Mistake in PermZeroDBeta

The loss doesn't match the Latex formula. I suggest the following corrected implementation

```python
def f(self, X):
        d = X.shape[0]
        j = np.arange(1, d + 1)
        res = np.sum(
            [
                np.sum((j + self.beta) * (X**i - (1 / j) ** i)) ** 2
                for i in range(1, d + 1)
            ]
        )
        return res
```