Awesome stuff. Searched this video because I was trying to figure out why the scores/sum scores approach wouldn't work and you addressed it first thing. Great job.
What a great explanation! Thank you very much. The why do we choose this formula versus this formula explanation is truly makes everything clear. Thank you once again :)
Thank you!!! This is so much clearer and straighter than 2 20-minutes videos on Softmax from "Machine Learning with Python-From Linear Models to Deep Learning" from MIT! To be fair, the latter explains multiple perspectives and is also good in its sense. But you deliver just the most importaint first bit of what is softmax and what are all these terms are about.
please note that the outputs of Softmax are NOT probabilities but are interpreted as probabilities. This is an important distinction! The same goes for the Sigmoid function. Thanks
well after applying sigmoid you get only one probability p (the other one you can calculate as 1-p) so actually you only need one number in case of sigmoid
I am new to Data Sceince. However, why would a model output 100, 101 and 102 as three outputs unless the input had similarity to all three classes. Even in our daily lives, we would ignore 2 dollar variance on $100 think but complain if something which was originally free but now costs 2 dollars. Question is, why would we give up the usual practice and use some fancy transformation function here ?
What does dP_i/dS_j = -P_i * P_j mean and how did you get it? I understand dP_i/dS_i because S_i is a single variable. But dP_i/DS_j is a whole set of variables (Sum(S_j) = S_1 + S_2 ... S_n) rather than a single one. How are you taking a derivative of that?
4:00 I thank you have express it in a wrong way you wanted to say that we need to go into depth and not just focus on the application that is the façade which here's deriving formula
Hey, thank you for a great video! I have a question: in your example, you said that probabilities between 0,1 and 2 should not be different from 100, 101, and 102. But in the real world, the scale which is used to assess students makes difference and affects probabilities. The difference between 101 and 102 is actually smaller than between 1 and 2, because in the first case the scale is probably much smaller, so the difference between scores is more significant. So wouldn't a model need to predict different probabilities depending on the assessment scale?
My point of view is that the softmax scenario is different from sigmoid scenario. In the sigmoid case, we need to capture the changes in relative scale because subtle changes around the 1/2 prob. point result in significant prob. changes(turns the whole thing around, drop out or not); whereas in the softmax case, there are more outputs and our goal is to select the very case which is most likely to happen, so we are talking about an absolute amount rather than a relative amount(final judge). I guess that's why ritvik said" change in constant shouldn't change our model'.
Oh.. softmax is for multiple classes and sigmoid is for two classes. I get that your i here is the class. In the post below though, is their i observations and k the classes? stats.stackexchange.com/questions/233658/softmax-vs-sigmoid-function-in-logistic-classifier
are you crazy. the moment he did that, I knew it would be fun listening to him. He was focused. Like he said, theory is relevant only in context of practicality.
I really love how you progress step by step instead of directly throwing out the formulas! The best video on TH-cam on the Softmax! +1
For a non-mathematician like myself, this was crystal clear, thanks very much!
tutorials with boards noww...nice one dude...underrated channel I must say!
Much appreciated!
agreed. greetings from russia!
I love it! The choice of examples and the way you explain ... just bravo!
Thank you! 😃
This is excellent! I saw your video on the sigmoid function and both of these explain the why behind their usage.
Glad it was helpful!
This channel is literally the best.
Thanks!
the person who is going to be responsible for me kick starting my ML journey with a good head on my shoulders, thank you ritvik, very enlightening
Awesome stuff. Searched this video because I was trying to figure out why the scores/sum scores approach wouldn't work and you addressed it first thing. Great job.
What a great explanation! Thank you very much.
The why do we choose this formula versus this formula explanation is truly makes everything clear. Thank you once again :)
Thnx. Very clear explanation of the rationale for employing exponential fns instead of linear fns
Great to hear!
Thank you!!! This is so much clearer and straighter than 2 20-minutes videos on Softmax from "Machine Learning with Python-From Linear Models to Deep Learning" from MIT! To be fair, the latter explains multiple perspectives and is also good in its sense. But you deliver just the most importaint first bit of what is softmax and what are all these terms are about.
Glad it helped!
The introduction to softmax which explains why softmax exists helped me a lot understanding it
Great explenations, your addition of the story to the objects really help understanding the material
The only video I needed to understand the SOFTMAX function. Kudos to you!!
Wow...teaching from first principles...I love that!
Glad you liked it!
Woooow ,really liked our teaching approach, awesome!
Now i know why lot of your videos answers WHY question. You give importance to application not the theory alone. concept is very clear. thanks
This is a glorious explanation.
An excellent and straightforward way of explaining. So helpful! Thanks a lot :)
I came for the good-looking teacher but stayed for the really clear an good explanation.
