Question: Instead of splitting your data into parts (train/validation) why didn't you create a user - items matrix and then mask some values for the validation?
Copied: Consider using the user-items matrix with masking if: Your primary data source is implicit feedback Temporal order of recommendations is crucial You are interested in matrix factorization techniques Traditional train/validation splitting might be better if: You have lots of explicit ratings Your matrix is extremely sparse Computational speed is a major priority
Hi Abhishek, can you do a testing and deployment series of ML/DL models. I really wanted to get that insight and love the content. I have learnt a lot from your videos and live sessions
Hey Abhishek, great stuff, eagerly waiting your best ____ competition to learn from (begginer/advanced) series. Thank you for your content, it is always something interesting and helpful to me
Hi. Still strange how it run w/o the torch import as well as without self.optimizer arg in fetch_scheduler(). You didn't define self.optimizer anywhere. Unless tez is taking care of it or something.
Im a bit green as far as OOP is concerned. In the MovieDataSet class, in the 3rd function _getitem_, I didn't understand what you are doing. Can you please explain?
hi @Abhishek Thakur, you had a video last time on how to approach classification i think and the different types of classification. Can you link the video in which you talk about that? I cant seem to find it anymore. Thanks for all these videos by the way!!!
Could someone tell me the interface he's using, It looked like VS Code but immediately got some jupyter lab environment. Is it some kind of remote setup in the browser?
Can you please tell how to install tez. do we need to clone it from your repo and then run the install command. As for me git install tez is not working. Please reply at the earliest
Nice one Abhishek! Was wondering how would you handle the cold start problem with this, does this only allow for users seen before in the training set ?
cold-start is always a problem with recommenders. a simple way to handle this would be to show the most popular items. if there are some known user characteristics, you can find the most similar user and then most popular items., or random items :)
hi can you please specify how after training the model, one can find say 20 recommendations for each user. Do we do that using two loops where we predict the rating for each user and for each item and then, pick top max rating?? Please specify
Hi Abhishek, a big fan. Could you please upload a video on 'Extracting text from structured images (Form-Like) using graphCNN (or other algorithms)' please? I read a lot and couldn't find a clear cut explanation anywhere.
you can install pandas using pip. but for now, i would recommend you to use the environment in kaggle notebooks. we will talk about custom libraries in a few days!
Note to self: Could measuring the time for a step indicate if work is done? Calculating a better function should require more work more time. Time is cold or constant heat since there is no build up.
Can you please cover any kaggle problem with classification examples having both train and test csv files as an example. I bought your book and it very good. Thank you 🙏
How about starting a regular coding live stream on twitch or on youtube itself? (much like George hotz and tourist's stream; casual interaction over your fun projects like these) Please give it a thought.
this stratify parameter in train_test_split doesn't work for me in multi labels and it throws error saying found one class but needs atleast two...my labels were 25k...so what i did for my project is reduce the labels to consider atleast more 200 times (arbitrary) present in dataset and it reduces to 300+ labels and then use third party stratification library to distribute labels properly and it works... don't know why scikit learn stratify don't work for me...
Hello! Thanks for the great videos, they are very helpful. May I ask how to avoid overfitting the training set when creating entity embeddings in DNN? I just tried extracting the embeddings for a lightgbm and actually reduces my score on test.
Question: Instead of splitting your data into parts (train/validation) why didn't you create a user - items matrix and then mask some values for the validation?
Copied:
Consider using the user-items matrix with masking if:
Your primary data source is implicit feedback
Temporal order of recommendations is crucial
You are interested in matrix factorization techniques
Traditional train/validation splitting might be better if:
You have lots of explicit ratings
Your matrix is extremely sparse
Computational speed is a major priority
That is a very self-containing talk, thanks Abhishek!
Hi Abhishek, can you do a testing and deployment series of ML/DL models. I really wanted to get that insight and love the content. I have learnt a lot from your videos and live sessions
bro you have this dataset now.please give github link
😢😢 going to copy this and show off in my college
Rmit ?
