Corrections: 6:17 I should have said that the blue points have twice the density of the purple points. 7:08 There should be a 0.05 in the denominator, not a 0.5. Support StatQuest by buying my book The StatQuest Illustrated Guide to Machine Learning or a Study Guide or Merch!!! statquest.org/statquest-store/
Thanks very much for the informative lecture and it is really helpful. UMAP is more and more popular now, could you explain it and compare with tSNE as well? Thanks in advance.
@@statquest UMAP is great, I dont know if it is more popular. There are more stringent reductions out there like ICA. I wonder the thoughts of Josh about it?
@@CompBioQuest I guess it largely depends on the field. Right now, genetics and molecular biology are going bonkers over UMAP. However, ICA is very interesting. Thanks to your question, I found this article which is fascinating: gael-varoquaux.info/science/ica_vs_pca.html
I am always blown away by how you make statistics & machine learning algorithms so simple to understand and how you graciously share your knowldege. Keep up the great work man, you are awesome!
I'm writing this comment while having watched only half way into this video, which is pretty unusual for me! It is so clearly explained! I once glanced at the t-SNE paper and didn't understand it. If this is what it does then this is how things like this should be explained! Really, we need people explaining science like this! It's possible to read scientific papers, but what they fail to do is properly communicate the core idea to the reader so that the reader quickly grasps the big picture and the intent of the mathematical details without getting lost in the details. Frequently, even a missing definition can make reading papers much harder for non experts.
I regret I can't put 1000 likes! I read about 20 articles about t-SNE, they are similar to one another, almost identical - and they don't get me closer to the point. But your video - I watched it 4 times (because the topic is hard, at least for me) with making some and drawing - but finally I understand how it works, up to the point that I can explain it to someone else. So many thanks to you!
It's impressive how you managed to explain the essential concepts of this chain of algorithms in such a clear way! I'm sharing this video with my beginner fellows, who normally flee as soon as I say words like nearest-neighbor or stochastic. Thank you very much!
This explanation almost makes tSME sound like a clustering technique not a reduction technique..... That said, this was by far the best explanation I've heard to date.
@@statquest Now if you can explain how to interpret a tSME plot. This would help immensely as it's virtually impossible to determine the correct perplexity number without understanding how to interpret the plot. This seems like one of those "blackbox" methods which we just trust. Keep up the great work!
Thank you. I am not sure if you remember me from the PCA video. I have a job now. My job do not have high salary, but I could now support you by donating and thank you now. 😊
Josh, i literally love your videos, they are really helping me get through my ADV CS degree. I am going to buy one of your shirts, and wear it on campus as a thank you!
Came here for understanding the t-SNE plots used in single cell transcriptomics - which I finally did, thanks! Overall, you helped me out already plenty of times! To display cells in during cell fate transition/acquisition e.g. different time points during neurodevelopment, often pseudo-temporal ordering is used. Since scRNA seq is becoming more and more popular, this might be a good next topic
Just hear about t-SNE and I did not quite understand how it works so I crossed my fingers hoping that josh did a video of this and of course he did!! haha I have my popcorn ready to enjoy this video :)
this is such an awesome explanation of tsne that i dont need to watch any other video or read any other website/book. I dont think there can be a better explanation. Superlike.
Very nice way of teaching ! ML concepts CLEARLY EXPLAINED and BAM adds lot of curiosity in the videos :) Thanks for your videos. And not to forget your songs are really nice :)
Fantastic video. I really appreciate all the slides that you made to get the animation effect. It really helped. Possibly the best explanation of t-SNE around. Keep up the good work.
Awesome explanation, thank you so much! I read a few papers/books multiple times and barely have a clue, but with your vid I understand the concept just by watching it once!
Hi Josh, I can't thank you enough for how much I have benefitted from your videos even though I do data science as part of my day job. Thank you so much for sharing your knowledge! One request for a video: could you do a video of when to use which methods / models in a typical data science problem? Much appreciated.
