add tf.nn.elu activation function. #294

Author: Oceania2018

src/TensorFlowNET.Core/APIs/tf.nn.cs

@@ -97,6 +97,9 @@ public static (Tensor, Tensor) dynamic_rnn(RNNCell cell, Tensor inputs, TF_DataT
                 throw new NotImplementedException("");
             }
 
+            public static Tensor elu(Tensor features, string name = null)
+                => gen_nn_ops.elu(features, name: name);
+
             public static (Tensor, Tensor) moments(Tensor x,
                 int[] axes,
                 string name = null,

src/TensorFlowNET.Core/Operations/NnOps/gen_nn_ops.cs

@@ -139,6 +139,27 @@ public static Tensor bias_add_grad(Tensor out_backprop,
             });
 
             return _op.outputs[0];
+        }
+
+        /// <summary>
+        /// Computes exponential linear: <c>exp(features) - 1</c> if &amp;lt; 0, <c>features</c> otherwise.
+        /// </summary>
+        /// <param name="features">
+        /// </param>
+        /// <param name="name">
+        /// If specified, the created operation in the graph will be this one, otherwise it will be named 'Elu'.
+        /// </param>
+        /// <returns>
+        ///    The Operation can be fetched from the resulting Tensor, by fetching the Operation property from the result.
+        /// </returns>
+        /// <remarks>
+        ///    See [Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
+        ///    ](http://arxiv.org/abs/1511.07289)
+        /// </remarks>
+        public static Tensor elu(Tensor features, string name = "Elu")
+        {
+            var op = _op_def_lib._apply_op_helper("Elu", name: name, args: new { features });
+            return op.output;
         }
 
         public static Tensor[] _fused_batch_norm(Tensor x,

test/TensorFlowNET.Examples/ImageProcessing/RetrainImageClassifier.cs

@@ -199,7 +199,7 @@ public bool Run()
                 RefVariable layer_biases = null;
                 with(tf.name_scope("biases"), delegate
                 {
-                    layer_biases = tf.Variable(tf.zeros((class_count)), name: "final_biases");
+                    layer_biases = tf.Variable(tf.zeros(class_count), name: "final_biases");
                     variable_summaries(layer_biases);
                 });