i have a column/placeholder x with shape [None, 5]
and i want to obtain the mean, which i'll have to sum over all the vectors, [1, 5]
and i also want to obtain the mean of outer product, which i'll have to sum over all the outer product for each vector $ \sum_{i}{( x_i \otimes x_i) } $ which will have shape [5, 5]
it seems right now the reduce_block have some constraints on one input -> one output exactly mapping, is there any reason ?
i have a column/placeholder x with shape [None, 5]
and i want to obtain the mean, which i'll have to sum over all the vectors, [1, 5]
and i also want to obtain the mean of outer product, which i'll have to sum over all the outer product for each vector $ \sum_{i}{( x_i \otimes x_i) } $ which will have shape [5, 5]
it seems right now the reduce_block have some constraints on one input -> one output exactly mapping, is there any reason ?