Automated Machine Learning using H2O AutoML

[@psw1]

Sir , How does automl handles the class imbalance problem? Does. It use techniques like random overaampling , undersampling , synthetic generation or just by tuning the class_weight parameter in algorithms such as xgboost ,lgb or ranfom forest. ?

[@AIEngineeringLife]

Pradeep.. Most automl out there does undersampling or oversampling. In case of H2O automl it is controlled by "balance_classes " parameter along with few other parameters on how to control it. I am not sure of class_weight can be controlled in H2O as we can directly in xgboost using scale_pos_weight. Will try and update if I find any

[@psw1]

@@AIEngineeringLife Sure sir. Thank you!

[@anantwag19]

H2O Can perform better ? or I can create use Basic Algorithms train on my Existing datasets . , Which one should be performing better ?

[@ijeffking]

Simply fantastic. You are a blessing. I have been struggling with H2O for quite some time. Your explanation has been easy-to-understand and really useful! Thank you very much.

[@AIEngineeringLife]

Thanks Jeff for the note. Code to the video is in video description in case if you want to try it

[@ijeffking]

@@AIEngineeringLife Surely I will be exploring the code myself. Thank you for making the code available.

[@RoyalBengalCub]

Hi Jeff! We are researchers in human-computer interaction (HCI) looking for people who have taken an initiative to recently learn Machine Learning on their own, for career, course or curiosity. People from any major/background may participate. Tell us here about your experiences and any difficulties you faced while self-teaching ML and how you overcame them, and have a chance to win $50 giftcard. You can help this project by taking out 5-10 minutes to participate in our study. For more details, see here: www.surveymonkey.ca/r/SelfLearning_ML Please share this request with your colleagues or friends who fit this description. The survey will be open until July 20, 2020.

[@andreapalladino7999]

One of the best and simplest video on how to use H20. Thank you !

[@DanielWeikert]

h20 states it's different from pandas as it does not store data in memory. So where does it store the data when running h20 in colab? Would the same be true when running it locally (is that even possible)? Thanks. And great channel. Really like it! BR

[@AIEngineeringLife]

H2O data frame is a distributed data frame and stores data in its cluster nodes. On single node, local node is cluster nodes and it reads data during processing and stores it in memory so algorithms run fast. So it will be very similar to pandas on single node but on distributed memory is distributed and processing happens on each node. They have confused all with their statement in website :) . Here is from their documentation "H2O’s core code is written in Java. Inside H2O, a Distributed Key/Value store is used to access and reference data, models, objects, etc., across all nodes and machines. The algorithms are implemented on top of H2O’s distributed Map/Reduce framework and utilize the Java Fork/Join framework for multi-threading. The data is read in parallel and is distributed across the cluster and stored in memory in a columnar format in a compressed way. H2O’s data parser has built-in intelligence to guess the schema of the incoming dataset and supports data ingest from multiple sources in various formats."

[@DanielWeikert]

@@AIEngineeringLife Thanks a lot for your detailed response. In colab we normally have only one node like in desktop or not? H2o.init() creates the cluster. So we would not have to build our own hadoop cluster or spark context? BR

[@AIEngineeringLife]

@@DanielWeikert .. yes that is right . Colab is one node like local machine. H20 creates his own cluster and if u want spark h20 then additionally u need spark as well installed

[@ashirbaddas2573]

Nice once SIr.. Beautifully explained.

[@Yog4All]

This is a very helpful video sir

[@TejasPhase]

Thank You Very Much @AIEngineering Sir for posting this resource. Just want to know, if I further want to explore this library and want to experiment with it then where to look for it?

[@AIEngineeringLife]

Tejas.. You can look at their documentation here - docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html They have good info and details

[@TejasPhase]

@@AIEngineeringLife Sir, Thank You Very Much for this reply. Just one more little question, like this Automation for Structured Data, are there any Automation libraries for "Image Processing" as a Preprocessing step before applying to CNN?

[@Cricketpracticevideoarchive]

Sir, thanks for such an informative video .

[@MSK-qc4lt]

very helpful

[@karndeepsingh]

Hello Sir! while training models this is the error showing up.."Local server has died unexpectedly. RIP." WHAT TO DO NEXT?

[@AIEngineeringLife]

Kamdeep.. did u try in colab or in local ?

[@valerysalov8208]

can you make a video series on mlflow? from tracking to deployment?

[@AIEngineeringLife]

Sure will do it later this month

[@bankaiQPL]

nice

[@kornellewicki4236]

Great explanation, thanks a lot!

[@abilashkanagasabai3508]

Sir please make videos about auto sklearn and auto keras

[@AIEngineeringLife]

Abilash.. I have plan to male auto keras next. Will see if I can accommodate auto sklearn as well sometime this month

[@karimbaig8573]

Sir great tutorial. Like in autoviml we pass in a text column. What do we do if we want to use nlp text classification using h2o automl?

[@AIEngineeringLife]

There is no direct way in h2o for text but you can check this on how to do it- kzbin.info/www/bejne/e3nXkKtup5aYrrc

[@karimbaig8573]

Thank you sir. Could you update the tags of KZbin video to include AutoML, H2O into it.

[@AIEngineeringLife]

@@karimbaig8573 sure will do over weekend. By the way does tag show up in search. Curious to know how is tag used. I never paid much attention in detail to it

[@karimbaig8573]

No, I searched your list first for nlp and h2o, but it did not show up. Then on google still this video dis not appear on first 1 pages. If you change it's name to include h2o or add tags to KZbin video

[@VigneshVicky-cn8ek]

Great Effort

[@akhil18vydyula12]

Sir, I have couple of questions : 1. Why did you remove Customer_Id from X_train? 2. Is H20 directly does hyperparameter tuning for all models? 3. How does feature engineering works here? 4. How can we know which optimization techniques it does? 5. What if for prediction and time series data sets we have? same way of it or any different? Thanks in advance !!

[@AIEngineeringLife]

Akhil.. I removed it as it was unique column but AutoML will remove it as well. Yes and one can customize hyper parameters as well.. Time Series needs separate way of handling. You can check my AutoML playlist where I have covered some part of it

[@bijaynayak6473]

simple and excellent to understand h2o

[@stoufa]

Yet another great video, Thanks for sharing. 😊

[@adinarayanapalvadi2711]

An excellent resource for beginners who wants to learn h2o models. Thank you sir

[@مغيرةمؤيدعزيزاحمد]

Very good

[@Suraj1996786]

Hey! you have an excellent video. I am new to machine learning, my question is does H2O handles the one-hot encoding or there are some ways we can one-hot encode the dataset for the categorical values? I've seen the documentation but couldn't understand. Thank You!

[@AIEngineeringLife]

depends on the model you are using you can pass it is a parameter in model and H2O takes care of it Say in GBM you can pass parameter as below H2OGradientBoostingEstimator(categorical_encoding = "One_Hot_Implicit")

[@Suraj1996786]

@@AIEngineeringLife Thank you!