Great video - thank you! One small insertion that I think would improve it: at ~2:07, right after showing what projecting on to the X or Y axis would look like, show one more example of projecting onto an arbitrary line to try to retain as much variance as possible (basically PCA). I think this could be done in 15-20 seconds, and would be helpful in comparing t-SNE to one of its most popular alternatives, which is helpful in deciding *when* to use an algorithm - one of the hardest things for beginners like myself.
Hey, love your videos! Just a typo but it should be 0.05 on the values to the right at 07:19. Confused me for a second so might clear things up for others.
Thank you so much! Right now everyone in our department (Systems Genetics at NYU Langone) is using UMAP. There aren't many great videos about it - it would be awesome if you could help us understand what all the hype is about!
Great explanation! Thank you so much... I think their is a typo @7:08. Oh oh... On upper part, sum of all scores is 0.24+0.5 instead of 0.24+ 0.05. BAM. Same mistake on the other equation with same denominator. Double BAM. Results are correct. Triple BAM :-)
Notes for my future revision. STEP 1) CREATE SIMILARITY MATRIX For each data point, get a Similarity Score vs every other data point -> Similarity Matrix. Similarity Score = Distances between points converted/related to Probabilities -> pdf of points on pdf curve = Unscaled Similarities -> Scaled Similarities = Similarities of data are normalised regardless of spread (variation) STEP 2) CREATE NEW "PICTURES" ON A NEW (REDUCED) DIMENSION For each and all data points, position data against other such that the new similarities scores (i.e. the probability score) follow the original similarities score. Objective: Find the new values for each data points (in the new dimensions) such that the difference between the resulting new probability distribution and the original probability distribution is minimised. The origin original distribution is (assumed?) Normal. But the new probability distribution is based on t-distribution, so that the new distances between the data (in the new dimensions) is exaggerated at the edges -> the new distribution spread the data clusters wider. This helps to show the distinct groups of data.
Hi Josh, great videos as always! I'm not sure if there's a video about this already, but could you do one with all the clustering or classification or dimensionality reduction methods compiled together and then compare their differences and similarities and talk about situations when we should use which? For example, after looking at many of the videos, I think I'm already a little lost on if I should use PCA or MDS or t-SNE on my data. Ty.
Hey, love your videos! We are actually using it to help explain key concepts in our application-focused courses. I'd love to see UMAP (similar to t-SNE), which is a bit more scalable.
@@statquest Awesome! I'm using your content in my courses - Students love it. PCA, K-Means, & t-SNE. Will be using your ML videos as well. Your explanations are the best!
t-SNE in concept is a little dense to me so I am watching this video multiple times to think about the nitty gritty of it… I have three perhaps very naive questions so far: 1) with really high dimensional feature space for some data, how do t-SNE algorithms decide how many dimensions to use for the simplified data? In PCA it can be specified by inspecting the variance of data in each of the components to decide that new feature’s “contribution” in grouping/separating the datapoints, is there a similar measure that is used to decide how many dimensions are used in t-SNE? 2) Why is it only used as a visualization technique and not a true dimension-reduction method for data pre-processing in machine learning pipelines? 3) is it possible that the data do not converge in low dimensional space (i.e., you just could not move the second matrix so that it is similar enough to the first one)? I dug out the original 2008 paper from SkLearn citation and as usual was amazed by how you explained the fairly abstract idea in section 2 of the paper in a mere 20-minute long unhurried video, down to the analogy of the repelling and attraction of mapped data in the low dimensional space (the original paper interpreted the gradient decent method used to locate the low dimensional mapping of points as “springs between every point and all other points”) - no important detail is lost in your video yet they are organized in such a way that they follow a clear logic and do not overwhelm. That is mastery of the art of elucidation ❤ Thanks as always for digesting these complicated items for the benefit of the students and present them in simplified yet informative ways, as always!
