# Unsupervised pretraining with autoencoder and supervised training

**URL:** <https://community.konduit.ai/t/unsupervised-pretraining-with-autoencoder-and-supervised-training/685>\
**Category:** DL4J\
**Created:** [July 5, 2020, 1:54am UTC](https://community.konduit.ai/t/unsupervised-pretraining-with-autoencoder-and-supervised-training/685 "2020-07-05T01:54:48Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Rutzno](https://avatars.discourse-cdn.com/v4/letter/r/bbe5ce/32.png) [@Rutzno](https://community.konduit.ai/u/Rutzno)\
**Post date:** [July 5, 2020, 1:54am UTC](https://community.konduit.ai/t/unsupervised-pretraining-with-autoencoder-and-supervised-training/685/1 "2020-07-05T01:54:48Z")

</div>

How to do unsupervised pretraining in dl4j with autoencoder and then doing supervised training in order to classify the differentes labels ?

Thanks in advance !

---

<div class="post-metadata">

**Author:** ![treo](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/treo/32/47_2.png) [@treo](https://community.konduit.ai/u/treo)\
**Post date:** [July 8, 2020, 7:36am UTC](https://community.konduit.ai/t/unsupervised-pretraining-with-autoencoder-and-supervised-training/685/2 "2020-07-08T07:36:47Z")

</div>

Aside from calling `fit` to train on labeled data, you can also call `pretrain` to pretrain your model in an unsupervised fashion.

---

<div class="post-metadata">

**Author:** ![Rutzno](https://avatars.discourse-cdn.com/v4/letter/r/bbe5ce/32.png) [@Rutzno](https://community.konduit.ai/u/Rutzno)\
**Post date:** [July 8, 2020, 8:15am UTC](https://community.konduit.ai/t/unsupervised-pretraining-with-autoencoder-and-supervised-training/685/3 "2020-07-08T08:15:13Z")

</div>

That helps as well !  
But to answer to my asked question, I did **transferLearning** to take pretrained model in autoencoder(model.fit(input, input)) and apply dense neural network with activation function **softmax** in output layer and doing supervised training.