deep-learning-for-indentation
deep-learning-for-indentation copied to clipboard
Code issues of the Transfer learning
Thanks for sharing this novel algorithm to identify the material properties.
I still have a few questions about transfer learning:
-
For the function “validation_exp_cross_transfer”, it seems that only the 2D and 3D FEM dataset are involved in the Pre-training. Can we consider the experimental dataset to be involved in pre-training and saved as a model file?
-
If I want to perform a 2DFEM+3DFEM+EXP training network as in Figure 4 and apply it to the identification of properties of other unknown materials, how can this be achieved?
-
What is the difference between the function “validation_exp_cross2”and function “validation_exp_cross3”? It looks like the input and output of these two functions are the same.
Thanks.
- Yes. If you have those data, you can also train on them.
- You can use this function https://github.com/lululxvi/deep-learning-for-indentation/blob/42033c345831c7976d9e9d87c54395906d083487/src/nn.py#L217 Basically, the high-fideliyt is 3D FEM + Exp.
- They are pretty similar. The cross validation data is slightly different.
Thanks a lot for your reply.
-
It seems that the function "validation_exp_cross3" is designed for predict the "E, sigmay, sigma033" of "expdata2" by using the network built with the "FEMdata", "BerkovichData" and "dataexp1". While, the "validation_exp_cross" can only perform a 2DFEM+3DFEM training network and predict the expdata.
-
For the function "validation_exp_cross3" in nn.py, I don't understand the purpose of setting the the 10 times of interation. In additon, the result data is presented in 10 columns. I am not sure which column can be the solution with highest reliability.
Because neural network algorithm has randomness.