and i will explain you what what it means so you can contact me on any channel linkedin or twitter twitter my messages are open so you can contact me there anytime you want uh if if it concerns the book so i i try to look at each and every message and respond to them so the first chapter of the book is setting up your working environment and in this chapter i explain how you can use anaconda or mini conda i'm using miniconda and create an environment and this is this is like a two-page chapter i'm i have also shared the environment file with you guys so you can use that environment file um let me show you my in my copy of the book so after that after you have set up the environment i jump a little bit into uh supervised and unsupervised learning so in this chapter i i don't cover a lot of things you know because you don't need to cover a lot of things so let me answer some of the questions which are coming in so inr price is 800 rupees i have already mentioned that um can we get a colored version yes you can get a colored version but it's going to cost you four times the price so i will launch the color version tomorrow and any plans for computer vision book computer vision is the third book in this series of approaching almost that i have a pdf version i i don't i don't really want to do that and some people have already done create a pdf of the kindly books and doing a lot of illegal pirated sharing i i don't really like that so in the first chapter in the real first chapter the supervised and unsupervised learning i jump a lot so i go i explain a little bit about what classification regression is and then i jump into some clustering approaches i talk about tsne so everyone's favorite uh mnist data set yes i'm using mnist because this is starting and i cannot i cannot go like to a very complicated data set in the very beginning of the book and uh an amnesty when it came it was complicated believe me um then then i have a chapter on uh cross validation as you can see so in cross validation i'm again using a data set um some of you maybe you guys are quite familiar with the data set is called wine dataset redwine dataset predicting the quality of red wine and then i have let me fix this thing a bit okay yeah and then i have um some so here i describe a lot of things related to what kind of cross validation should you use so if you see that uh if you look at some of some of the what overfitting is so i try to describe that and then i try to give kind of a definition which is uh not available uh in many books and uh then i i jump into how to create the cross validation folds so this is the first step that you should do whenever you're starting with a machine learning problem the first thing that you do is to create the cross validation folds and that's what i'm doing in this book so you see uh like how to how to split the data and okay i'm not very good with keeping the book in focus how to uh split the data what the training and validation data that should be like and even uh like the code uh for it so the best way to read the book would be to maybe read a page or read couple of pages and look at the code and then implement the code on your own that would be the best way but you you can obviously find a lot of other ways and uh then i go from um k4 cross validation to stratified k4 what is stratified k4 what is the hold out based validation how to do uh cross validation for time series data then i have a few more things like how to how to convert a regression problem uh to a stratified k4 so this is probably something new for some of you then i have i i have a general way of creating cross validation so you can you can create a uh class cross-validation class and you can uh have different kinds of cross-validation there so aditya sony it's not three four eight seven yeah it's 800. don't ask me that question again please um indian version is not uh has not has not been uh announced yet it will be soon and i don't i don't set the random state in the book because i mean even if you sent the random state it's not going to be the same across different workstation or machines so it's like that um then i have then i go into another chapter which is very important and um i have tried to cover a lot of important evaluation metrics but obviously i cannot cover all of uh evaluation metrics um so uh yeah so these are the evaluation metrics that i have covered so i have accuracy precision recall f1 score auc log loss precision at k ap at k map at k mean squared error mean absolute error and all kinds of other regression metrics so this is what i have this is what i have covered and uh as you dig into this chapter you will see that i have not only tried to define these different kinds of metrics and even like things like true positive false negative and like one liners uh using an example of a pneumothorax classification but i have also implemented them so you have the implementation and you can you can also see like i have covered i have tried to compare the implementation with uh second learn implementation uh yeah you can as a beginner you can use the book i i don't see why you will you will you will get a lot of terms so in the first 10 pages you will have a lot of terms that you might not know but you can obviously google it try to google the terms and learn but it's more like a code first book and you have to code along to learn and to get the most out of this book so some other questions um update on paperback version in india i've already given that update so um why did i choose not to use jupiter yeah good question so i chose not to use jupiter because jupiter is good for eda purposes and if you see like i don't have a lot of eda in the book but when it comes