Clearest explanation about softmax.. thank you
Glad it was helpful!
please note that the outputs of Softmax are NOT probabilities but are interpreted as probabilities. This is an important distinction! The same goes for the Sigmoid function. Thanks
1:14, how is it a single dimensional for sigmoid? Shouldn't it be two dimensions?
well after applying sigmoid you get only one probability p (the other one you can calculate as 1-p) so actually you only need one number in case of sigmoid
Great explanation! Easy and helpful!
awesome man..your videos make me less anxious about math..
You can do it!
What an amazing, simple explanation. thank you!
Thank you so much. I now understand why exp is used instead of simple calc.😊
Of course!
Another great content, thank you so much!
My pleasure!
Just great! Thanks, man.
You're welcome!
Great explanation, thank you!
Dude! I really love you.
Thank s for uploading ❤❤❤
Absolutely beautiful.
Very clear explained , thank you, subscribed
thank you very much, you are very good at teaching, very well prepared!
I like the hierarchy implied by the indices on the S vector ;)
How great was this video! thank you
How to dealing with high Xi values? I got 788, 732 for Xi value, and if I exp(788) it gives error bcs it exp results near to infinity
Awesome explanation.... thanks !!!
@ritwikmath want to understand why you chose the subscript N to describe the features; they should be S_1..S_M isn't it?
You're amazing. great teacher
How does this relate to multinomial logistic regression?
Thanks Professor!
really enjoyed that. thanks!
I wish you were my teacher haha great explanation :D Thank you so much ♥
Could you do a video about the maxout unit? I read it on Goodfellow's Deep Learning book, but I did not grasp the intuition behind it clearly.
Great tutorials!
Glad you like them!
Bravo! + Thank you very much!
bro you're a legend
more people should watch this
Thank you very much, clear and helpful to me as a beginer😗
Thanks for such intuitive explanation Sir :)
Very clearly explained!
Beautiful!
Maybe this was explained in a past video, but why is "e" chosen over any other base (like 2 or 3 or pi)...
Solid video, subscribed!
hi there, what is the meaning of the square summation?
I am new to Data Sceince. However, why would a model output 100, 101 and 102 as three outputs unless the input had similarity to all three classes. Even in our daily lives, we would ignore 2 dollar variance on $100 think but complain if something which was originally free but now costs 2 dollars. Question is, why would we give up the usual practice and use some fancy transformation function here ?
What does dP_i/dS_j = -P_i * P_j mean and how did you get it? I understand dP_i/dS_i because S_i is a single variable. But dP_i/DS_j is a whole set of variables (Sum(S_j) = S_1 + S_2 ... S_n) rather than a single one. How are you taking a derivative of that?
that was so much sweet and inspiring
a clear explanation!
Glad you think so!
3:18 very good teacher
Keep going buddy
Holy shit! That makes so much sense
Awesome Brother!
thank you so much !!
You teach better than my grad school professor 😂
Thank you so much! You made it very clear :)
Very helpful!!! Thx!
good video
Thanks 👍
No problem 👍
Do you do one on one tutoring?
I don't understand *why* it's weird that 0 maps to 0 or why we need the probability to be the same for a constant shift...
👍🏻👍🏻👍🏻👍🏻👍🏻👍🏻
4:00 I thank you have express it in a wrong way you wanted to say that we need to go into depth and not just focus on the application that is the façade which here's deriving formula
Yet again an Indian dude is saving me!
Lol 😂
Hey, thank you for a great video! I have a question: in your example, you said that probabilities between 0,1 and 2 should not be different from 100, 101, and 102. But in the real world, the scale which is used to assess students makes difference and affects probabilities. The difference between 101 and 102 is actually smaller than between 1 and 2, because in the first case the scale is probably much smaller, so the difference between scores is more significant. So wouldn't a model need to predict different probabilities depending on the assessment scale?
same question!
My point of view is that the softmax scenario is different from sigmoid scenario. In the sigmoid case, we need to capture the changes in relative scale because subtle changes around the 1/2 prob. point result in significant prob. changes(turns the whole thing around, drop out or not); whereas in the softmax case, there are more outputs and our goal is to select the very case which is most likely to happen, so we are talking about an absolute amount rather than a relative amount(final judge). I guess that's why ritvik said" change in constant shouldn't change our model'.
Thank you
thank you!
Nice video
PRETTY GOOD
thanks
Amazing!
Very good video
Thanks!
Oh.. softmax is for multiple classes and sigmoid is for two classes.
I get that your i here is the class. In the post below though, is their i observations and k the classes?
stats.stackexchange.com/questions/233658/softmax-vs-sigmoid-function-in-logistic-classifier
Hello Boltzmann distribution we meet again, cool nickname
what a shame that this dude is not a professor!!!!!!!!
Why do I need to go to school?
I knew things were about to go down when he flipped the pen.
are you crazy. the moment he did that, I knew it would be fun listening to him. He was focused. Like he said, theory is relevant only in context of practicality.
Very different from how *cough* Siraj *cough* explained this lol
god
Binod stop ads
minute 11-12.30 you are not very clear and going too fast
hey thanks for the feedback, will work on it