Thank you very much! For t-SNE, I'm pretty sure it's always used to generate a 2 (or at most 3) dimensional graph that can be visualized. This is because, unlike PCA, where the axes (or PCs) actually represent something (the directions of the most variance), the axes in t-SNE are completely arbitrary. So there's no way to quantify or rank the axes in order of importance. And it is probably possible to have the low dimensional graph fail to converge. That said, if you'd like more details on t-SNE, check out my videos on UMAP - a related technique that is a little more popular: ruclips.net/video/eN0wFzBA4Sc/видео.html and ruclips.net/video/jth4kEvJ3P8/видео.html
1. In Flow Cytometry we use median for almost all data analysis because it best describes the central tendency of the data. Is geo mean anyway better describe Flow Cytometry data or geomean is better for some types of Flow Cytometry experiments? 2. What are the drawbacks of downsampling? If there are any way to identify when to avoid downsampling? 3. What is the batch effect? How to identify and remove it? What is the basic principle of identification? What are the strategies to avoid begin with?
Thank you a lot for the video Josh. Let me point something out, and by minute 10:40, it looks like that t-sne perform a sort of the matrix, instead of minimizing the loss function by gradient descent.
One sincere request .... Can you please make one consolidate video ( could be long video ) which one or two examples of each machine learning concepts you have explained in your channel, also comparing why we are using that particular concept to solve the issue.. what would be issues with other algorithms... Comparison video will surely help to further enhance understanding....
thanks for your great explaination. I just wonder from 5:00 - 5:45, Why when you plot the distance on the normal curve the red and the orange is on different sides of normal curve. I thought distance didn't have direction. Can you please explain more detail about this different direction of the red and orange?
@@statquest yeah, i understood. Because we take p as similarities values so right or left is the same. Thanks a lot. Your videos help me a lot in my machine learning studying.
The mean of the normal curve is 0, the distance from the point we are calculating similarities to and itself. The standard deviation is a function of the density of the points around it and, I believe, the perplexity fudge factor. I can't remember the formula off the top of my head, but the higher the density of point, the smaller the standard deviation, and the lower the density of points, the higher the standard deviation.
Hi @StatQuest with Josh Starmer, great video! The thing I am missing is what is happening in each of this steps to move each point. What are the 'attract' and 'repel' real values and how they are use to make the Similarity Matrices closer each of these steps?
This is a good question. The actual math is a little too messy to put in this comment, however, the idea is that the matrices are made similar using Gradient Descent, and that's where the attractions and repulsions come in. Here's a quote from the original paper (the link to the paper comes after the quote): Physically, the gradient [ minimized by gradient descent ] may be interpreted as the resultant force created by a set of springs between the [low dimensional point A] yi and all other [low dimensional points] yj. All springs exert a force along the direction (yi −y j). The spring between yi and y j repels or attracts the map points depending on whether the distance between the two in the map is too small or too large to represent the similarities between the two high-dimensional datapoints. The force exerted by the spring between yi and y j is proportional to its length, and also proportional to its stiffness, which is the mismatch (p j|i −qj|i + pi| j −qi| j) between the pairwise similarities of the data points and the map points. Here's the link to the paper: www.jmlr.org/papers/volume9/vandermaaten08a/vandermaaten08a.pdf
Hi Josh, quality content! This channel continuously helps me to understand the idea behind so that the dry textbook explanations actually make sense. I still have a question. When you calculate the unscaled similarity score, how do you exactly determine the width of your guassian? I get it in the example that we already know the cluster. If I only want to visualize the data without having pre-defined clusters, what happens then?
I talk more about the details of t-SNE and how it works in my videos on UMAP: ruclips.net/video/eN0wFzBA4Sc/видео.html and ruclips.net/video/jth4kEvJ3P8/видео.html
Hi Josh, great video, many thanks! Anyway, I still don't get how do you determine the distribution properties (like standard deviation) for calculating unscaled similarity between two points. When you introduced half as dense cluster as the others, you used normal distribution with standard deviation doubled, what is quite intuitve. But you knew that this cluster is just half as dense as the others. The question is, how to know the properties of these distribution curves?
Corrections:
6:17 I should have said that the blue points have twice the density of the purple points.
7:08 There should be a 0.05 in the denominator, not a 0.5.