to uh real kind of machine learning engineering where when you have to like let's say put something to production you cannot you cannot use uh jupyter notebooks for that and a lot of people try to do that but it's not recommended at all so uh as a machine learning engineer you can definitely use a book um i don't know how useful it's going to be for you um given like maybe you already know a lot of stuff um but you can definitely try to learn a few things that i have explained here so after after uh true positive false negative for precision recall i go into uh auc and try to define like uh what kind of aoc is good or bad so that's something you can probably learn here and then i go into how to choose the best possible threshold which has a small mistake but yeah let you guys will figure it out i'll leave it to you guys to figure it out uh so the way i the way i say how to choose threshold doesn't have any mistake on the only mistake that i have here is the threshold that i have written that you should choose but that has been updated in a new version so the version in india will be correct um then i then then i go into uh like other kinds of approaches like micro averaged macro averaged and weighted scores and i try to generalize it as much as possible like so that you can implement them on your own so it doesn't matter what kind of metric it is if you're asked for micro micro average you can implement it on your own okay let me answer some of the questions it's running quite fast now is computer vision cover we will come to that and um as a data scientist you can definitely use a book you will probably learn a few new things not maybe not a lot of new things um okay i'm not going to hiring and mathematics at the moment um thank you tomb2141 and okay thanks pawan i've seen your message your message a few times this is not a church [Music] yeah you must be familiar with some some data science and machine learning concepts and you must be familiar with coding and yeah so i i'm into scripts i'm not into jupiter so you don't find anything which works in jupiter here obviously everything works in jupiter you should just copy paste it to jupiter so everything will work and then uh this is this is a quite huge chapter i would say i've invested quite a lot of time in evaluation metrics and then um i i discuss a few like uh important metrics in the end of the chapter like matu's correlation coefficient i discuss correct weighted copper so these kind of metrics are discussed like i i just touched them so it's not very detailed mcc has an implementation obviously and uh then i dive into arranging machine learning projects now this is this is like a very important chapter for me and uh like the people who have been asking why not jupiter uh well read this chapter you will understand and it's because here i show how you can uh try to arrange your machine learning project like uh in a proper way that you can uh even let's say i'll give you an example let's say when you start with a kaggle competition and you start with building one model you build one model then you change a lot of core you build the second model and then go into the third model and like that right so instead of doing that uh build something where you can just specify what kind of model you want to use and um then you have you don't have to code after like two weeks so you just have to change a few things so you don't have to change a lot of code so for that purpose and also for um your job i think this is a very important chapter then i then the next chapter that i have so this is this is important and it's quite small it's not a lot of pages for this chapter and um then we move to the next chapter uh let me take some questions first so i expected a lot of depth uh in this book yeah the book is great for beginners yeah definitely there's not a lot of depth in the book um it's impossible to cover everything in 300 pages believe me and uh i will talk about the course in the end of the video so yeah i i use vs code so you can use vs code i use i i like to use the code server version of vs code um kindle version is there others it has been there for a long time um i plans on winning any candle challenges i i don't think they let me win anymore it's a lot of competition uh [Music] hard copy will be released in india soon how to set up the environment that's the first chapter deep and reinforcement reinforcement learning is not covered i'm sorry uh upcoming books in the end of the video i will give a brief about the course i'm going to start in the end of the video no you don't nee i mean i i'm not against jupiter notebooks so a lot of people think like that so you don't have to say no to notebooks notebooks are nice and it's good for early evaluation purposes if you're doing exploratory analysis it's it's really very good to use notebooks even even for writing simple functions and even for creating the first model it's really very useful but the problem with notebooks is you run cell 4 and then you can go back to cell 2 and run cell 2 and come back to cell 4 and the values will change so things like that you need to avoid in production and i i would rather distribute the software or what i have written as scripts which are easy to run so and in a good way like use by using docker so the next chapter is approaching categorical variables this is very extensive chapter and i can i mean most of the people like when you talk about feature engineering most of the people talk about categorical variables and uh there is a very so i have tried to make this chapter um really extensive and try to cover a lot of different things like um what are the different types of categorical variables and how to handle them