Error - out of memory during training phase classification

5 次查看（过去 30 天）

Swapnil Sayan Saha 2019-6-6

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/465816-error-out-of-memory-during-training-phase-classification

评论： Swapnil Sayan Saha 2019-6-7

I am trying to use a large predictor matrix (196072 X 3042) for classification (using classification learner and other functions).

However, the training phase fails every time and the message points to running out of memory. Note that my computer has 16 GB of RAM and the predictor matrix is around 3.7 GB.

Is there a way to use tall arrays or any other out of memory processing options that I can use? Functions I will be using mainly include fitcensemble and fitcsvm.

3 个评论
显示 1更早的评论隐藏 1更早的评论

Swapnil Sayan Saha 2019-6-7

编辑：Swapnil Sayan Saha 2019-6-7

I did deduce the memory issue by looking at task manager. MATLAB at times eats away 13/14 GB of RAM. The PC does not respond during those few seconds. The "Error - Out of memory" message was also provided by MATLAB.

Could you provide tools for batch training?

The funny thing is I trained shallow neural networks using MATLAB Neural Network Pattern Recognition on the same PC using the same large predictor matrix later on without any issues.

Swapnil Sayan Saha 2019-6-7

I found out I could in fact use tall arrays with treebagger function. But it seems not all of the elements of the function are supported if I use tall arrays.

请先登录，再进行评论。

请先登录，再回答此问题。