Support StatQuest by buying my book The StatQuest Illustrated Guide to Machine Learning or a Study Guide or Merch!!! statquest.org/statquest-store/
Thanks very much for the informative lecture and it is really helpful. UMAP is more and more popular now, could you explain it and compare with tSNE as well? Thanks in advance.
@@linweitao6470 I should have a UMAP StatQuest ready in a few weeks. I'm working on it right now.
@@statquest Thanks again!
@@statquest UMAP is great, I dont know if it is more popular. There are more stringent reductions out there like ICA. I wonder the thoughts of Josh about it?
@@CompBioQuest I guess it largely depends on the field. Right now, genetics and molecular biology are going bonkers over UMAP. However, ICA is very interesting. Thanks to your question, I found this article which is fascinating: gael-varoquaux.info/science/ica_vs_pca.html
I am always blown away by how you make statistics & machine learning algorithms so simple to understand and how you graciously share your knowldege. Keep up the great work man, you are awesome!
Thank you very much! :)
Whenever I find statistics technique I have never seen in scientific article, I always visit your channel. Thanks a lot!!
Happy to help! :)
I'm writing this comment while having watched only half way into this video, which is pretty unusual for me!
It is so clearly explained! I once glanced at the t-SNE paper and didn't understand it. If this is what it does then this is how things like this should be explained!
Really, we need people explaining science like this! It's possible to read scientific papers, but what they fail to do is properly communicate the core idea to the reader so that the reader quickly grasps the big picture and the intent of the mathematical details without getting lost in the details.
Frequently, even a missing definition can make reading papers much harder for non experts.
I'm glad you liked this video so much! :)
I regret I can't put 1000 likes! I read about 20 articles about t-SNE, they are similar to one another, almost identical - and they don't get me closer to the point. But your video - I watched it 4 times (because the topic is hard, at least for me) with making some and drawing - but finally I understand how it works, up to the point that I can explain it to someone else. So many thanks to you!
HOORAY!!! TRIPLE BAM! I'm glad the video was helpful. BAM! :)
I never leave comments, but I really feel the need to thank you for being able to explain this in such a simple way
Thank you! :)
Josh is so far my favorite RUclipsr that is able to explain complex stats concepts so smoothly.
Thank you so much! :)
I never knew machine learning could be as simple as... BAM
Thats like the most important lesson.
Double bam 💥
Just a random comment so that someone can say triple bam
Triple bam 💥
hurayyyy we have made it to the END !!!
As entertaining as watching a Walt t-SNE movie!
You made me laugh out loud! BAM! :)
Best stat-word-play of the year! 😂
I am a student in Japan.
I'm not good at English, but it was very easy to understand and I learned a lot:)
Awesome! :)
The only educational channel which brings a smile to my face.
bam!
Josh.. Your explanation is always "simple and easy to understand" even for layman.You are simply "The life Saviour" !!!
Thank you so much :)
Hooray! I'm glad my video was helpful. :)
It's impressive how you managed to explain the essential concepts of this chain of algorithms in such a clear way! I'm sharing this video with my beginner fellows, who normally flee as soon as I say words like nearest-neighbor or stochastic.
Thank you very much!
Thank you very much! :)
🤣🤣🤣🤣it's that terrifying?!? Barbara Oakley in her book, "a mind for numbers" called them zombies🤣🤣🤣
This explanation almost makes tSME sound like a clustering technique not a reduction technique..... That said, this was by far the best explanation I've heard to date.
That's a good observation. In many ways t-SNE is a hybrid method that reduces dimensions by clustering.
@@statquest Now if you can explain how to interpret a tSME plot. This would help immensely as it's virtually impossible to determine the correct perplexity number without understanding how to interpret the plot. This seems like one of those "blackbox" methods which we just trust. Keep up the great work!
Great explanations! Can you please do a video explaining UMAP and potentially how it compares to t-SNE? Thanks!
+1
+1
+1
+1
+1
I was so confusing about t-SNE until I watched this. It's clear and very easy to understand! Thank you! Like your BAM. :D
BAM! :)
It's rare to come across such a brilliant explanation.