what are uh what are the different kinds of feature engineering can you do on categorical variables like you go for binary one hot encoding so in which situation uh what is good what is better and whether you should go for binarization whether you should go for one hot encoding and there is also there is obviously a lot of code here that you can you can take a look at and and then i have some kind of some feature engineering that you can use for categorical variables so i define that how to combine different kinds of categorical variables how do you split your data for categorical variables or how do you handle new categories or rare categories so these kind of things so that that's the chapter on categorical variables but and it's a huge chapter i think i dedicated around 50 60 pages to this chapter so uh there's a lot of things a lot of cool stuff here hopefully for you and i also define if you can uh how how you can use uh categorical feature uh for um data for models like logistic regression or for models with like xg boost or any kind of tree based models i also have like a section on entity embeddings so and this is like the one one of the only uh places where i actually use tensorflow so uh i use tensorflow later on for it for something else but here i use it to build the model and then i again have uh one more chapter on feature engineering so this covers what has not been covered in categorical variables and how to combine categorical variables with normal numerical features so let's take some of some of the questions again um so computer vision so maybe we will come to that uh software engineering related concepts not in this video uh tutorials with tensorflow i use i'm far away from tensorflow now so uh i do like tensorflow but i find pie torch much easier and better but i if if people want i can do some tensorflow videos so i've been doing tpu videos anyways [Music] beginners in machine learning can use this book uh use it at your own risk so i mean there there are some things that you need to know that you already need to know uh i don't know how this book compares to other books i'm sorry i have not read other books many other books i've read some [Music] no it's not specific to ml being used in competition i don't think you can get a kind of a good score in competition just by using this book but maybe it will help you think a little bit more um this book is very useful for if you are working as a data scientist or want to work as a data scientist um how many days do we have to wait for print book in india i mean if people had been wearing masks and they were home my book would have been released in india last month india was my first country but now it's kind of being the last country so after feature engineering chapter in which i i show you all these things um like combining categorical numerical features how to create features from date and time data these kind of things i move to feature selection which is also like a very important topic and uh in feature selection i try to create a kind of like a general uh feature selection class that you can probably use with any kind of data set i talk about uh greedy feature selection so implementation of greedy feature selection how that's done and then i talk about some some of the feature selection methods like rfe and i talk about feature selection from models from different kinds of models and then um that's it on feature selection then i move to hyper parameter optimization now if you're if you like watching um this channel of mine i made a hyper parameter optimization video yesterday and everything that i described has been covered in this book i think the book has more than what i have described in the video or maybe the video has more so optina is not covered in the book but it's in the video um [Music] i have not been in this field for past 10 years man so i don't know and then i have like in hyper parameter optimization chapter i have everything like grid search random search i have uh using a custom metric how you can design your own custom metric and then do grid search or randomize search then i have then we have things like uh hyper opt uh using bioshin optimization using gaussian process that we have um and uh we don't have optional but then i also provide you with like a table of uh different parameters that you can use uh when you're tuning um different kinds of machine learning models so if you if you look at the book i have like i don't know maybe i've mentioned light jbm um a few times but uh i have not used them because all the tree based models are the same light gbm has a lot of parameters and extra boost was much easier to train then i have a chapter on approaching image classification and segmentation so that's the computer vision chapter i have i just want to touch it i could have rendered the book right here and said that that's it for machine learning but that's not everything right so in in classification and uh regression uh for images uh and segmentation sorry i have uh i'm using a data set for pneumothorax and like you can see here i have uh neutrax and non pneumothorax images and trying to show you how you can uh try to build a model for uh this these kind of data sets or any kind of image classification dataset and i go a little bit into uh different kinds of models like uh i i do talk about lx net and then i go into uh then i talk about what are the different different things like what is stride what is filters what is initialization why do you do initialization what is vanishing gradients but this is all has been covered in like uh very little space i talk about dilation i talk about max pooling uh then i show you how you can implement your own neural network