I want the outcome like this : A: You can explode all values with this code: $result=explode(‘,’,$searchFor); Then each value of the row would be in $result[1],$result[2] etc. Then you can run a foreach loop over $result and implode (join) them, so the outcome is the same as you want: foreach($row as $row) { echo implode(‘ ‘,$row); } Q: Uniqueness of regularization parameter without loss of generality If I have a problem $f_\theta^* = \mathrm{argmin}_f L(f)+\lambda\Omega(f)$, where $\Omega$ is some regularizer, I usually need to solve this problem by first selecting a $\lambda_0$, then minimizing the following: $f_\lambda^* = \mathrm{argmin}_f \frac{1}{2}||y-f||^2 +\lambda \Omega(f)$ This minimization problem is computationally much more efficient because we don’t need to evaluate the regularization term for every sample in the training set. However, I want to ensure that the regularization term will remain relevant for my predictions. So when I write a function, I want to make sure that the parameters of the regularization $\Omega$ are tuned to optimize the regularized loss function. The question then becomes: What do I need to make sure is the regularization term $\Omega$ is not tuned to overfit the data? In other words, how do I know that the regularization term is actually fitting “something meaningful” to the data, rather than overfitting? Is there a way to bound the number of parameters of $\Omega$ so I can see that it is under-fitting? A: The regularization term can overfit the data and if this happens, then the resulting model could be unstable, because the regularization term can even change the sign of the gradient. The way to