Computes generalized degrees of freedom as originally proposed in [1].
Here, we use a Monte Carlo approximation, as described in [2]
Note that the model is retrained multiple times, which can take time.
Also, the model state changes due to this process.
[1] Ye, Jianming. “On Measuring and Correcting the Effects of Data Mining and Model Selection.” Journal of the American Statistical Association, vol. 93, no. 441, 1998, pp. 120–31, https://doi.org/10.2307/2669609. Accessed 3 May 2022.
[2] Gao, Tianxiang, Vladimir Jojic. ‘Degrees of Freedom in Deep Neural Networks’. arXiv preprint arXiv: Arxiv-1603. 09260 (2016): n. pag. Print.
Computes generalized degrees of freedom as originally proposed in [1]. Here, we use a Monte Carlo approximation, as described in [2]
Note that the model is retrained multiple times, which can take time. Also, the model state changes due to this process.
[1] Ye, Jianming. “On Measuring and Correcting the Effects of Data Mining and Model Selection.” Journal of the American Statistical Association, vol. 93, no. 441, 1998, pp. 120–31, https://doi.org/10.2307/2669609. Accessed 3 May 2022. [2] Gao, Tianxiang, Vladimir Jojic. ‘Degrees of Freedom in Deep Neural Networks’. arXiv preprint arXiv: Arxiv-1603. 09260 (2016): n. pag. Print.