neuralogic.nn.functional

sigmoid(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

tanh(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

signum(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

relu(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

leaky_relu(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

lukasiewicz(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

exp(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

sqrt(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

inverse(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

reverse(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

log(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

identity(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

transp(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

softmax(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

sparsemax(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

norm(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

slice(entity: BaseRelation | Function = None, *, rows: ellipsis | Tuple[int, int] = Ellipsis, cols: ellipsis | Tuple[int, int] = Ellipsis) → BaseRelation | Function[source]

Slices a value into a new value that is created by taking values on specified rows and columns.

Rows and Cols coordinates are specified either as ..., which means all rows/cols or by a tuple of two elements [from_index, to_index].

Parameters:

entity (Union[BaseRelation, Function]) – Relation to apply the function on. Default: None
rows (Union[type(Ellipsis), Tuple[int, int]]) – Default: ...
cols (Union[type(Ellipsis), Tuple[int, int]]) – Default: ...

reshape(entity: BaseRelation | Function = None, *, shape: None | Tuple[int, int] | int) → BaseRelation | Function[source]

Change the shape/type of the value to a new shape. The shape can be either None, int, or a tuple of two ints.

If None, the underlying value will be converted to a scalar. E.g., a matrix value of one element [[1]] will be converted to scalar 1.
If int, then the value will be converted to scalar (if the int is 0) or to a column vector.
If a tuple of two ints, the value will be converted to a scalar if the tuple is (0, 0). Into a row vector if the shape is (len, 0) or a column vector for shape (0, len). For other tuples (n, m), the value will be reshaped to matrix \(n \times m\).

Parameters:

entity (Union[BaseRelation, Function]) – Relation to apply the function on. Default: None
shape (Union[None, Tuple[int, int], int]) – The new shape of the value

count_comb(entity: BaseRelation = None) → BaseRelation | Function[source]

product_comb(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

elproduct_comb(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

softmax_comb(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

sparsemax_comb(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

crosssum_comb(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

concat_comb(entity: BaseRelation | Function = None, *, axis: int = -1) → BaseRelation | Function[source]

cossim_comb(entity: BaseRelation | Function = None) → BaseRelation | Function[source]

concat(entity: BaseRelation = None, *, axis: int = -1) → BaseRelation | Function[source]

softmax_agg(entity: BaseRelation = None, *, agg_terms: Sequence[str] = None) → BaseRelation | Function[source]