SciSharp · Oceania2018 · Mar 4, 2019 · Mar 4, 2019
@@ -0,0 +1,33 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Framework
+{
+    public static class common_shapes
+    {
+        /// <summary>
+        /// Returns the broadcasted shape between `shape_x` and `shape_y
+        /// </summary>
+        /// <param name="shape_x"></param>
+        /// <param name="shape_y"></param>
+        public static Tensor broadcast_shape(Tensor shape_x, Tensor shape_y)
+        {
+            var return_dims = _broadcast_shape_helper(shape_x, shape_y);
+            // return tensor_shape(return_dims);
+            throw new NotFiniteNumberException();
+        }
+        /// <summary>
+        /// Helper functions for is_broadcast_compatible and broadcast_shape.
+        /// </summary>
+        /// <param name="shape_x"> A `TensorShape`</param>
+        /// <param name="shape_y"> A `TensorShape`</param>
+        /// <return> Returns None if the shapes are not broadcast compatible,
+        /// a list of the broadcast dimensions otherwise.
+        /// </return>
+        public static Tensor _broadcast_shape_helper(Tensor shape_x, Tensor shape_y)
+        {
+            throw new NotFiniteNumberException();
+        }
+    }
+}
@@ -0,0 +1,14 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+using Tensorflow;
+
+namespace Tensorflow.Operations.Distributions
+{
+    public enum DistributionEnum
+    {
+
+
+
+    }
+}
@@ -6,7 +6,7 @@
 
 namespace Tensorflow
 {
-    abstract class _BaseDistribution : Object
+    abstract class _BaseDistribution : Python
     {
         // Abstract base class needed for resolving subclass hierarchy.
     }
@@ -22,8 +22,8 @@ class Distribution : _BaseDistribution
         public ReparameterizationType _reparameterization_type {get;set;}
         public bool _validate_args {get;set;}
         public bool _allow_nan_stats {get;set;}
-        public Dictionary<object, object> _parameters  {get;set;}
-        public List<object> _graph_parents  {get;set;}
+        public Dictionary<string, object> _parameters  {get;set;}
+        public List<Tensor> _graph_parents  {get;set;}
         public string _name  {get;set;}
 
         /// <summary>
@@ -82,7 +82,21 @@ private Distribution (
     /// </summary>
     class ReparameterizationType
     {
+        public string _rep_type { get; set; }
+        public ReparameterizationType(string rep_type)
+        {
+            this._rep_type = rep_type;
+        }
 
+        public void repr()
+        {
+            Console.WriteLine($"<Reparameteriation Type: {this._rep_type}>" );
+        }
+
+        public bool eq (ReparameterizationType other)
+        {
+            return this.Equals(other);
+        }
     }
 
 

@@ -1,10 +1,80 @@
+using System.Collections.Generic;
+
 namespace Tensorflow
 {
     class Normal : Distribution
     {
-        public Normal (Tensor loc, Tensor scale, bool validate_args=false, bool allow_nan_stats=true, string name="Normal")
+        public Tensor _loc { get; set; }
+        public Tensor _scale { get; set; }
+
+        Dictionary<string, object> parameters = new Dictionary<string, object>();
+        /// <summary>
+        /// The Normal distribution with location `loc` and `scale` parameters.
+        /// Mathematical details
+        /// The probability density function(pdf) is,
+        /// '''
+        /// pdf(x; mu, sigma) = exp(-0.5 (x - mu)**2 / sigma**2) / Z
+        /// Z = (2 pi sigma**2)**0.5
+        /// '''
+        /// where `loc = mu` is the mean, `scale = sigma` is the std.deviation, and, `Z`
+        /// is the normalization constant.
+        /// </summary>
+        /// <param name="loc"></param>
+        /// <param name="scale"></param>
+        /// <param name="validate_args"></param>
+        /// <param name="allow_nan_stats"></param>
+        /// <param name="name"></param>
+        public Normal (Tensor loc, Tensor scale, bool validate_args=false, bool allow_nan_stats=true, string name="Normal") 
+        {
+            parameters.Add("name", name);
+            parameters.Add("loc", loc);
+            parameters.Add("scale", scale);
+            parameters.Add("validate_args", validate_args);
+            parameters.Add("allow_nan_stats", allow_nan_stats);
+
+            with(new ops.name_scope(name, "", new { loc, scale }), scope => 
+            {
+                with(ops.control_dependencies(validate_args ? new Operation[] { scale.op} : new Operation[] { }), cd =>
+                {
+                    this._loc = array_ops.identity(loc, name);
+                    this._scale = array_ops.identity(scale, name);
+                    base._dtype = this._scale.dtype;
+                    base._reparameterization_type = new ReparameterizationType("FULLY_REPARAMETERIZED");
+                    base._validate_args = validate_args;
+                    base._allow_nan_stats = allow_nan_stats;
+                    base._parameters = parameters;
+                    base._graph_parents = new List<Tensor>(new Tensor[] { this._loc, this._scale });
+                    base._name = name;
+                });
+
+            });
+
+        }
+        /// <summary>
+        /// Distribution parameter for the mean.
+        /// </summary>
+        /// <returns></returns>
+        public Tensor loc()
+        {
+            return this._loc;
+        }
+        /// <summary>
+        /// Distribution parameter for standard deviation."
+        /// </summary>
+        /// <returns></returns>
+        public Tensor scale()
+        {
+            return this._scale;
+        }
+
+        public Tensor _batch_shape_tensor()
+        {
+            return array_ops.broadcast_dynamic_shape(array_ops.shape(this._loc), array_ops.shape(this._scale));
+        }
+
+        public Tensor _batch_shape()
         {
-
+            return array_ops.broadcast_static_shape(new Tensor(this._loc.shape), new Tensor(this._scale.shape));
         }
 