Good job Abhishek, your videos are so relaxing to watch and learn!
Hey Abhishek, great stuff, eagerly waiting your best ____ competition to learn from (begginer/advanced) series.
Thank you for your content, it is always something interesting and helpful to me
How to integrate this with a website development like django or any other
Thank you very much for the helpful giude! Pls make a video about how to get recommended movies by user ID using this model ) kinda stuck with it
what a time to be alive
Hi. Still strange how it run w/o the torch import as well as without self.optimizer arg in fetch_scheduler(). You didn't define self.optimizer anywhere. Unless tez is taking care of it or something.
Im a bit green as far as OOP is concerned. In the MovieDataSet class, in the 3rd function _getitem_, I didn't understand what you are doing. Can you please explain?
please can you share link of flask deployment video.im unable to find it
nice sir. Question: I have created model based CF and predicted the values. How can i recommend items to new user using this model?
Hi Abhishek, The video was very informative can you please upload part 2 of it soon. Thanks 😇
I am not able to find the dataset on kaggle that abhishek showed in video. Can someone post the link of dataset ?
www.kaggle.com/c/predict-movie-ratings/
It can only be downloaded by kaggle api
Can you provide downloadable link, I'm can't able to create kaggle api
@@abhishekkrthakur Still getting 403 - Forbidden error by using kaggle api
i downloaded it yesterday using api yesterday and it worked fine. did you set up ur api key?
@@abhishekkrthakur yes just download the kaggle.json file and try to download on collab.
after resolving other errors still got 403 forbidden.
Thanks, looking forward for deploying part.
But how do you input a movie name and then get recommendations based on it??
Hello do you have any video on binary or multiclass semantic segmentation
great video!!...looking forward to a video on the transformer recommender
Another amazing video. Many thanks, Abhishek!
glad you liked it
hey Abhishek why don't you use autocomplete
hi @Abhishek Thakur, you had a video last time on how to approach classification i think and the different types of classification. Can you link the video in which you talk about that? I cant seem to find it anymore. Thanks for all these videos by the way!!!
Great stuff,but unable to find code stuff on github
it looks good video , but i was waiting for test it and show us the recommend item for user !
in next video: tomorrow 😉
@@abhishekkrthakur
Is recommend the item code available?
Could someone tell me the interface he's using, It looked like VS Code but immediately got some jupyter lab environment. Is it some kind of remote setup in the browser?
jupyter lab, its now available in vs code
Hello,
What software do you use for making videos? I would like to know how can I make a screencapture with me in the corner. Thanks in advance!
That was very informative. Thanks
Can you please tell how to install tez. do we need to clone it from your repo and then run the install command. As for me git install tez is not working. Please reply at the earliest
Nice one Abhishek! Was wondering how would you handle the cold start problem with this, does this only allow for users seen before in the training set ?
cold-start is always a problem with recommenders. a simple way to handle this would be to show the most popular items. if there are some known user characteristics, you can find the most similar user and then most popular items., or random items :)
Thanks for all your hard work on the videos. Question: Can I live with Keras or should I move to PyTorch?
hi can you please specify how after training the model, one can find say 20 recommendations for each user. Do we do that using two loops where we predict the rating for each user and for each item and then, pick top max rating?? Please specify
Awesome video as always. Thanks man!
Once again a very informative concise video.
Sir,can we use this to design a system for tourism purpose.!!!!!!!
so how to use this movie recommendation ?
Do you use any chrome extension for inverting colors of web pages? Looks good.
dark reader
@@abhishekkrthakur Thanks! Great video btw !!
Can anyone help me with where to train this model or where did sir did it? It's my first model.
Hi Abhishek, a big fan. Could you please upload a video on 'Extracting text from structured images (Form-Like) using graphCNN (or other algorithms)' please? I read a lot and couldn't find a clear cut explanation anywhere.
