Linear regression kind of is machine learning in the most basic sense. I think it's actually quite instructive to call it that. It's the grand daddy technique.
Machine learning is like I said data science speak for a lot of regressions.
I think it is more the social scientist in me is being cranky. I am not as bad as my senior person. I think she wants to just burn down that description a million times over.
0
Options
ShivahnUnaware of her barrel shifter privilegeWestern coastal temptressRegistered User, Moderatormod
lots of professor listings lately there are biology but with a machine learning component
i wonder if i could trick the olds into thinking i do machine learning
step 1) what is it tho????
That sounds like it was tailored for me
But I don't want to academia
0
Options
BrodyThe WatchThe First ShoreRegistered Userregular
lots of professor listings lately there are biology but with a machine learning component
i wonder if i could trick the olds into thinking i do machine learning
step 1) what is it tho????
They just need professors who can teach machines, which are quickly replacing human students at universities across the world.
"I will write your name in the ruin of them. I will paint you across history in the color of their blood."
How's it feel to play second fiddle to a fish Preacher?
Well since on my birthday I'm choosing to take my son to an aquarium really its just how this birthday is going to go. He cares little I'm getting older he just wants to see the fishies
If it makes you feel better VishNub's engagement was overshadowed by a dead crustacean
How's it feel to play second fiddle to a fish Preacher?
Well since on my birthday I'm choosing to take my son to an aquarium really its just how this birthday is going to go. He cares little I'm getting older he just wants to see the fishies
If it makes you feel better VishNub's engagement was overshadowed by a dead crustacean
Its perfectly fine, I don't mind my birthday but I never make a big deal out of it really. Like I normally take it off (don't have to this year since its a saturday), and since it falls either on or near mothers day I never really get to celebrate it very much anyway.
My wife got me some star wars stuff, and the weather looks nice to take my son to a fun thing (we're all wearing shark shirts too, so I'm daddy shark, wife is mommy shark and we got our son baby shark we're that family), so really its all I can hope for. Plus I have a communicable disease I can spread around, so good marks all over.
I would like some money because these are artisanal nuggets of wisdom philistine.
How's it feel to play second fiddle to a fish Preacher?
Well since on my birthday I'm choosing to take my son to an aquarium really its just how this birthday is going to go. He cares little I'm getting older he just wants to see the fishies
This is absolutely what I was thinking of for the OP
Honestly the one part that always catches me now is when I see notice of someone dying and they are a similar age and instead of comments about "oh man they were young" its "meh they were middle aged" that's the part that hits me in the feels, right in the feely bits.
I would like some money because these are artisanal nuggets of wisdom philistine.
I meant to ask Shiv, which self-guided course work did you use for the ML stuff? PM if needed.
I did Andrew Ng's basic introduction on Coursera, then most of Geoffrey Hinton's old course (no longer enrolling) there. I feel like there was other stuff, but I can't recall off the top of my head. I also did a lot of going to blogs and struggling through math and occasionally reading papers on ArXiv. I am sure there is a lot more advanced shit I could learn in another course, but the introduction to machine learning one taught me all the building blocks (linear/logistic regression, neural networks, the math behind them, and how to combat overfitting) and now I have the knowledge to expand outside of that arena without guidance (e.g., recurrent networks including LSTMs (which I've implemented despite never taking a course that discussed them), convolutional networks, etc).
I get annoyed when I see linear regression marked as "machine learning" on anything. I know machine learning is basically data science speak for regressions but ugh.
At least random forest is interesting compared to old fashioned linear regression in this presentation.
Also linear regression seems like a terrible fit for the model they were making.
*grumbles*
Oh and stop using base R graph outputs, they are ugly.
It is machine learning
I think it's important to call algorithms like linear regression machine learning so that people stop thinking ~*~Machine Learning~*~ is the special all-powerful magic thing
Really though FF6 had some of the best music of the whole series. Had a real super villain in it. Had a whale ship going to the moon. It had everything.
Machine learning is super broad, though, and I think most applications of it are probably not linear regression. I feel like neural nets are used for most of the recent advances in machine learning (e.g., ImageNet stuff).
I tend to distinguish between 'traditional machine learning' and 'deep learning'
HOWEVER
Deep learning is often not appropriate for any given task! It really depends on the data and on the problem. XGboost is a very good algorithm that will outperform neural nets on many tasks. Logistic regression is a good algorithm that will outperform neural nets on some tasks! In the task that I just completed, a CNN did best on most tasks, but for the one binary classification problem, a non-deep learning method worked best.
