Gaussian fit with provided mean and variance

Hi guys, I'm a real newbie to Matlab, haven't used it in years so bear with me please

I'm studying a paper regarding CNN, apparently it uses MatLab to perform some kind of preprocessing on the input data and, if possible, I would like your help on the matter or at least a direction to follow as I have no idea even if the task is possible at all

Basically I'm working with the the AVA dataset which provides a huge set of aesthetic scores for around 250k images, for each image 10 values are provided which are the amount of votes it received for each aesthetic "bucket" 1,2,3,..,10 (how many users rated with score 1/10, 2/10, 3/10,...,10/10). The paper provides a way to "denoise" those values by fitting each set into a Gaussian curve and then to discretize the values at x=1,2,3,..,10. In particular, it doesn't just use 10 values to obtain the fit but it also provides as input the mean and variance: the first is the classic arithmetic one while the variance is the minimum between half the original value or 1.5 "A Gaussian distribution is thus fitted to the raw score distribution of each image using μ and σ^2_denoised". Note that the 10 values can be weighted using the bucket they refer to, thus mean and variance can be computed without problem

I usually program with Python and there we have the possibility to define the required curve fitting function, such as a Gaussian formula, and also to provide a suggested starting point for the function's parameters such as the mean and variance. Sadly, this solution only suggests those values and doesn't force thier use to the system, in fact I noticed that the fitted functions obtained with normal variance or its denoised version are basically the same. How does MatLab works in this regard? Given that the paper provides different CNN training results when using the fitted input data obtained with the two possible σ^2, I can't proceed with Python. Can you help me with this task?