     }

@@ -239,5 +239,34 @@ public static Tensor squeeze(Tensor input, int[] axis = null, string name = null
         {
             return gen_array_ops.squeeze(input, axis, name);
         }
+
+        public static Tensor identity(Tensor input, string name = null)
+        {
+            return gen_array_ops.identity(input, name);
+        }
+        /// <summary>
+        /// Computes the shape of a broadcast given symbolic shapes.
+        /// When shape_x and shape_y are Tensors representing shapes(i.e.the result of
+        /// calling tf.shape on another Tensor) this computes a Tensor which is the shape
+        /// of the result of a broadcasting op applied in tensors of shapes shape_x and
+        /// shape_y.
+        /// For example, if shape_x is [1, 2, 3] and shape_y is [5, 1, 3], the result is a
+        /// Tensor whose value is [5, 2, 3].
+        /// This is useful when validating the result of a broadcasting operation when the
+        /// tensors do not have statically known shapes.
+        /// </summary>
+        /// <param name="shape_x"> A rank 1 integer `Tensor`, representing the shape of x.</param>
+        /// <param name="shape_y"> A rank 1 integer `Tensor`, representing the shape of y.</param>
+        /// <returns> A rank 1 integer `Tensor` representing the broadcasted shape.</returns>
+        public static Tensor broadcast_dynamic_shape(Tensor shape_x, Tensor shape_y)
+        {
+            return gen_array_ops.broadcast_args(shape_x, shape_y);
+        }
+
+        public static Tensor broadcast_static_shape(Tensor shape_x, Tensor shape_y)
+        {
+            return Framework.common_shapes.broadcast_shape(shape_x, shape_y);
+        }
+
     }
 }

@@ -178,5 +178,21 @@ public static Tensor squeeze(Tensor input, int[] axis = null, string name = null
 
             return _op.outputs[0];
         }
+
+        /// <summary>
+        /// Return the shape of s0 op s1 with broadcast.
+        /// Given `s0` and `s1`, tensors that represent shapes, compute `r0`, the
+        /// broadcasted shape. `s0`, `s1` and `r0` are all integer vectors.
+        /// </summary>
+        /// <param name="s0"> A `Tensor`. Must be one of the following types: `int32`, `int64`.</param>
+        /// <param name="s1"> A `Tensor`. Must have the same type as `s0`.</param>
+        /// <param name="name"> A name for the operation (optional).</param>
+        /// <returns> `Tensor`. Has the same type as `s0`.</returns>
+        public static Tensor broadcast_args(Tensor s0, Tensor s1, string name = null)
+        {
+            var _op = _op_def_lib._apply_op_helper("BroadcastArgs", name, args: new { s0, s1, name });
+
+            return _op.outputs[0];
+        }
     }
 }