like a network like alex net in pytorch uh then then we move a little bit more forward and i show how you can define the classification data set data loaders for pytorch and then i have things like uh then i have things like uh here i where i show like implementation of uh after implementation of lx net i go into resnet so how you can how you can easily switch between models what do you need to change what you don't need to change so these kind of things and uh then using the same problem like for pneumothorax versus non-pneumothorax images i go a little bit deep and try to implement a unit and then i show you how you can not just detect pneumothorax or not but you can also detect how it's um how the segmentation is done so how you're segmenting the image so let me i also talk a little bit about what are the different kinds of augmentations that you can try and do when it comes to images and then in the end we try to build this model for segmentation for image segmentation so which you can apply anywhere any kind of image segmentation problem after this chapter we have uh approaching text classification and regression i tried to when i started writing this chapter i i think it didn't take me much time to write this chapter and i decided to write a book on nlp link if you message once i can see your message man come on so uh nlp book will be released very soon so i've already started working on that uh then in this chapter you will learn some of some of the basics of uh natural language processing so like what bag of words is what is count vectorizer what is tf idf how to use them what is nltk how to tokenize words what is stemming what is lamentization uh you will also read a little bit about uh what are the different transformer models i don't go into details of the transformer models uh but before that you will learn how to use lstms or gru's one-dimensional cnns for building a model for nlp and you will also learn some some something about world embeddings so to use word embedding so these kind of things i also have uh an example on bird so because everybody likes bird transformer model so you can just like switch the model from birth to something else if you want and it will work and then i talk about fast text embeddings i talk about globe embeddings i obviously don't go into too much details that i will tell you now so if you're thinking like i discuss all the details in 300 pages that's not possible man and uh then i i i show you some of the ways in which you can try to load the embeddings from different kinds of uh um embedding uh vectors like glow for fast x so let me answer a few questions so nlp release nlp book will be released um in some time i cannot promise anything right now and um this is going to be very extensive uh nlp book um so vs studio for machine learning or yeah obviously you can use any ide you want i mean i started with sublime it was a text editor so you can also use sublime uh any online course for machine learning and planning i mean if you're doing the courses from android i think they're the best then you should start with so my book is based on pure pytorch it does not use my own framework so i've i've used my own framework once only once in this month that's it and um i think that was in the cnn chapter where i've used my own framework i've used it to show how you can use it for um plant classification one of the problems and um what do what do we have do we have any more versions is deployment covered let's see if the deployment is covered in like next two three two minutes two to three minutes uh ibm data science i'm sorry man i don't know i have not done that [Music] time series data is not really covered so maybe i have like a paragraph on it on how you can do um cross validation for time series but that's it uh feature importance are covered yes but i mean what do you mean it's a huge topic right so it is covered uh i've covered a little bit of it devops knowledge is required in data science i mean depends on your data science job then my gpu and laptop setup i will talk about it very soon maybe this week if we get lucky um deploying machine learning models at scale i will i have some things planned there if you have time can you please explain code on page two for one more specifically after the comment that says choose first sample and create a dictionary okay let me see if i if i cannot then ask me somewhere else so choose first sample and create a dictionary of feature names and their scores from svd you can change the sample index variable to get dictionary of any other sample okay so choose the first sample and then create a dictionary your feature names so sample index so you have to you have to put sample index to zero and no it's not in play store and i have two okay so you have to make the sample index 0 and then that's that's the 0th component 0th index of the components that you have and then you can create a dictionary with feature names so that's how you do uh topic modeling stuff so if you do that then you have like only the first one then you can run multiple times and so you have like all the different topics for different sentences so zero it's all only for the first one first sentence um but i can explain to you in more details uh later okay uh how much time does someone need to go through this thoroughly spend this book into an end i don't know man if you are doing like a chapter or a day then one chapter a day is difficult i think maybe three weeks or maybe a month so okay moving on um i will move to the next chapter after nlp chapter so after nlp chapter i have uh this chapter on um on something and stacking so i decided to spend some time there and just to explain what ensembling is what