Thank you! :)
I really can't appreciate you enough for your videos.
Books and blogs only make sense after I watch your videos!
Thank you very much! :)
"This is Josh Starmer, and you're watching Tisney Channel!"
Triple BAM! :)
Thank you. I am not sure if you remember me from the PCA video. I have a job now. My job do not have high salary, but I could now support you by donating and thank you now. 😊
WOW! Thank you so much. And congratulations on getting a job!!! HOORAY!!! TRIPLE BAM! :)
@@statquest Keep doing great work sir! Also, it would be great if you could make a video about the comparation between clustering methods. 😁
@@tuongminhquoc Thanks and I'll keep that in mind!
Josh, i literally love your videos, they are really helping me get through my ADV CS degree. I am going to buy one of your shirts, and wear it on campus as a thank you!
That would be awesome!!! Thank you very much! :)
Came here for understanding the t-SNE plots used in single cell transcriptomics - which I finally did, thanks! Overall, you helped me out already plenty of times!
To display cells in during cell fate transition/acquisition e.g. different time points during neurodevelopment, often pseudo-temporal ordering is used.
Since scRNA seq is becoming more and more popular, this might be a good next topic
Same here, and I did not expect to understand so fast and clearly!
Just hear about t-SNE and I did not quite understand how it works so I crossed my fingers hoping that josh did a video of this and of course he did!! haha
I have my popcorn ready to enjoy this video :)
Worth it!
BAM! :)
Why I couldn't stop bamming the like button??!! You're the best Josh!!
Thanks!
Very clearly explained!
Loved the way you explained such a complicated concept so intuitively.
Thank you.
Glad it was helpful!
this is such an awesome explanation of tsne that i dont need to watch any other video or read any other website/book. I dont think there can be a better explanation. Superlike.
you are the hero, keep explaining complex thing into simple. thankss
Thank you! :)
Very nice way of teaching ! ML concepts CLEARLY EXPLAINED and BAM adds lot of curiosity in the videos :) Thanks for your videos. And not to forget your songs are really nice :)
Thank you!
You are incredible, Josh Starmer!! I loved this
Thank you! :)
The Best tutorial and explanation for TSNE so far! It's of great help! Thanks a lot!
Thanks! :)
Brilliant explanation, this has been bugging me all day, thank you!!
Glad it helped!
Very well explained ! Your video was recommended to us by our professors at ETH-Zürich.:)
Fantastic video. I really appreciate all the slides that you made to get the animation effect. It really helped. Possibly the best explanation of t-SNE around. Keep up the good work.
Excellently explained! I really like your simple, clear, concise explanation - those 3 factors make a world of difference. And, great animations.
Awesome, thank you!
I just love the way you start all your videos! Stat-Questtttttt :)
BAM! :)
This is the best video for t-SNE that I have ever seen. Thanks a lot, man
Awesome explanation, thank you so much! I read a few papers/books multiple times and barely have a clue, but with your vid I understand the concept just by watching it once!
Thanks a lot!! These videos are much more clear than any article!
A video explaining UMAP (related to t-SNE) would be awesome !
I'm working on UMAP. For now, however, know that it is almost 100% the same as t-SNE. The differences are very subtle.
Hi Josh, I can't thank you enough for how much I have benefitted from your videos even though I do data science as part of my day job. Thank you so much for sharing your knowledge!
One request for a video: could you do a video of when to use which methods / models in a typical data science problem? Much appreciated.
That's a great idea.
One word reaction after watching this video --> AWESOME!!
Thank you so much 😀!
Thank you so much for this great resource and how much investment you have made into it. I have understood this well.
Glad it was helpful!
Hello Josh, thank you for coming with such incredible videos. Data scientist’s life becomes easy.😬
Thank you! :)
StatQuest with Josh Starmer Hi a request to do a tutorial of UMAP.
