Trjtdtk

How to recreate this effect in Photoshop?

Is a manifold-with-boundary with given interior and non-empty boundary essentially unique?

How to bypass password on Windows XP account?

What is this single-engine low-wing propeller plane?

When -s is used with third person singular. What's its use in this context?

Is it ethical to give a final exam after the professor has quit before teaching the remaining chapters of the course?

Does surprise arrest existing movement?

What do you call a phrase that's not an idiom yet?

Determinant is linear as a function of each of the rows of the matrix.

What are the pros and cons of Aerospike nosecones?

Is 1 ppb equal to 1 μg/kg?

Why is there no army of Iron-Mans in the MCU?

What is the longest distance a 13th-level monk can jump while attacking on the same turn?

Single word antonym of "flightless"

When to stop saving and start investing?

How does a Death Domain cleric's Touch of Death feature work with Touch-range spells delivered by familiars?

Bonus calculation: Am I making a mountain out of a molehill?

Proof involving the spectral radius and the Jordan canonical form

Can a non-EU citizen traveling with me come with me through the EU passport line?

Why is black pepper both grey and black?

Is the Standard Deduction better than Itemized when both are the same amount?

How can I make names more distinctive without making them longer?

What's the difference between `auto x = vector<int>()` and `vector<int> x`?

How can I fade player when goes inside or outside of the area?

Multi task learning with missing labels in Keras tutorial question

0

$begingroup$

https://www.dlology.com/blog/how-to-multi-task-learning-with-missing-labels-in-keras/

I followed this tutorial to create a multi task model for binary classification with missing labels in keras. The dataset that I'm using has inputs in the form of 167 length bit strings consisting of ones and zeroes (represent chemical structures), and the output is 12 binary labels representing toxicity of assays. I pretty much copied the tutorial exactly except for replacing their data with my curated dataset, and I'm getting 11% accuracy on my test set. I'm extremely confused about how the tutorial's masked_loss_function works, and I would appreciate any input on how to improve my model:

from keras import backend as K
def masked_loss_function(y_true, y_pred):
 mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
 return K.binary_crossentropy(y_true * mask, y_pred * mask)

model.compile(loss=masked_loss_function, optimizer='adam', metrics=['accuracy'])

Thanks so much!

classification keras beginner multitask-learning

share|improve this question

edited Apr 1 at 22:29

oW_♦

3,341933

asked Apr 1 at 22:18

scouts9scouts9

11

$endgroup$

• 
  
  
  

  
  

  
  
  
  
  
  

  
  $begingroup$
  Try applying KL divergence loss with similar and dissimilar pairs. ;)
  $endgroup$
  – William Scott
  Apr 2 at 0:40
  

  
  
  
  

add a comment | 

0

$begingroup$

https://www.dlology.com/blog/how-to-multi-task-learning-with-missing-labels-in-keras/

I followed this tutorial to create a multi task model for binary classification with missing labels in keras. The dataset that I'm using has inputs in the form of 167 length bit strings consisting of ones and zeroes (represent chemical structures), and the output is 12 binary labels representing toxicity of assays. I pretty much copied the tutorial exactly except for replacing their data with my curated dataset, and I'm getting 11% accuracy on my test set. I'm extremely confused about how the tutorial's masked_loss_function works, and I would appreciate any input on how to improve my model:

from keras import backend as K
def masked_loss_function(y_true, y_pred):
 mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
 return K.binary_crossentropy(y_true * mask, y_pred * mask)

model.compile(loss=masked_loss_function, optimizer='adam', metrics=['accuracy'])

Thanks so much!

classification keras beginner multitask-learning

share|improve this question

edited Apr 1 at 22:29

oW_♦

3,341933

asked Apr 1 at 22:18

scouts9scouts9

11

$endgroup$

• 
  
  
  

  
  

  
  
  
  
  
  

  
  $begingroup$
  Try applying KL divergence loss with similar and dissimilar pairs. ;)
  $endgroup$
  – William Scott
  Apr 2 at 0:40
  

  
  
  
  

add a comment | 

$begingroup$

https://www.dlology.com/blog/how-to-multi-task-learning-with-missing-labels-in-keras/

I followed this tutorial to create a multi task model for binary classification with missing labels in keras. The dataset that I'm using has inputs in the form of 167 length bit strings consisting of ones and zeroes (represent chemical structures), and the output is 12 binary labels representing toxicity of assays. I pretty much copied the tutorial exactly except for replacing their data with my curated dataset, and I'm getting 11% accuracy on my test set. I'm extremely confused about how the tutorial's masked_loss_function works, and I would appreciate any input on how to improve my model:

from keras import backend as K
def masked_loss_function(y_true, y_pred):
 mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
 return K.binary_crossentropy(y_true * mask, y_pred * mask)

model.compile(loss=masked_loss_function, optimizer='adam', metrics=['accuracy'])

Thanks so much!

classification keras beginner multitask-learning

share|improve this question

edited Apr 1 at 22:29

oW_♦

3,341933

asked Apr 1 at 22:18

scouts9scouts9

11

$endgroup$

from keras import backend as K
def masked_loss_function(y_true, y_pred):
 mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
 return K.binary_crossentropy(y_true * mask, y_pred * mask)

model.compile(loss=masked_loss_function, optimizer='adam', metrics=['accuracy'])

share|improve this question

edited Apr 1 at 22:29

oW_♦

3,341933

asked Apr 1 at 22:18

scouts9scouts9

11

share|improve this question

edited Apr 1 at 22:29

oW_♦

3,341933

asked Apr 1 at 22:18

scouts9scouts9

11

edited Apr 1 at 22:29

oW_♦

3,341933

edited Apr 1 at 22:29

oW_♦

3,341933

asked Apr 1 at 22:18

scouts9scouts9

11

asked Apr 1 at 22:18

scouts9scouts9

11