pdfs: initial work on random vector representation

will be later used for more powerful ProdCPdf

doc/pdfs.rst

@@ -5,6 +5,13 @@ Probability Density Functions
 .. automodule:: pybayes.pdfs
    :no-members:
 
+Random Variables and their Components
+=====================================
+
+.. autoclass:: RV
+
+.. autoclass:: RVComp
+
 Probability Density Function prototype
 ======================================
 

pybayes/__init__.py

@@ -7,5 +7,5 @@
 Bayesian statistics... TODO
 """
 
-from pdfs import CPdf, Pdf, UniPdf, GaussPdf, ProdPdf, MLinGaussCPdf, ProdCPdf
+from pdfs import RVComp, RV, CPdf, Pdf, UniPdf, GaussPdf, ProdPdf, MLinGaussCPdf, ProdCPdf
 from kalman import Kalman

pybayes/pdfs.pxd

@@ -8,6 +8,19 @@ cimport cython
 from numpywrap cimport *
 
 
+cdef class RVComp(object):
+    cdef readonly int dimension
+    cdef readonly str name
+
+
+cdef class RV(object):
+    cdef readonly int dimension
+    cdef readonly str name
+    cdef readonly list components
+
+    cpdef bint contains(self, RVComp component)
+
+
 cdef class CPdf(object):
     cpdef int shape(self) except -1
     cpdef int cond_shape(self) except -1

pybayes/pdfs.py

@@ -16,6 +16,57 @@
 from numpywrap import *
 
 
+class RVComp(object):
+    """Atomic component of a random value"""
+
+    def __init__(self, name, dimension):
+        """Initialise new component of a random variable. It will be named
+        `name` with dimension `dimension`"""
+
+        self.name = str(name)
+        if not isinstance(dimension, int):
+            raise TypeError("dimension must be integer (int)")
+        if dimension < 1:
+            raise ValueError("dimension must be non-zero positive")
+        self.dimension = dimension
+
+
+class RV(object):
+    """Representation of a random variable made of one or more components
+    (see RVComp)"""
+
+    def __init__(self, *components):
+        """Initialise new random variable made of one or more components. You may
+        also pass one or more existing RVs, whose components will be reused"""
+        self.dimension = 0
+        self.name = '['
+        self.components = []
+        for component in components:
+            if isinstance(component, RVComp):
+                self._add_component(component)
+            elif isinstance(component, RV):
+                for subcomp in component.components:
+                    self._add_component(subcomp)
+            else:
+                raise TypeError('component ' + component + ' is neither an instance '
+                              + 'of RVComp or RV')
+        self.name = self.name[:-2] + ']'
+
+    def _add_component(self, component):
+        # TODO: check if component is already contained? (does it matter somewhere?)
+        self.components.append(component)
+        self.dimension += component.dimension
+        self.name += component.name + ", "
+
+    def contains(self, component):
+        """Return True is this random value contains the exact same instance of
+        the `component`"""
+        for comp in self.components:
+            if id(comp) == id(component):
+                return True
+        return False
+
+
 class CPdf(object):
     """Base class for all Conditional Probability Density Functions
 

pybayes/tests/test_pdfs.py

@@ -12,6 +12,36 @@
 from support import PbTestCase
 
 
+class TestRVComp(PbTestCase):
+    """Test random variable component"""
+
+    def test_init(self):
+        rvcomp = pb.RVComp("pretty name", 123)
+        self.assertEquals(rvcomp.name, "pretty name")
+        self.assertEquals(rvcomp.dimension, 123)
+
+    def test_invalid_init(self):
+        self.assertRaises(TypeError, pb.RVComp, "def", 0.45)
+        self.assertRaises(TypeError, pb.RVComp, "def", "not a number")
+        self.assertRaises(ValueError, pb.RVComp, "abc", -1)
+
+
+class TestRV(PbTestCase):
+    """Test random variable representation"""
+
+    def test_init(self):
+        comp_a = pb.RVComp("a", 1)
+        comp_b = pb.RVComp("b", 2)
+        rv_1 = pb.RV(comp_a, comp_b)
+        self.assertEquals(rv_1.name, "[a, b]")
+        self.assertEquals(rv_1.dimension, 3)
+        self.assertTrue(rv_1.contains(comp_a))
+        self.assertTrue(rv_1.contains(comp_b))
+        rv_2 = pb.RV(rv_1)
+        self.assertEquals(rv_2.name, "[a, b]")
+        self.assertEquals(rv_2.dimension, 3)
+
+
 class TestCpdf(PbTestCase):
     """Test abstract class CPdf"""