I was looking for perfect GD in short video, and here it is, literally the perfect one so far,, great job and effort I must say. keep making short videos like these.
35:00 Here batch sizes are in exponents of 2 because thats how GPU allocates the chunk. If you put the batch size as 100 then it would still reserve space for 128 and would only use 100. Also, the allocated space that its not going to use will cause issue as GPU can only reserve continuous chunk unlike CPU where it can allocate chunks which are not necessarily continuous.
Batch gradient Descent : an old man , who is walking with a stick slowly but towards his home. Stochastic : his grandchild , who is drunk , but heading towards home. epochs: nopof steps
very very nicely explained. as i always feel and tell everyone for one stop channel for data science this is the best channel as content is real quality and the way Nitish you explains makes it really easy for a beginner to understand the complex topics also. Thank you so much
16:24 but i feel that SGD will take lesser time , because if we apply Batcg GD , it will update weights using whole data set , say its a data set which has 2 million entries , it will take more time , vese SGD will also take good amout of time in this case , i think we cant say which one will take more time in such cases , time taken depends on a lot of factors , so we cant clearly say that BGD will take lesser time .
Good Content, Great Explanation and an exceptionally gifted teacher. Learning is truly made enjoyable by your videos. Thank you for your hard work and clear teaching Nitish Sir.
Sir we are finding the difficult to crack the job in data science placement so please make a dedicated 100 days series of placement tricks & tips, Requested everyone to upvote it , so that sir notice it
While updating the weights of first layer we use weights of last layer in SGD so which one we will consider the old values of them or the updated values. please answer this doubt of mine
Awesome man ... big thanks.. I was very confused about batch_size .. After watching this video i found to know that there is relationship between batch size and gradient decent ❤️❤️ Now deep learning is getting interesting to me
Hey Nitish, can you please tell me which tracking/writing pad did you use in your Machine Learning playlist videos. In the current ones your using a Samsung tab. What about the previous videos? Also could you please tell me the size of it too.
sir ,can you explain what is the difference between GD and Backprop Algorithm , though both are calculating derivatives (chain rule in Neural nets) and then updating weights. I know its kinda silly question to ask
Hi sir i think u have mixed sorcastic and batch when u were explaining the 2 graphs u plotted so batch will give u not stable curve but sorcastic will give u more stable or smooth graphs Bcs more points u have smoother the graph will be less points u have not smooth the graph will be but in batch u have less no of times u have updated the w and b so less points So i dint follow ur logic Also sir even stocastic gradient descent uses for product to calculate y pred so both places ur using dot product so please can you explain me why are you saying that dot product is replacing the loop on batch bcs ur using dot in both the algorithm
![Felix "Unfair" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/Oswujxm2Ag0/mqdefault.jpg)
I was looking for perfect GD in short video, and here it is, literally the perfect one so far,, great job and effort I must say. keep making short videos like these.
You are an excellent teacher; you explain everything in detail so that concepts become clear. Thank you so much for your efforts. You are truly a gem.
35:00 Here batch sizes are in exponents of 2 because thats how GPU allocates the chunk. If you put the batch size as 100 then it would still reserve space for 128 and would only use 100.
Also, the allocated space that its not going to use will cause issue as GPU can only reserve continuous chunk unlike CPU where it can allocate chunks which are not necessarily continuous.
Batch gradient Descent : an old man , who is walking with a stick slowly but towards his home.
Stochastic : his grandchild , who is drunk , but heading towards home.
epochs: nopof steps
Sir videos kafi acche hai seekhne ko bhi bht kuch milta hai bs sir time continuity banaye rkhie sir, kafi wait krna pad jata hai apke videos ke liye
very very nicely explained. as i always feel and tell everyone for one stop channel for data science this is the best channel as content is real quality and the way Nitish you explains makes it really easy for a beginner to understand the complex topics also. Thank you so much
You are an excellent teacher; you explain everything in detail so that concepts become clear. Thank you so much for your efforts.
can u explain what is batch size . is it the no of rows
@@hitanshramtani9076 Yes geek, number of rows or points taken at a time to vectorise and sum up their loss and update the parameters once per batch
What a fantastic teaching method; great job.
wow, 50 jagah gradient descent samajhane k try kiya , ab jake smjh aya
great work
16:24 but i feel that SGD will take lesser time , because if we apply Batcg GD , it will update weights using whole data set , say its a data set which has 2 million entries , it will take more time , vese SGD will also take good amout of time in this case , i think we cant say which one will take more time in such cases , time taken depends on a lot of factors , so we cant clearly say that BGD will take lesser time .
i thankyou from bottom of my heart i am a dataengineer looking for career transition and u made concepts very clear.Thankyou!!!!!!
excellent man.. i salute you. Brilliantly explained the concept these optimization technique..