I need to play FF12 at some point. Also 6, I've only ever played like an hour or two of 6 and then I get sidetracked. But I have a SNES Classic, it's right there!
lol I was just listening to the new Reply All and it gets to the end of a really moving segment about depression and what that looks like and what it looks like when you move past it and it's got me a little teary-eyed and then it just pops so quickly into an ad read ("sign up for ZipRecruitr!") like okay nevermind I guess
Machine learning is super broad, though, and I think most applications of it are probably not linear regression. I feel like neural nets are used for most of the recent advances in machine learning (e.g., ImageNet stuff).
I tend to distinguish between 'traditional machine learning' and 'deep learning'
HOWEVER
Deep learning is often not appropriate for any given task! It really depends on the data and on the problem. XGboost is a very good algorithm that will outperform neural nets on many tasks. Logistic regression is a good algorithm that will outperform neural nets on some tasks! In the task that I just completed, a CNN did best on most tasks, but for the one binary classification problem, a non-deep learning method worked best.
No
Deep learning is literal magic and the shareholders demand it
0
Options
Hi I'm Vee!Formerly VH; She/Her; Is an E X P E R I E N C ERegistered Userregular
How is Jon Bois so fucking good at making videos?
The Bob Emergency might be the best one he's ever done, and that's saying a lot. What an incredible talent. And it's not even done, there's gonna be a part 2!
anyway I'm putting together a presentation on machine learning algorithms that we used to solve a problem for the contract that just got shuttered, and I'm super annoyed that the senior manager guy put in this extra slide that was like 'here's a table showing what algorithm we recommend for each problem' and he left it blank
I actually hit him up on skype chat and was like "hey are the numbers I included in this presentation uninterpretable? I highlighted in red the ones that are highest."
as in--can you seriously not look at the tables and see which algorithm had the highest metrics because seriously
they are written out and also highlighted in red
why did you leave the table blank
why did I code everything in this project AND ALSO make the powerpoint, not just the parts about the data but also all the part about project objectives etc, rewriting everything from scratch because the original one that the managers made was awful?
lol I was just listening to the new Reply All and it gets to the end of a really moving segment about depression and what that looks like and what it looks like when you move past it and it's got me a little teary-eyed and then it just pops so quickly into an ad read ("sign up for ZipRecruitr!") like okay nevermind I guess
unfortunately it isnt a new reply all so im pretty annoyed
Okay, going to just go through this thought process here. But I know this is a weird fault like I have noticed between social science folks (sociologist, epis, and so on) and data science.
My brain doesn't tend to connect linear regression with Machine Learning (yes, again regressions are Machine Learning I know this). That is because to me they are regressions and just a long running part of the social science I was taught. My brain tends to attach machine learning to usually newer algorithms. Not specifically neural nets but more of the things that are developed more heavily with computing or what my data science teacher tended to call "black box" algorithms.
And that is very much due to my training and connection to social science which has been using a lot of the "machine learning" regressions in research for a long time before it became a data science thing.
There are biases here I fully accept and recognize. And I can even trace how they have developed through my education and the type of folks I work with.
0
Options
ChanusHarbinger of the Spicy Rooster ApocalypseThe Flames of a Thousand Collapsed StarsRegistered Userregular
Machine learning is super broad, though, and I think most applications of it are probably not linear regression. I feel like neural nets are used for most of the recent advances in machine learning (e.g., ImageNet stuff).
I tend to distinguish between 'traditional machine learning' and 'deep learning'
HOWEVER
Deep learning is often not appropriate for any given task! It really depends on the data and on the problem. XGboost is a very good algorithm that will outperform neural nets on many tasks. Logistic regression is a good algorithm that will outperform neural nets on some tasks! In the task that I just completed, a CNN did best on most tasks, but for the one binary classification problem, a non-deep learning method worked best.
Yeah I think the simpler or more physically derived your task is, the better conventional shallow learning works. One of its biggest assets is resistance to overfitting and easier inspection of the resulting classifier or transform.
Thousands of hot, local singles are waiting to play at bubbulon.com.
lol I was just listening to the new Reply All and it gets to the end of a really moving segment about depression and what that looks like and what it looks like when you move past it and it's got me a little teary-eyed and then it just pops so quickly into an ad read ("sign up for ZipRecruitr!") like okay nevermind I guess
unfortunately it isnt a new reply all so im pretty annoyed
Posts
Sexy dust
Machine learning is like I said data science speak for a lot of regressions.