Great video - thank you! One small insertion that I think would improve it: at ~2:07, right after showing what projecting on to the X or Y axis would look like, show one more example of projecting onto an arbitrary line to try to retain as much variance as possible (basically PCA). I think this could be done in 15-20 seconds, and would be helpful in comparing t-SNE to one of its most popular alternatives, which is helpful in deciding *when* to use an algorithm - one of the hardest things for beginners like myself.
Thanks for the tip!
Difficult concept made so simple. Just brilliant!!!!
Thanks a lot 😊!
Thanks a lot. I really struggled to understand the concept first time I came across it in a book. Your video helped a lot. Great job!
Great as always. I've heard of t-SNE before, but this was my first real introduction to it. Definitely want to go look at some more resources now.
excellent explanation , this intuition helps to follow maths behind t-SNE
Hey, love your videos!
Just a typo but it should be 0.05 on the values to the right at 07:19. Confused me for a second so might clear things up for others.
OH God, this is a great explanation, as Radel mention below, it would be nice to have an extended video of the algorithm as the one from PCA!!
Thank you! Yes, one day I'll break the actual equations down and do "step-by-step" explanation of t-SNE.
Looking forward to this.
Kudos, I understood so effortlessly....tripple BAM!!!
Thanks! :)
Thanks for such a clear explanation. You know, your channel already in the top list for me and very soon I'll watch all your videos..
i am a huge fan of this channel! greetings from brazil ^^
Muito obrigado! :)
You make a complex idea becomes so simple and understanding ! Great video. Thanks a lot
Super Mega BAM !! So great at what you do as always ... Tons of love sent your way ! Keep up the amazing work :D
Thanks so much!!
I need to watch 3 more times to fully understand. TRIPLE BAM!!!
:)
Dude this is super clear. Love the content! BAM
Thank you very much! :)
Excellent video! Perhaps you could add another video where you go through the actual algorithm and how the moves is actually computed.
yes!!! pleasee!!
Love the vid. I was wondering how tsne works and you broke it down great and the explanation for the t distribution was short and to the point.
Thank you! :)
"Clearly Expalined" indeed!
Thank you so much! Right now everyone in our department (Systems Genetics at NYU Langone) is using UMAP. There aren't many great videos about it - it would be awesome if you could help us understand what all the hype is about!
UMAP is on the to-do list. I hope to get to it in the spring.
thank you so much for this nice explanation. will help me a lot in my exams
Glad to hear that!
Wish I could *Triple Bam* like this video! Such a simple explanation. Thanks a lot Josh :-)
Glad you liked it!
"Bam, I made that terminology up" :D :D , great vid, keep up the good work.
Thanks! 😁
I never thought I'd not understand a statquest video! :(
Bummer. What time point was confusing?
Subscribed because that intro gave me life!
Ha!!! Thanks! :)
Amazing work! perfectly explained!!!
Thanks a lot!
Your speak like Kevin from The Office. Great explanation, thanks a lot:)
Great explanation! Thank you so much... I think their is a typo @7:08. Oh oh... On upper part, sum of all scores is 0.24+0.5 instead of 0.24+ 0.05. BAM. Same mistake on the other equation with same denominator. Double BAM. Results are correct. Triple BAM :-)
Thanks! I added that note to the pinned comment.
Thank you very much Josh . You made it easier to understand.
Hooray! I'm glad the video was helpful! :)
your explanation is very very good! thanks!!!
Thank you! :)
great explanation especially for beginners.Thanks
Thank you! :)
Notes for my future revision.
STEP 1)
CREATE SIMILARITY MATRIX
For each data point, get a Similarity Score vs every other data point
-> Similarity Matrix.
Similarity Score
= Distances between points converted/related to Probabilities
-> pdf of points on pdf curve
= Unscaled Similarities
-> Scaled Similarities
= Similarities of data are normalised regardless of spread (variation)
STEP 2)
CREATE NEW "PICTURES" ON A NEW (REDUCED) DIMENSION
For each and all data points, position data against other such that the new similarities scores (i.e. the probability score) follow the original similarities score.
