Trjtdtk

It is correct to match light sources with the same color temperature?

Which one is the true statement?

Won the lottery - how do I keep the money?

Is it ok to trim down a tube patch?

What flight has the highest ratio of timezone difference to flight time?

Is fine stranded wire ok for main supply line?

Scary film where a woman has vaginal teeth

Towers in the ocean; How deep can they be built?

Why is the US ranked as #45 in Press Freedom ratings, despite its extremely permissive free speech laws?

Is dried pee considered dirt?

Traveling with my 5 year old daughter (as the father) without the mother from Germany to Mexico

What is the difference between "hamstring tendon" and "common hamstring tendon"?

(How) Could a medieval fantasy world survive a magic-induced "nuclear winter"?

Reshaping json / reparing json inside shell script (remove trailing comma)

Does higher Oxidation/ reduction potential translate to higher energy storage in battery?

Can you teleport closer to a creature you are Frightened of?

Aggressive Under-Indexing and no data for missing index

Is there a way to save my career from absolute disaster?

Help understanding this unsettling image of Titan, Epimetheus, and Saturn's rings?

Players Circumventing the limitations of Wish

Expressing the idea of having a very busy time

Redefining symbol midway through a document

Is a distribution that is normal, but highly skewed, considered Gaussian?

How to find image of a complex function with given constraints?

Extract features (Topics) from lstm

0

$begingroup$

I have searched a lot this question but sadly could not find a hint how to do it.
I am using an LSTM autoencoder architecture modeled here. I would like to use this code to extract topics from it.
the code I have written brings me 600 features because the sequence being defined is 60 and I asked for 10 topics so it returns back 600 features.
However, I think these 600 features are only hidden state at each time.

So my question arises when we fit the model what is the reasonable thing to do to extract features from the middle layer?
(I have done this using autoencoder, it is pretty simple and understandable as we just to extract topic from the encoder part). but in LSTM model it seems different, where can I extract features?

What is the best approach while trying LSTM autoencoder to come up with topics?

deep-learning lstm feature-extraction autoencoder topic-model

share|improve this question

asked Mar 26 at 22:45

sariiisariii

18510

$endgroup$

add a comment | 

0

$begingroup$

I have searched a lot this question but sadly could not find a hint how to do it.
I am using an LSTM autoencoder architecture modeled here. I would like to use this code to extract topics from it.
the code I have written brings me 600 features because the sequence being defined is 60 and I asked for 10 topics so it returns back 600 features.
However, I think these 600 features are only hidden state at each time.

So my question arises when we fit the model what is the reasonable thing to do to extract features from the middle layer?
(I have done this using autoencoder, it is pretty simple and understandable as we just to extract topic from the encoder part). but in LSTM model it seems different, where can I extract features?

What is the best approach while trying LSTM autoencoder to come up with topics?

deep-learning lstm feature-extraction autoencoder topic-model

share|improve this question

asked Mar 26 at 22:45

sariiisariii

18510

$endgroup$

add a comment | 

$begingroup$

I have searched a lot this question but sadly could not find a hint how to do it.
I am using an LSTM autoencoder architecture modeled here. I would like to use this code to extract topics from it.
the code I have written brings me 600 features because the sequence being defined is 60 and I asked for 10 topics so it returns back 600 features.
However, I think these 600 features are only hidden state at each time.

So my question arises when we fit the model what is the reasonable thing to do to extract features from the middle layer?
(I have done this using autoencoder, it is pretty simple and understandable as we just to extract topic from the encoder part). but in LSTM model it seems different, where can I extract features?

What is the best approach while trying LSTM autoencoder to come up with topics?

deep-learning lstm feature-extraction autoencoder topic-model

share|improve this question

asked Mar 26 at 22:45

sariiisariii

18510

$endgroup$

share|improve this question

asked Mar 26 at 22:45

sariiisariii

18510

share|improve this question

asked Mar 26 at 22:45

sariiisariii

18510

asked Mar 26 at 22:45

sariiisariii

18510

asked Mar 26 at 22:45

sariiisariii

18510