Why does every beginner data scientist fall for the "deep learning trap"?
A thread. 🧵
#DataScience #DeepLearningTrap
A thread. 🧵
#DataScience #DeepLearningTrap
When I was first learning data science this cost me at least 6-months. Seriously...
I was building a model for predicting which quotes would become orders.
I had just finished using a linear regression (didn't know about logistic yet) to make a predictive model.
I was building a model for predicting which quotes would become orders.
I had just finished using a linear regression (didn't know about logistic yet) to make a predictive model.
Yeah I know - I was a noobie using regression instead of classification. So what?!
Well eventually I found out about logistic regression and I actually built my first usable model. Win!
Here's what happened next.
Well eventually I found out about logistic regression and I actually built my first usable model. Win!
Here's what happened next.
I tried to improve my model with Deep Learning. HAHAHAHAH. Big mistake.
Oh god this is funny.
So I began researching deep learning BECAUSE I saw someone tweet about TensorFlow and Keras on twitter.
Oh god this is funny.
So I began researching deep learning BECAUSE I saw someone tweet about TensorFlow and Keras on twitter.
And then I saw another person talk about Torch on LinkedIn.
And I heard them saying, "this is so crazy, I can predict images."
What were they doing?
Predicting cats and dogs.
And I heard them saying, "this is so crazy, I can predict images."
What were they doing?
Predicting cats and dogs.
But like a noobie, I said, "I gotta learn this."
You know, if they can predict CATS AND DOGS, then this deep learning stuff has to be good, right?!
WRONG. ❌
You know, if they can predict CATS AND DOGS, then this deep learning stuff has to be good, right?!
WRONG. ❌
I spent 3-months researching tensorflow.
Then I spent another month learning Keras because tensorflow was a flipping nightmare.
Then I spent another month trying to build a classification model.
Then I spent another month learning Keras because tensorflow was a flipping nightmare.
Then I spent another month trying to build a classification model.
You know, trying to add dense layers and lstm layers, and all sorts of stuff that silly noobies try to do that MAKES no sense.
So anyways, after 6-months I finally got SOMETHING that worked.
And it was WORSE than my linear regression model.
So anyways, after 6-months I finally got SOMETHING that worked.
And it was WORSE than my linear regression model.
I'm not even talking about my logistic regression model, the one that was actually getting decent results.
I'm talking LINEAR REGRESSION that has no business being used for CLASSIFICATION.
So yeah, deep learning was worse than my linear regression classification model. 😡
I'm talking LINEAR REGRESSION that has no business being used for CLASSIFICATION.
So yeah, deep learning was worse than my linear regression classification model. 😡
But the good news is that Logistic Regression model made my company $15,000,000.
So at the end of the day I still got promoted. And a hefty 100% increase in salary.
My point is when you're learning, there's a 1000 things you CAN learn.
But what SHOULD you learn?
So at the end of the day I still got promoted. And a hefty 100% increase in salary.
My point is when you're learning, there's a 1000 things you CAN learn.
But what SHOULD you learn?
I'd like to help.
I put together a FREE 40-minute webinar that consolidates the 10 things that helped me the most in my journey.
👉 Watch Here: buff.ly
I put together a FREE 40-minute webinar that consolidates the 10 things that helped me the most in my journey.
👉 Watch Here: buff.ly
جاري تحميل الاقتراحات...