# Training accuracy (loss) increases (decreases) in a zigzag way?

**URL:** <https://discuss.pytorch.org/t/training-accuracy-loss-increases-decreases-in-a-zigzag-way/21112>\
**Category:** Uncategorized\
**Created:** [July 13, 2018, 3:05pm UTC](https://discuss.pytorch.org/t/training-accuracy-loss-increases-decreases-in-a-zigzag-way/21112 "2018-07-13T15:05:14Z")\
**Posts on this page:** 1\
**Showing post:** 11

<div class="post-metadata">

**Author:** ![kevinj22](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/k/8491ac/32.png) [@kevinj22](https://discuss.pytorch.org/u/kevinj22)\
**Post date:** [December 19, 2018, 1:54pm UTC](https://discuss.pytorch.org/t/training-accuracy-loss-increases-decreases-in-a-zigzag-way/21112/11 "2018-12-19T13:54:58Z")

</div>

There’s a similar post here [Strange behavior with SGD momentum training](https://discuss.pytorch.org/t/strange-behavior-with-sgd-momentum-training/7442/2) which also shows a saw toothed loss. A suggestion by smth is to do sampling with replacement.

---

_[View the full topic](https://discuss.pytorch.org/t/training-accuracy-loss-increases-decreases-in-a-zigzag-way/21112)._