Objective: Find the new values for each data points (in the new dimensions) such that the difference between the resulting new probability distribution and the original probability distribution is minimised.
The origin original distribution is (assumed?) Normal. But the new probability distribution is based on t-distribution, so that the new distances between the data (in the new dimensions) is exaggerated at the edges -> the new distribution spread the data clusters wider. This helps to show the distinct groups of data.
Noted
Thanks really great videos understood concepts so well
Glad it was helpful!
Hi Josh, great videos as always! I'm not sure if there's a video about this already, but could you do one with all the clustering or classification or dimensionality reduction methods compiled together and then compare their differences and similarities and talk about situations when we should use which? For example, after looking at many of the videos, I think I'm already a little lost on if I should use PCA or MDS or t-SNE on my data. Ty.
Thanks! I'll keep that in mind.
Great videos! Great channel! Big thumbs UP!
Big thanks!
Hey, love your videos! We are actually using it to help explain key concepts in our application-focused courses. I'd love to see UMAP (similar to t-SNE), which is a bit more scalable.
Thank you so much! It's on the to-do list. :)
@@statquest Awesome! I'm using your content in my courses - Students love it. PCA, K-Means, & t-SNE. Will be using your ML videos as well. Your explanations are the best!
VERY CLEAR EXPLANATIONS :) THANK YOU FOR ALL YOUR VIDEOS
best channel ever
Thank you!
Holy shit, man's actually still replying on a 6 year old video
@@tedchirvasiu :)
I am at the intro and love it already!
BAM! :)
Excellent intro to tSNE
Thank you! :)
Thanks for this wonderful video❤️
Glad you enjoyed it!
Incredibly helpful and well presented. Thank you.
t-SNE in concept is a little dense to me so I am watching this video multiple times to think about the nitty gritty of it… I have three perhaps very naive questions so far: 1) with really high dimensional feature space for some data, how do t-SNE algorithms decide how many dimensions to use for the simplified data? In PCA it can be specified by inspecting the variance of data in each of the components to decide that new feature’s “contribution” in grouping/separating the datapoints, is there a similar measure that is used to decide how many dimensions are used in t-SNE? 2) Why is it only used as a visualization technique and not a true dimension-reduction method for data pre-processing in machine learning pipelines? 3) is it possible that the data do not converge in low dimensional space (i.e., you just could not move the second matrix so that it is similar enough to the first one)?
I dug out the original 2008 paper from SkLearn citation and as usual was amazed by how you explained the fairly abstract idea in section 2 of the paper in a mere 20-minute long unhurried video, down to the analogy of the repelling and attraction of mapped data in the low dimensional space (the original paper interpreted the gradient decent method used to locate the low dimensional mapping of points as “springs between every point and all other points”) - no important detail is lost in your video yet they are organized in such a way that they follow a clear logic and do not overwhelm. That is mastery of the art of elucidation ❤
Thanks as always for digesting these complicated items for the benefit of the students and present them in simplified yet informative ways, as always!
Thank you very much! For t-SNE, I'm pretty sure it's always used to generate a 2 (or at most 3) dimensional graph that can be visualized. This is because, unlike PCA, where the axes (or PCs) actually represent something (the directions of the most variance), the axes in t-SNE are completely arbitrary. So there's no way to quantify or rank the axes in order of importance. And it is probably possible to have the low dimensional graph fail to converge. That said, if you'd like more details on t-SNE, check out my videos on UMAP - a related technique that is a little more popular: ruclips.net/video/eN0wFzBA4Sc/видео.html and ruclips.net/video/jth4kEvJ3P8/видео.html
can you explain the math more?
the only info that's stuck clearly in my head in BAM..
😂😂😂
Well done! I would love to see videos on handling data outliers for regressions. Thanks!
1. In Flow Cytometry we use median for almost all data analysis because it best describes the central tendency of the data. Is geo mean anyway better describe Flow Cytometry data or geomean is better for some types of Flow Cytometry experiments?
2. What are the drawbacks of downsampling? If there are any way to identify when to avoid downsampling?
3. What is the batch effect? How to identify and remove it? What is the basic principle of identification? What are the strategies to avoid begin with?
