Trjtdtk

Can each chord in a progression create its own key?

How to politely respond to generic emails requesting a PhD/job in my lab? Without wasting too much time

What information about me do stores get via my credit card?

Word for: a synonym with a positive connotation?

Using `min_active_rowversion` for global temporary tables

Nested ellipses in tikzpicture: Chomsky hierarchy

Word to describe a time interval

Why can't devices on different VLANs, but on the same subnet, communicate?

Single author papers against my advisor's will?

Windows 10: How to Lock (not sleep) laptop on lid close?

Am I ethically obligated to go into work on an off day if the reason is sudden?

How to read αἱμύλιος or when to aspirate

Categorical vs continuous feature selection/engineering

Button changing its text & action. Good or terrible?

How to substitute curly brackets with round brackets in a grid of list

Didn't get enough time to take a Coding Test - what to do now?

What can I do if neighbor is blocking my solar panels intentionally?

What happens to a Warlock's expended Spell Slots when they gain a Level?

How to handle characters who are more educated than the author?

Is every episode of "Where are my Pants?" identical?

Is an up-to-date browser secure on an out-of-date OS?

Why doesn't shell automatically fix "useless use of cat"?

Are spiders unable to hurt humans, especially very small spiders?

Pretty sure I'm over complicating my loops but unsure how to simplify

intercept correction in deep learning

2

1

$begingroup$

Say I have an imbalanced data set, and I decided to over/undersample it during model training. I would then like to predict on new records but using the original, true imbalance in the classes as an apriori for the model.

i.e my classes are distributed in a ratio of 1:100, I undersample to 1:2, but would like the model to know that class A is very rare so be careful in predicting it.

For logistic regression I am familiar with a method of intercept correction, detailed here:

My question is: what if I use deep leaning instead of logistic regression?
My reasoning is that since the last layer of the NN is basically the same a logistic regression, I can use the same method for the intercept of this layer.
Can you think of any objections? any reason this doesn't make sense? If so, how would you go about correcting the model?

Also, is there a generalized method for intercept correction for a case when I have more than 2 classes?

deep-learning class-imbalance

share|improve this question

edited Apr 1 at 19:12

Ethan

701625

asked Apr 1 at 17:08

Oren MatarOren Matar

111

$endgroup$

• 
  
  
  

  
  

  
  
  
  
  
  

  
  $begingroup$
  Welcome to this site! I agree with you, since the proposed correction does not make any assumption about the feature space $boldsymbolx$, therefore, all layers from input $boldsymbolx$ to the last layer, act as a transformation that produces a new feature space $boldsymbolx'$, and the rest is the same.
  $endgroup$
  – Esmailian
  Apr 1 at 17:21
  
  

  
  
  
  

add a comment | 

2

1

$begingroup$

Say I have an imbalanced data set, and I decided to over/undersample it during model training. I would then like to predict on new records but using the original, true imbalance in the classes as an apriori for the model.

i.e my classes are distributed in a ratio of 1:100, I undersample to 1:2, but would like the model to know that class A is very rare so be careful in predicting it.

For logistic regression I am familiar with a method of intercept correction, detailed here:

My question is: what if I use deep leaning instead of logistic regression?
My reasoning is that since the last layer of the NN is basically the same a logistic regression, I can use the same method for the intercept of this layer.
Can you think of any objections? any reason this doesn't make sense? If so, how would you go about correcting the model?

Also, is there a generalized method for intercept correction for a case when I have more than 2 classes?

deep-learning class-imbalance

share|improve this question

edited Apr 1 at 19:12

Ethan

701625

asked Apr 1 at 17:08

Oren MatarOren Matar

111

$endgroup$

• 
  
  
  

  
  

  
  
  
  
  
  

  
  $begingroup$
  Welcome to this site! I agree with you, since the proposed correction does not make any assumption about the feature space $boldsymbolx$, therefore, all layers from input $boldsymbolx$ to the last layer, act as a transformation that produces a new feature space $boldsymbolx'$, and the rest is the same.
  $endgroup$
  – Esmailian
  Apr 1 at 17:21
  
  

  
  
  
  

add a comment | 

$begingroup$

Say I have an imbalanced data set, and I decided to over/undersample it during model training. I would then like to predict on new records but using the original, true imbalance in the classes as an apriori for the model.

i.e my classes are distributed in a ratio of 1:100, I undersample to 1:2, but would like the model to know that class A is very rare so be careful in predicting it.

For logistic regression I am familiar with a method of intercept correction, detailed here:

My question is: what if I use deep leaning instead of logistic regression?
My reasoning is that since the last layer of the NN is basically the same a logistic regression, I can use the same method for the intercept of this layer.
Can you think of any objections? any reason this doesn't make sense? If so, how would you go about correcting the model?

Also, is there a generalized method for intercept correction for a case when I have more than 2 classes?

deep-learning class-imbalance

share|improve this question

edited Apr 1 at 19:12

Ethan

701625

asked Apr 1 at 17:08

Oren MatarOren Matar

111

$endgroup$

share|improve this question

edited Apr 1 at 19:12

Ethan

701625

asked Apr 1 at 17:08

Oren MatarOren Matar

111

share|improve this question

edited Apr 1 at 19:12

Ethan

701625

asked Apr 1 at 17:08

Oren MatarOren Matar

111

edited Apr 1 at 19:12

Ethan

701625

edited Apr 1 at 19:12

Ethan

701625

asked Apr 1 at 17:08

Oren MatarOren Matar

111

asked Apr 1 at 17:08

Oren MatarOren Matar

111