stacking is what blending is and i try to explain it with a nice diagram of a elephant and i also have a function that you can use to optimize any kind of uh metric so here i'm doing auc but you can optimize any kind of metric you want and then i also discuss stacking in the same chapter so what stacking is and how difficult it is it's not very difficult and that's that's all about that chapter uh then i have approaching reproducible code and model serving so here i talk about docker i talk about how you can dockerize your machine learning application that you have built till now different kinds and i talk about what is nvidia's uh docker uh container um so what to use what not to use uh talk about what to use not what about what not to use and then i talk about how you can use docker to deploy machine learning models but i don't talk about how to deploy machine learning models because when i was i was right trying to write that i uh i ended up like taking screenshots and writing okay go to figure one and click on this go to figure two and click on that so um i decided not to write about how to deploy machine learning models so i was i was thinking about kubernetes cube flow uh bean stock uh i was wrong i was thinking about uh what was the other one i forgot the name okay anyways so i was thinking about these things but i didn't i didn't write them in the book and i will uh sage maker i was thinking about stage maker um but i will be making youtube videos uh for that and i will show you how to deploy these models so in this book i have tried to cover like almost everything that you need to know when you are a data scientist or when you want to be a data scientist but there's also a lot of things that i've not covered and what makes this book different from other books is this is a code first book so that's the thing that you must keep in your mind if you want to buy this book then remember that it's a code first book and it expects you to code so if you like if you like copy pasting then don't buy this book please and if you have already bought the book and if you like it then do not forget to write a review on amazon um so that's it for the book i will scroll and see uh what are the different questions i have and try to answer some of the questions so uh there was a question from um there was a question from rashmi right about the code in the book so please message me on twitter and we can take a look into that in more details but if you just print after that line you will understand what's happening how much how much devops knowledge on this one i have already discussed we have done the time series so if you have some questions then feel free to ask me if you don't then we can end the stream pipelines in the book uh so pipelines in the book that i have discussed my work or competition pipelines are very much similar maybe not my work pipelines but the combination pipelines are quite similar and but even in my current work and my previous jobs we have used docker and all that stuff all the time so yeah it's very similar to work too exploratory data analysis is not covered my friend i'm sorry about that and i have tried to like explain um a few things when dealing with different kinds of data so maybe that's exploratory data analysis but i have not gotten gone into like very much detail arranging ml project means uh how you should arrange your project so that it's uh when it's distributed to other people they don't have to face a lot of problems and you don't have to face a lot of problems when you're experimenting from one model to another model uh [Music] how much time does okay i've already answered this question so any books on big data spark i don't know not not from me sorry um really sorry um industrial time series project if you have any suggestions on industrial time series brothers you can tell me if you have some kind of data set and i can try to create a tutorial around it uh tensorflow or by dodge what is better i mean both of them are good both of them are really nice and you just have to choose one so just choose one and stick to it and know about the other um okay i also wanted to talk about this course thing that i was thinking about so i'm planning to make a applied machine learning course in which uh which will be three months of course uh there will be so details are not decided yet i have not uh thought a lot about the details but some of the details that i know um that i think right now table of contents of this book okay so let's see the table of contents of this book i've already talked about all the chapters so let's see the table of contents first then i will talk about the course thing so this is the table of contents so very simple very easy and it's also in the preview of the book so you can i think 30 pages of the booker for preview so that's 10 of the book so you can see that for free so the course um i'm thinking about a basic python course python for data science course like a crash course maybe a week maybe 10 days and i'm also thinking about a course on machine learning and applied machine learning mostly so in which i will be talking about some of the basic stuff and implementation how to um so starting with linear regression logistic regression to going into details of uh uh gradient descent then going into neural networks uh these kind of things so it's divided into three months the first month we will be studying um simple machine learning stuff and also deep learning stuff and how to implement them in buy docs and all that kind of things and um in the second month um we will be studying everything about natural language processing so going from back of words to transformer-based models to model summarization and