Thanks man but i did flow the video but i can't import tez. Can you help me?
Hello, Abhishek. Great video as always. Can you please do a video on the recsys challenge from Spotify?
I don't know how to install pandas, it say "Import "pandas" could not be resolved from source Pylance". I don't find a solution
you can install pandas using pip. but for now, i would recommend you to use the environment in kaggle notebooks. we will talk about custom libraries in a few days!
Please how can I use the same code to build for Amazon product
Can you suggest a way of selecting a model with low inference time and high accuracy.
sir do you have the implementation of artificial immune system algorithms?
How to shuffle data randomly after each epoch?
How much u charge for making a video recommendation system for Android app?
What can we do,if we have data in form of strings?
Can you explain difference bw tensorflow vs pytorch, which will be better ?
What algorithm do you use?
Note to self: Could measuring the time for a step indicate if work is done? Calculating a better function should require more work more time. Time is cold or constant heat since there is no build up.
When I download tez and import it .. I get this error :
module 'torch.optim.lr_scheduler' has no attribute 'SAVE_STATE_WARNING
That has been depreciated ...?
@@shriharimutalik3231 ill fix it. please use torch==1.7.1 and this will work
@@abhishekkrthakur Sure , thanks
What if we want to train on multidimensional data??
Is that possible to build a recommendation system based on implicit feedback using CNN?
Very helpful video😊
Btw which IDE you are using🤔
Vscode via code server
Okey thanks😊..... Can we use spyder for this
Or jupyter notebook
Can you please cover any kaggle problem with classification examples having both train and test csv files as an example. I bought your book and it very good. Thank you 🙏
Sir could you make a video on Information extraction from documents.
Can you please share the link for dataset
How can I get Github link of the code explained in the video?
sorry, most of my videos are code along
Thanks man , good video.
Cool stuff, thanks for sharing!
Thank you very much ...
Plz make vd on multi label news classification using deep learning
Hi, could someone explain why we're using nn.Embedding ?
How about starting a regular coding live stream on twitch or on youtube itself? (much like George hotz and tourist's stream; casual interaction over your fun projects like these) Please give it a thought.
Can you help me with recommendation system for blogs!!! Plz
Could you please send me fashion recommendation system idea if u have near you??
Thank you very for this tutorial. Can you tell me which IDE you use?
vscode via code-server
Hello sir
I m working on Graph convolutional neural network for recommendation system can you help me out with improving it by latent factor models
But doesn't adding new users or movies break the system?
yep, CF doesn't solve cold start problem
thanks for your content! Can you tell me how to export file csv which has min RMSE? Thanks!
this stratify parameter in train_test_split doesn't work for me in multi labels and it throws error saying found one class but needs atleast two...my labels were 25k...so what i did for my project is reduce the labels to consider atleast more 200 times (arbitrary) present in dataset and it reduces to 300+ labels and then use third party stratification library to distribute labels properly and it works... don't know why scikit learn stratify don't work for me...
stratify does not work for multi label :)
@@abhishekkrthakur 😔
use iterstrat instead
@@abhishekkrthakur That's what i used and results were amazing...find it from kaggle post after long time
Where is the next video
Hello! Thanks for the great videos, they are very helpful. May I ask how to avoid overfitting the training set when creating entity embeddings in DNN? I just tried extracting the embeddings for a lightgbm and actually reduces my score on test.
got any error while importing Tez
ModuleNotFoundError: No module named 'psutil'
You install psutil
Awesome sir
Can we have a video on malaria detection using machine learning
do you have the data?
@@abhishekkrthakur yes
thanks. man.
love u from algeria
can you share the code please, thanks
Abhishek, thank you very much for what you are doing. Can you please do multilable classification using BERT?
Great vid! Can you add the predict function also?
wowwww this is neat
Code?
If you got the code please share
Hi, Abhi
👌🤟
❤️❤️