Partial Dependence Plots (Opening the Black Box)

[@jacob_dmn]

Man, lemme say this.. I’m learning Data Science from your channel not because i don’t know other resources, but because you were one of the most intuitive and realistic teachers i knew after a long journey searching the best resources, i love this channel pleaase keep lighting us 🙏💚

[@ritvikmath]

Glad to hear it!

[@hammasmajeed3715]

Thanks so much. I was reading the PDP from tutorial, did not understand anything, thought that this topic is really hard to understand. This 10 minute video cleared everything and made it very easy. 🤩 going to look for other videos in the channel.

[@ginervs]

Fantastic, watched this once and understood the concept after trying to read obscure articles/study manual that only gave formulas and not real life example. Thank you!

[@micmaw7553]

Thank you very much for this video! Only someone who knows his subject and wants to share his knowledge can do that. It's very rare!!! Thank you for your generosity!!!!

[@szymonk.7237]

Very clearly and intuitivly explained. Thank you ! ❤️

[@ritvikmath]

Glad it was helpful!

[@xxshogunflames]

Ritvikmath the disclaimer at the end was the cherry on top, I definitely am excited to have this in my tool kit (especially with final projects around the corner) Thank you!

[@vreesje]

Very clear and concise explanation. After this video, the concept is very clear. Thanks!

[@aephidayatuloh]

Great explanation! And as a non-native English speaker, I can clearly understand in a simple way.

[@mx555551]

hey thanks so much man. your explanation was amazing, you're a life saver!

[@MarinaRosellMurillo]

Great explanation, super clear and easy to understand! Thank you!

[@MrMoore0312]

Very cool technique! Post-prediction analysis is an area I'm most def interested in seeing more of

[@fabiolainkorea]

you are a great teacher...

[@aditipawde5894]

I was scared to look at the notations and integrals used in the PDP literature. After watching your video, my reaction was 'Eureka!!'. Very well explained!

[@khonghengo]

Really nice explaination. I really appriciate your work.

[@ritvikmath]

I appreciate your comment!

[@seemaaswani8369]

amazing explanation

[@keithmaly904]

nicely articulated. ty!

[@weydesign]

I really like your videos, everything is very easy to understand and it helped me a lot!

[@lautarocisterna3339]

Amazing pace of teaching and explanation, thanks!! Are you planning to make a video related to shap values? Thanks again

[@maty6091]

Amazing video. Thank you!

[@janezou8611]

This is very helpful! Thank you! Can you please do video for ICE plot?

[@ΔιονύσηςΓιαννούλησ-δ6γ]

You were so helpful ,thank you so much

[@daesoolee1083]

Super neat lecture

[@ritvikmath]

Glad you think so!

[@Aquila67280]

Wow, that was a great explanation! Thanks!

[@lakshman587]

Thank you so much for the video!!

[@MartyAckerman310]

Needs to do a video on the probability of Ritvik not catching his marker after tossing it into the air.

[@xhinker]

clear and nice, Thank you! the video is underated.

[@yashshah4172]

Hi, I enjoyed the video, can you please explain for continuous feature, how it works ? do we take all the different data points available on the dataset or we take all the numbers from the range in a specific interval ? Thanks.

[@teddy5474]

Crystal clear!

[@louisg3713]

Very clear video I just have a question regarding what you're saying at the end. You're talking about an alternative algorithm that takes the probability distribution of the p_hat for the plot. What is the name of this algorithm ?

[@Cath38639]

Great explanation!

[@experiacare1383]

Well explained! Thanks!

[@ajithshenoy5566]

kindly create a video on SHAP

[@ritvikmath]

good suggestion!

[@PuneetBhushan93]

Which Blackboxes/algorithms use PDPs to derive predictions?

[@jamalnuman]

what is the difference between the PFI, SHAP, LIME and inherent FI?

[@jamalnuman]

great thanks

[@qiguosun129]

You can find a detailed explanation for each new word here, thank you!

[@sownikturaga800]

Can correlation of a feature with other features in the dataset cause flat line in the partial dependency plot?

[@dobaolinh5165]

Will 2-feature partial dependence plot solve the independence problem? For example, you plot average predicted values for every combinations of games won and games lost in your dataset?

[@kayae6659]

Great vid, thanks!

[@akashbhateja7410]

Why further videos are hidden??? Please make them available!

[@Set_Get]

Thank you. I enjoyed it. I have a question: do we use each plot separately for prediction or do we combine the output of them to get the final prediction?

[@ritvikmath]

The interpretation of each plot is the isolated effect of each feature on the final score so typically they are used independently

[@PuneetBhushan93]

How did we calculate the individual predicted probabilities for the artificial data?---p-hat

[@konstantinosmexis6561]

What if variable #GAMES LOST was a continuous variable instead of just taking discrete (integer) values?

[@Justin-uz7br]

good video

[@rebbo263]

I am asking this myself for a long time - why make predictions for all observations and then average? Couldn't you just 1) train the model 2) set all variables to their mean except the one you are visualizing 3) Evaluate the function of the model at different points of the variable of interest 4) Plot ..? In that way you only compute lets say 10 points per variable instead going through all observations. Thanks!

[@ocamlmail]

Hi. Thank you for video. Is partial dependence is the same concept as en.wikipedia.org/wiki/Conditional_expectation ? Put it simply, it is visualization of conditional expexctation.