Good Content, Great Explanation and an exceptionally gifted teacher. Learning is truly made enjoyable by your videos. Thank you for your hard work and clear teaching Nitish Sir.
The explanation of that little nuance about which is faster in SGD and BGD ❤❤
Sir we are finding the difficult to crack the job in data science placement so please make a dedicated 100 days series of placement tricks & tips, Requested everyone to upvote it , so that sir notice it
watch sir website for this
@@Harshupadhyay-u9d please provide name of course and is it paid or not ?
@@janardhan1853 paid
Best explanation on gradient descent
Great learning 😀😀😀
While updating the weights of first layer we use weights of last layer in SGD so which one we will consider the old values of them or the updated values.
please answer this doubt of mine
great, thanks a lot bro, quite detailed explanation!
Wow!!clear explanation i never got from anywhere
Thank you so much sir 🙏🙏🙏
from where we can get this collab notebook
this video contains good content so pls change the thumbnail and put it attractivly like ur photo with background this topics
what is the use of shuffling the data set in SGD , vese bhi randomly samples choose ho rhe hai , in Mini Batch it is fine
Mind-blowing class
Awesome man ... big thanks.. I was very confused about batch_size .. After watching this video i found to know that there is relationship between batch size and gradient decent ❤️❤️ Now deep learning is getting interesting to me
yes sir very very helpful thanks a lot
how can u shre one note ,note with us
Thanks for the wonderful explanation!
Always like your video before watching it.❤
Easy explanation , Thank You
8:34 50 points ka loss aik sath calculate ho kae avg ho ga? otherwise tou loss buhat bara ae ga i guess.
Thank You Very Much Sir.
How to apply mini batch gradient descent. Please show the implementation
Bahut aala... 😊
Hey Nitish, can you please tell me which tracking/writing pad did you use in your Machine Learning playlist videos.
In the current ones your using a Samsung tab. What about the previous videos?
Also could you please tell me the size of it too.
Wacom 6 inch * 8 inch
@@campusx-official thanks so much
19:15 i feel like results will be different in case of bigger data sets
sir,please suggest me some projects od deep learning
for college resume
sir ,can you explain what is the difference between GD and Backprop Algorithm , though both are calculating derivatives (chain rule in Neural nets) and then updating weights. I know its kinda silly question to ask
In back propagation to calculate the loss gradient descent is used.
backpropagation is a type of training method in which gradient descent is used.
Nice video sir
Awesome
Bole tho jhakkas Bhai!!..
So good
Sir apne Lectures ko compile kr k book bna do...
Need to buy
Thank you so much sir
can a get teh dataset please
Sir please provide short notes for this which will help full for us
Hi sir i think u have mixed sorcastic and batch when u were explaining the 2 graphs u plotted so batch will give u not stable curve but sorcastic will give u more stable or smooth graphs
Bcs more points u have smoother the graph will be less points u have not smooth the graph will be but in batch u have less no of times u have updated the w and b so less points
So i dint follow ur logic
Also sir even stocastic gradient descent uses for product to calculate y pred so both places ur using dot product so please can you explain me why are you saying that dot product is replacing the loop on batch bcs ur using dot in both the algorithm
he didn't. Google this topic on geeksforgeeks "ML | Stochastic Gradient Descent (SGD)"
25:20 😂🤣🤣🤣
Best Explanation
u r great as always nitish!!
bhai tune to youtube chod diya tha na .. can anyone clear me on this what actually that video was about?
April fool
😂🤣🤣
Nice 1 bhaiyya
great Video
intuitive
Thanks 🎉
fantastic....
Outstanding
best explaination
GREAT CONTENT
Bhaiya ML projects ke playlist me kuch add Karo na please!!!
❤❤❤
Thank you Man
Thanks Sir
sir ek Python interview ka series bna dijiya sir if it is possible.
❤
❤❤👏👏👏
great explanation sir
best
Very best 👌 sir
Please covet Adam and other optimization techniques also weight untilaization process ,leqening rate decay
Sir i wanna recommend your videos to my south Indian and international friends.. plz make some videos in only English too..thanks:)
na bhai hindi me hi bana ne do aap krish sir ka video dekho ruclips.net/user/krishnaik06
Revising my concepts.
August 11, 2023😅
image processing
finished watching
who r u man🤣
kindly sir
kindly sir aghar ap k sare one note k notes mel jae
please add English Subtitles to your Videos
🤍
Ambani mere 🦜🦜
💓
Thanks sir