# Disadvantages of Mean Squared Error?

I'm using mean squared error as reconstruction error for my autoencoder. The dataset is ECG (time series) and model is conv1d.

I assumed MSE as the best option for reconstruction error, but it's failing sometimes. What can be the plausible reasons for it not working.

In general when does MSE doesn't works or what are the disadvantages of MSE.

One basic disadvantage with Mean Squared Error is related to basic statistical concept which is Variance. Just like in Variance or "Mean" used in Variance, is prone to outliers. MSE is also prone to outliers as it uses the same concept of using mean in computing each error value.

Population Variance Formula:

We know there can be few outliers in our population data and hence when we compute

• summation of (X - mean)**2

When we calculate our mean there is a high chance it being a very large value due to consideration of one large outlier. This can produce high errors.

MSE formula :

Now, you can see the similarity in both formulas and understand that the same "mean" is used and both the formulas and are prone to Outliers.

Solution :