Great questions!
Thanks a million for this masterpiece !!!
Thank you!
Thank you a lot for the video Josh.
Let me point something out, and by minute 10:40, it looks like that t-sne perform a sort of the matrix, instead of minimizing the loss function by gradient descent.
Good point. I represented it as a matrix because, internally, all of the similarity scores are maintained that way.
One sincere request .... Can you please make one consolidate video ( could be long video ) which one or two examples of each machine learning concepts you have explained in your channel, also comparing why we are using that particular concept to solve the issue.. what would be issues with other algorithms...
Comparison video will surely help to further enhance understanding....
That's a good idea, a worked out machine learning example from start to finish, and I'll put it on the to-do list.
StatQuest with Josh Starmer thanks a lot Joshh... Waiting for it
thanks for your great explaination. I just wonder from 5:00 - 5:45, Why when you plot the distance on the normal curve the red and the orange is on different sides of normal curve. I thought distance didn't have direction. Can you please explain more detail about this different direction of the red and orange?
The normal curve is symmetrical, so we can puts the dots on either side. In this case, I used both sides so that not all the dots would overlap.
@@statquest yeah, i understood. Because we take p as similarities values so right or left is the same. Thanks a lot. Your videos help me a lot in my machine learning studying.
Great video. Just if you could explain a bit who the shape of normal curve has been determined would be wonderful! I'm a bit confused there at 4:41.
The mean of the normal curve is 0, the distance from the point we are calculating similarities to and itself. The standard deviation is a function of the density of the points around it and, I believe, the perplexity fudge factor. I can't remember the formula off the top of my head, but the higher the density of point, the smaller the standard deviation, and the lower the density of points, the higher the standard deviation.
Excellent work, thank you !!
Thanks!
Hi @StatQuest with Josh Starmer, great video!
The thing I am missing is what is happening in each of this steps to move each point. What are the 'attract' and 'repel' real values and how they are use to make the Similarity Matrices closer each of these steps?
This is a good question. The actual math is a little too messy to put in this comment, however, the idea is that the matrices are made similar using Gradient Descent, and that's where the attractions and repulsions come in. Here's a quote from the original paper (the link to the paper comes after the quote):
Physically, the gradient [ minimized by gradient descent ] may be interpreted as the resultant force created by a set of springs between the [low dimensional point A] yi and all other [low dimensional points] yj. All springs exert a force along the direction (yi −y j). The spring between yi and y j repels or attracts the map points depending on whether the distance between the two in the map is too small or too large to represent the similarities between the two high-dimensional datapoints. The force exerted by the spring between yi and y j
is proportional to its length, and also proportional to its stiffness, which is the mismatch (p j|i −qj|i + pi| j −qi| j) between the pairwise similarities of the data points and the map points.
Here's the link to the paper: www.jmlr.org/papers/volume9/vandermaaten08a/vandermaaten08a.pdf
Great explainations!
Thanks!
Hi Josh, quality content! This channel continuously helps me to understand the idea behind so that the dry textbook explanations actually make sense. I still have a question. When you calculate the unscaled similarity score, how do you exactly determine the width of your guassian? I get it in the example that we already know the cluster. If I only want to visualize the data without having pre-defined clusters, what happens then?
I talk more about the details of t-SNE and how it works in my videos on UMAP: ruclips.net/video/eN0wFzBA4Sc/видео.html and ruclips.net/video/jth4kEvJ3P8/видео.html
This is a great explanation thank you!
Glad you enjoyed it!
Nice explanation of t-SNE for beginners.
Very well explained.
Thanks! :)
Hi Josh, great video, many thanks! Anyway, I still don't get how do you determine the distribution properties (like standard deviation) for calculating unscaled similarity between two points. When you introduced half as dense cluster as the others, you used normal distribution with standard deviation doubled, what is quite intuitve. But you knew that this cluster is just half as dense as the others. The question is, how to know the properties of these distribution curves?
You estimate it from the data.