## Vertices

There are several families of vertices. Each family shares a common value type.

### Probabilistic

Probabilistic vertices are vertices that do not depend completely on their parent vertices. An example of this is a GaussianVertex which is non-deterministic and has a probability distribution controlled by parameters that come from its parents.

Changing the value of their parent vertices may change the probability density function but it will not directly cause a change of the value of the vertex.

### Non-Probabilistic

The value of these vertices are completely dependent on their parent verticesâ€™ values. For example,
given `C = A * B`

(for any vertices A and B), C is a non-probabilistic vertex. Even if A or B are probabilistic
vertices, C is still completely dependent on their values which means it is non-probabilistic.

### The Double Family

A DoubleVertex can be used by most arithmetic operators. They can be used to describe a problem that can be solved using gradient ascent optimization and their value is a double.

The currently available double vertices are

### The Integer Family

An IntegerVertex is similar to the DoubleVertex except its value is an integer.

The currently available integer vertices are

### The Boolean (true/false) Family

A BoolVertex can be used by most boolean operators. These can be observed directly and used in MCMC.

The currently available boolean vertices are

### The Generic (everything else) family

These are the vertices that can have any type as a value. For example, this type can be an Enum or any user defined object.
Letâ€™s look at an example of this in Keanu with the `CategoricalVertex`

which will return a value of the specified Enum `MyType`

.

```
public enum MyType {
A, B, C, D
}
public CategoricalVertex<MyType> getSelectorForMyType() {
LinkedHashMap<MyType, DoubleVertex> frequency = new LinkedHashMap<>();
frequency.put(A, new ConstantDoubleVertex(0.25));
frequency.put(B, new ConstantDoubleVertex(0.25));
frequency.put(C, new ConstantDoubleVertex(0.25));
frequency.put(D, new ConstantDoubleVertex(0.25));
return new CategoricalVertex<>(frequency);
}
```

The getSelectorForMyType() method returns a probabilistic vertex that would contain an object of type MyType A, B, C or D, 25% of the time respectively.

The currently available generic vertices are

### Tensors

Vertices also have a `shape`

, which describes the tensor shape contained within them. A vertex with shape
[2,2] represents a 2 by 2 matrix. A vertex of shape [1,3] represents a row vector of length 3. The shape
can have any number of dimensions and any length.

Read more about tensors here