I think it is more the social scientist in me is being cranky. I am not as bad as my senior person. I think she wants to just burn down that description a million times over.
That sounds like it was tailored for me
But I don't want to academia
They just need professors who can teach machines, which are quickly replacing human students at universities across the world.
The Monster Baru Cormorant - Seth Dickinson
Steam: Korvalain
If it makes you feel better VishNub's engagement was overshadowed by a dead crustacean
I have like this annoying cold right now, so I'm having celebratory mucous break up coughs.
pleasepaypreacher.net
Its perfectly fine, I don't mind my birthday but I never make a big deal out of it really. Like I normally take it off (don't have to this year since its a saturday), and since it falls either on or near mothers day I never really get to celebrate it very much anyway.
My wife got me some star wars stuff, and the weather looks nice to take my son to a fun thing (we're all wearing shark shirts too, so I'm daddy shark, wife is mommy shark and we got our son baby shark we're that family), so really its all I can hope for. Plus I have a communicable disease I can spread around, so good marks all over.
pleasepaypreacher.net
Same.
Mine are from quitting smoking, though. Going on 7 months and my lungs are starting to clear but still got some crap in there.
This is absolutely what I was thinking of for the OP
...
Absolutely.
pleasepaypreacher.net
I did Andrew Ng's basic introduction on Coursera, then most of Geoffrey Hinton's old course (no longer enrolling) there. I feel like there was other stuff, but I can't recall off the top of my head. I also did a lot of going to blogs and struggling through math and occasionally reading papers on ArXiv. I am sure there is a lot more advanced shit I could learn in another course, but the introduction to machine learning one taught me all the building blocks (linear/logistic regression, neural networks, the math behind them, and how to combat overfitting) and now I have the knowledge to expand outside of that arena without guidance (e.g., recurrent networks including LSTMs (which I've implemented despite never taking a course that discussed them), convolutional networks, etc).
It is machine learning
I think it's important to call algorithms like linear regression machine learning so that people stop thinking ~*~Machine Learning~*~ is the special all-powerful magic thing
Really though FF6 had some of the best music of the whole series. Had a real super villain in it. Had a whale ship going to the moon. It had everything.
Cyans theme the first time I heard it. I remember just putting the controller down and listening for a long while.
I will support ANYONE buying FF12 for any reason
It’s so good just remember Vaan is NOT the main character
I tend to distinguish between 'traditional machine learning' and 'deep learning'
HOWEVER
Deep learning is often not appropriate for any given task! It really depends on the data and on the problem. XGboost is a very good algorithm that will outperform neural nets on many tasks. Logistic regression is a good algorithm that will outperform neural nets on some tasks! In the task that I just completed, a CNN did best on most tasks, but for the one binary classification problem, a non-deep learning method worked best.
Oh shit yeah my bad. Im sad now.
but I remember the battle system being incredibly fun
Didn’t they just re-release that one?
Arch,
https://www.youtube.com/watch?v=t_goGR39m2k
No
Deep learning is literal magic and the shareholders demand it
The Bob Emergency might be the best one he's ever done, and that's saying a lot. What an incredible talent. And it's not even done, there's gonna be a part 2!
I actually hit him up on skype chat and was like "hey are the numbers I included in this presentation uninterpretable? I highlighted in red the ones that are highest."
as in--can you seriously not look at the tables and see which algorithm had the highest metrics because seriously
they are written out and also highlighted in red
why did you leave the table blank
why did I code everything in this project AND ALSO make the powerpoint, not just the parts about the data but also all the part about project objectives etc, rewriting everything from scratch because the original one that the managers made was awful?
annoying!
i'm 38, 39 this year
My brain doesn't tend to connect linear regression with Machine Learning (yes, again regressions are Machine Learning I know this). That is because to me they are regressions and just a long running part of the social science I was taught. My brain tends to attach machine learning to usually newer algorithms. Not specifically neural nets but more of the things that are developed more heavily with computing or what my data science teacher tended to call "black box" algorithms.
And that is very much due to my training and connection to social science which has been using a lot of the "machine learning" regressions in research for a long time before it became a data science thing.
There are biases here I fully accept and recognize. And I can even trace how they have developed through my education and the type of folks I work with.
Yeah I think the simpler or more physically derived your task is, the better conventional shallow learning works. One of its biggest assets is resistance to overfitting and easier inspection of the resulting classifier or transform.
this is an outrage