Comments (6)
I don’t think the perplexity is a evaluation metric for text classification task but language model. If you want the value, you can run the pretrained model and calculate the perplexity. Note that don’t forget to change the total loss (xentropy+l2) to the cross entropy loss only when evaluation.
from disan.
@taoshen58 For perplexity part, it seems that a word embedding matrix should also be multiplied. Which embedding matrix should be used? emb_mat_token or emb_mat_glove
from disan.
Hi @xuy2
emb_mat_token
includes the words from the training set where the embedding is initialized by glove if the corresponding word exists in the glove, otherwise is initialized by random.
emb_mat_glove
is the words from glove where the overlapping words with emb_mat_token
have been removed. This mat is certainly initialized by glove.
So, emb_mat_token
need fine-tuned during training while the emb_mat_glove
is not trainable but useful when testing.
from disan.
@taoshen58 It seems that I'd better use the matrix after generate_embedding_mat function, which includes 406079 words. I'll update the perplexity if my test is finished. Have you ever consider that using the disan model as transfer learning?
from disan.
You can reduce the vocabulary size by using the word frequency in your corpus.
Is the transfer learning
you mentioned similar to the OpenAI Transformer?
from disan.
https://s3-us-west-2.amazonaws.com/openai-assets/research-covers/language-unsupervised/language_understanding_paper.pdf
Like this paper.
from disan.
Related Issues (20)
- word embedding HOT 4
- Using tree structure not raw text in SNLI dataset HOT 3
- experiments on MSRP HOT 3
- Word Embedding tune HOT 6
- tensorboard graph did not show anything HOT 1
- report error in line94 snli_main.py
- How to process new data? What should be the format of input file HOT 3
- what to do about the var rep_mask in the disan.py? HOT 10
- Can you take examples for the Fast-Disa.py?
- input to fast-disan.py
- SICK dataset
- Excuse me, where is the code for visual attention weight?
- SST data process HOT 1
- question about config parameter "only_sentence" HOT 3
- Where is the trained model stored?
- How can I generate predictions for my non-SNLI dataset using SNLI-DiSAN model? HOT 1
- It seems that without a pretrained embedding input, the results become worse HOT 2
- The net seems use lstm HOT 2
- About the structure of the code HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from disan.