# Architecture of model in EmbeddingIntentClassifier

**URL:** <https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359>\
**Category:** Rasa Masterclass\
**Tags:** rasa-nlu-trainer\
**Created:** [February 5, 2020, 2:25am UTC](https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359 "2020-02-05T02:25:17Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![namlv1997](https://avatars.discourse-cdn.com/v4/letter/n/b4bc9f/32.png) [@namlv1997](https://forum.rasa.com/u/namlv1997)\
**Post date:** [February 5, 2020, 2:25am UTC](https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359/1 "2020-02-05T02:25:17Z")

</div>

Where can i find the NN architecture of model in EmbeddingIntentClassifier such as image or document? When i was reading code on github, i saw 3 inputs (a, b, all\_bs) and that make me confused.

---

<div class="post-metadata">

**Author:** ![JiteshGaikwad](https://avatars.discourse-cdn.com/v4/letter/j/77aa72/32.png) [@JiteshGaikwad](https://forum.rasa.com/u/JiteshGaikwad)\
**Post date:** [February 5, 2020, 2:57am UTC](https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359/2 "2020-02-05T02:57:24Z")

</div>

hey @namlv1997, Welcome to the community 🙂

You can read this, it may help you:

> **[Components](https://rasa.com/docs/rasa/components/)**
>
> An open source machine learning framework for automated text and voice-based conversations

---

<div class="post-metadata">

**Author:** ![namlv1997](https://avatars.discourse-cdn.com/v4/letter/n/b4bc9f/32.png) [@namlv1997](https://forum.rasa.com/u/namlv1997)\
**Post date:** [February 5, 2020, 3:08am UTC](https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359/3 "2020-02-05T03:08:48Z")

</div>

@JiteshGaikwadSo can you explain what is the different between model with hidden layer b and model without hidden layer b? Moreover, in description part, it said that “Supervised embeddings are trained by maximizing similarity between them. This algorithm is based on StarSpace”. But rasa just uses it to calculate the similarity. is it right?

---

<div class="post-metadata">

**Author:** ![JiteshGaikwad](https://avatars.discourse-cdn.com/v4/letter/j/77aa72/32.png) [@JiteshGaikwad](https://forum.rasa.com/u/JiteshGaikwad)\
**Post date:** [February 5, 2020, 3:12am UTC](https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359/4 "2020-02-05T03:12:01Z")

</div>

I think so @amn41 can explain this 🙂

---

<div class="post-metadata">

**Author:** ![amn41](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/amn41/32/20_2.png) [@amn41](https://forum.rasa.com/u/amn41)\
**Post date:** [February 6, 2020, 10:50am UTC](https://forum.rasa.com/t/architecture-of-model-in-embeddingintentclassifier/24359/5 "2020-02-06T10:50:58Z")

</div>

hi - we don’t have it written up anywhere, so looking at the code directly is the only way to get a precise answer. But essentially we have a few fully connected layers between the input and the embedding space where we measure similarity.