Neural network with small data set

Question

Lukas Schmitt 2021-6-26

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/866005-neural-network-with-small-data-set

评论： Lukas Schmitt 2021-7-18

Hey there,

first of all, I am just a beginner starting with ANN's.

I am trying to make predictions with a simple Neural Network out of very small data sets. For example I got 3 inputs at a sample size of 24 to make predictions for 1 output.

Now my questions: Do you think it is feasible to get adequate results out of such a small sample size? What can I do to get more accuracy?

What can you recommend regarding the amount of layers and neurons?

Thanks for your help!

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Siddharth Solanki 2021-7-13

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/866005-neural-network-with-small-data-set#answer_745503

编辑：Siddharth Solanki 2021-7-13

As per my understanding you are trying to achieve better accuracy with a very small dataset and would like to know ways of improving the accuracy and some suggestions on architecture.

You can try to augment your dataset to increase the effective size of your dataset. Refer to this link for more details on image input. If you have audio input dataset then you can refer to this link.

You can learn more about basics of neural net architecture from this link.

You can also try transfer learning with pretrained nets which might suit your purpose if you have an image dataset - link.

If the dataset allows for using a ML algorithm for solving the problem, then it might be a good way to go about it for a small dataset. You may refer to this link for basics.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Lukas Schmitt 2021-7-18

Thanks a lot for your answer.

My problem is just a basic regression problem. My inputs are some integer values (properties from defects in metallic specimens, like defect size, position and so on). Now I am not sure how to separate this small amount of data into training and validation sets and what validation method I should use.

Furthermore, I was thinking about data augmentation too, but I don't know how I should do this.

请先登录，再进行评论。