whatnot and in the third month we will be doing computer vision so starting from the basic stuff again to going into gans so these kind of things and it will be very much applied there will be projects i still need to think a lot about different kinds of projects there will be special sessions so the classes will be three times a week one and a half hours that's a lot of class and there will be one project for um two weeks so i um i will share i will share a link with you guys so i have not decided about the price of the course or anything but i will share a link so i'm taking this information from you what should the price of the course be and you tell me so here is a link and you can go and take a look at this link and fill up the form if you want if you're interested then i will contact you later on so there are not individual courses so if i have to make them individual it's going to be i cannot make the course free venkat i'm sorry about that i i cannot do that but youtube will always be free so i will always keep on making free tutorials here um on which platform will you release my course i will release it on my own platform uh we will do live classes so all the classes are going to be live so take a look at the description i can i can read a little bit from the description of what i've created and i can tell you but just take a look at the description so first month simple stuff maths basic concepts implementation some problems tabular data time series data these kind of things second month nlp third month computer vision no and three live classes per week 1.5 hours each there will be special sessions or channels to cleared out and there will be special sessions on few weekends so where i will also be inviting some industry experts to discuss some projects um there will be the projects will be industrial level the project will be a combination of uh industry and competitive uh machine learning kind of thing and we will also like i will also try to help you with your portfolio building there will be bi-weekly group projects and there will be one major project that will you will do for entire three months and if you're interested in like individual nlp or these kind of computer vision codes then yeah feel free to write in that form and i can take a look but that's not my current plan but i can definitely take a look so the book again the book i don't know when it's coming in india man sorry um i hope soon i'm hoping this week but last week i was hoping last week so uh prerequisites are you should have some knowledge of python if you don't have a knowledge of python i've been planning to give a crash course of python for but only limited to machine learning and data science stuff uh so like your python for data scientist um okay one more thing that i want to talk about the book the ebook version of the book there's a kindle ebook it doesn't open in cloud reader and it doesn't open in old kindle devices it's there is a simple reason for that one of the and the reason is like i love the formatting of the code and i don't want to miss the formatting and the old kindle devices or the cloud reader it's very difficult to fix the formatting for them so if you have old kindle device and you have bought my book write me a message on twitter and i will send you a copy and send me your suddenly the invoice that you bought the ebook i'm only doing it for the ebook not for the print copy so if you've bought the ebook and want to want it to work on your old kindle device send me a message on twitter with your invoice and i will send you the details on how to how you can get the copy for free if you've if you've already bought it you should get it for free right um [Music] extensive natural language generation um i don't know if i'm i i would want to write a book on energy just analogy but there is nlp book coming and that's going to be quite extensive um yeah rishi feel free to use the book so if you use the book and if like if uh if you're done with the book feel free to give it to others so um people who cannot afford or just give it to a library or if you don't like the book then give it to someone who would like the book i hope worldwide it will open soon so then you will be able to buy it pretty soon let me launch the book in india and then we will think about worldwide because the the company that i'm talking to in india they deliver worldwide setup environment for machine learning in windows um difficult question i will try but you can try to use in wsl computer vision book will be after nlp book and um i mean if you want to go for a data kind of job i would recommend going for data engineering or data science jobs [Music] right now uh okay so this is it this is for this is it for today's video and thank you very much for joining uh there were a lot of people and there were a lot of good questions um and uh thank you very much and i hope you like the book and uh if you buy it like let me know tag me if you want to and if you like it then do write a review on amazon because self-publishing is difficult and okay i will i will talk about the prices again so in india it's available for 400 rupees the ebook 800 rupees for print uh which will come soon and outside india it's available for 8.5 uh ebook 15 for the print version so and the print version in all countries is black and white if you want the colored version i will i'm launching it tomorrow and it's going to cost you a lot more so i think it's going to cost you 50 instead of 15 so thank you very much thank you for joining and i hope you will like the book and tell me about the contents that you want to see tutorials uh in comments or anywhere and i will try to make them happen thank you very much goodbye