Some researcher claimed that deep learning works perfectly well with small data. Really is it possible?
Razieh Rastgoo
I think the best way is to find out the most suitable or combined neural network for machine learning... using neural networks Like Kohonen's networks.
Dear Razeih
The main difference of deep learning from other Machine learning approaches is due to the amount of training data involved and the computational power.
https://towardsdatascience.com/why-go-large-with-data-for-deep-learning-12eee16f708
Thx for your kind reply. Yeah, but please visit the following link:
https://towardsdatascience.com/does-deep-learning-really-require-big-data-no-13890b014ded
Razieh Rastgoo
I think the best way is to find out the most suitable or combined neural network for machine learning... using neural networks Like Kohonen's networks.
Both of the model and data are important. Here, we want to know about the data. Generally, deep models are trained using a large amount of data. I also know that we can benefit from the powerful deep models but I really like to know that using these models ONLY with small data is efficient or not because data providing is so challenging!
Normally deep learning models needs larger datasets. If you dont have a larger data then you can use augmentation methods to increase your data size.
Generally speaking, it is not recommended to use deep learning if you have
< 100 samples. However, you can probably train the deep learning model in (some) small data settings, if you train your model carefully.
Yeah, it’s clear that it’s better to train the deep model using a large dataset but I am wondering how some researchers claim that they just use small data for deep models training! By the way, the following link is interesting:
https://www.kdnuggets.com/2015/03/small-data-specialized-deep-learning-yann-lecun.html
This article shows how you can train a deep learning model with small data-set.
https://beamandrew.github.io/deeplearning/2017/06/04/deep_learning_works.html
- Badges
- Science topic