There is an error raised already in the case described which state: "NotImplementedError: Gradient is not implemented for CUQI Beta. Conditioning variables ['x']. with conditioning variables ['x']". However, computing the gradient for Beta likelihood used to (before this PR) fail for another reason: Beta._gradient() takes 2 positional arguments but 3 were given. I fixed the latter error by updating the gradient argument to be consistent with, e.g., Gaussian gradient. I also added a test. Now the first error is raised when attempting to compute the gradient for Beta likelihood.
closes #159
There is an error raised already in the case described which state: "NotImplementedError: Gradient is not implemented for CUQI Beta. Conditioning variables ['x']. with conditioning variables ['x']". However, computing the gradient for Beta likelihood used to (before this PR) fail for another reason: Beta._gradient() takes 2 positional arguments but 3 were given. I fixed the latter error by updating the gradient argument to be consistent with, e.g., Gaussian gradient. I also added a test. Now the first error is raised when attempting to compute the gradient for Beta likelihood.