dynaml.tensorflow: Added Dynamic System simulation and inference

 - FiniteHorizonCTRNN: A continuous time recurrent neural network layer.
 - Loss for time slices of multivariate time series.

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/tensorflow/layers/FiniteHorizonCTRNN.scala

@@ -0,0 +1,61 @@
+/*
+Licensed to the Apache Software Foundation (ASF) under one
+or more contributor license agreements.  See the NOTICE file
+distributed with this work for additional information
+regarding copyright ownership.  The ASF licenses this file
+to you under the Apache License, Version 2.0 (the
+"License"); you may not use this file except in compliance
+with the License.  You may obtain a copy of the License at
+
+  http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing,
+software distributed under the License is distributed on an
+"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+KIND, either express or implied.  See the License for the
+specific language governing permissions and limitations
+under the License.
+* */
+package io.github.mandar2812.dynaml.tensorflow.layers
+
+import org.platanios.tensorflow.api.{Output, Shape, tf}
+import org.platanios.tensorflow.api.learn.Mode
+import org.platanios.tensorflow.api.learn.layers.Layer
+import org.platanios.tensorflow.api.ops.variables.{Initializer, RandomNormalInitializer}
+
+/**
+  * Represents a Continuous Time Recurrent Neural Network (CTRNN)
+  * The layer simulates the discretized dynamics of the CTRNN for
+  * a fixed number of time steps.
+  *
+  * @author mandar2812 date: 2018/03/06
+  * */
+case class FiniteHorizonCTRNN(
+  override val name: String, units: Int,
+  horizon: Int, timestep: Double,
+  weightsInitializer: Initializer = RandomNormalInitializer(),
+  biasInitializer: Initializer = RandomNormalInitializer(),
+  gainInitializer: Initializer = RandomNormalInitializer(),
+  timeConstantInitializer: Initializer = RandomNormalInitializer()) extends
+  Layer[Output, Output](name) {
+
+  override val layerType: String = "FHCTRNN"
+
+  override protected def _forward(input: Output, mode: Mode): Output = {
+
+    val weights      = tf.variable("Weights", input.dataType, Shape(units, units), weightsInitializer)
+    val timeconstant = tf.variable("TimeConstant", input.dataType, Shape(units, units), timeConstantInitializer)
+    val gain         = tf.variable("Gain", input.dataType, Shape(units, units), timeConstantInitializer)
+    val bias         = tf.variable("Bias", input.dataType, Shape(units), biasInitializer)
+
+    tf.stack(
+      (1 to horizon).scanLeft(input)((x, _) => {
+        val decay = x.tensorDot(timeconstant.multiply(-1d), Seq(1), Seq(0))
+        val interaction = x.tensorDot(gain, Seq(1), Seq(0)).add(bias).tanh.tensorDot(weights, Seq(1), Seq(0))
+
+        x.add(decay.multiply(timestep)).add(interaction.multiply(timestep))
+      }).tail,
+      axis = -1)
+
+  }
+}

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/tensorflow/learn/MVTimeSeriesLoss.scala

@@ -0,0 +1,37 @@
+/*
+Licensed to the Apache Software Foundation (ASF) under one
+or more contributor license agreements.  See the NOTICE file
+distributed with this work for additional information
+regarding copyright ownership.  The ASF licenses this file
+to you under the Apache License, Version 2.0 (the
+"License"); you may not use this file except in compliance
+with the License.  You may obtain a copy of the License at
+
+  http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing,
+software distributed under the License is distributed on an
+"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+KIND, either express or implied.  See the License for the
+specific language governing permissions and limitations
+under the License.
+* */
+package io.github.mandar2812.dynaml.tensorflow.learn
+
+import org.platanios.tensorflow.api.learn.Mode
+import org.platanios.tensorflow.api.learn.layers.Loss
+import org.platanios.tensorflow.api.ops.Output
+
+/**
+  * L2 loss for a time slice of a multivariate time series
+  *
+  * @author mandar2812 date 9/03/2018
+  * */
+case class MVTimeSeriesLoss(override val name: String)
+  extends Loss[(Output, Output)](name) {
+  override val layerType: String = "L2Loss"
+
+  override protected def _forward(input: (Output, Output), mode: Mode): Output = {
+    input._1.subtract(input._2).square.mean(axes = 0).sum()
+  }
+}

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/tensorflow/utils/GenRegressionMetricsTF.scala

@@ -0,0 +1,99 @@
+/*
+Licensed to the Apache Software Foundation (ASF) under one
+or more contributor license agreements.  See the NOTICE file
+distributed with this work for additional information
+regarding copyright ownership.  The ASF licenses this file
+to you under the Apache License, Version 2.0 (the
+"License"); you may not use this file except in compliance
+with the License.  You may obtain a copy of the License at
+
+  http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing,
+software distributed under the License is distributed on an
+"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+KIND, either express or implied.  See the License for the
+specific language governing permissions and limitations
+under the License.
+* */
+package io.github.mandar2812.dynaml.tensorflow.utils
+
+import com.quantifind.charts.Highcharts.{regression, title, xAxis, yAxis}
+import io.github.mandar2812.dynaml.evaluation.RegressionMetricsTF
+import io.github.mandar2812.dynaml.tensorflow.dtf
+import org.platanios.tensorflow.api._
+
+/**
+  * Generalisation of [[RegressionMetricsTF]] to more complex output structures, like
+  * matrix outputs/labels.
+  *
+  * @author mandar2812 date: 9/03/2018
+  * */
+class GenRegressionMetricsTF(preds: Tensor, targets: Tensor) extends RegressionMetricsTF(preds, targets) {
+  private val num_outputs =
+    if (preds.shape.toTensor().size == 1) 1
+    else preds.shape.toTensor()(0 :: -1).prod().scalar.asInstanceOf[Int]
+
+  private lazy val (_ , rmse , mae, corr) = GenRegressionMetricsTF.calculate(preds, targets)
+
+  private lazy val modelyield =
+    (preds.max(axes = 0) - preds.min(axes = 0)).divide(targets.max(axes = 0) - targets.min(axes = 0))
+
+  override protected def run(): Tensor = dtf.stack(Seq(rmse, mae, corr, modelyield), axis = -1)
+
+  override def generatePlots(): Unit = {
+    println("Generating Plot of Fit for each target")
+
+    if(num_outputs == 1) {
+      val (pr, tar) = (
+        scoresAndLabels._1.entriesIterator.map(_.asInstanceOf[Float]),
+        scoresAndLabels._2.entriesIterator.map(_.asInstanceOf[Float]))
+
+      regression(pr.zip(tar).toSeq)
+
+      title("Goodness of fit: "+name)
+      xAxis("Predicted "+name)
+      yAxis("Actual "+name)
+
+    } else {
+      (0 until num_outputs).foreach(output => {
+        val (pr, tar) = (
+          scoresAndLabels._1(::, output).entriesIterator.map(_.asInstanceOf[Float]),
+          scoresAndLabels._2(::, output).entriesIterator.map(_.asInstanceOf[Float]))
+
+        regression(pr.zip(tar).toSeq)
+      })
+    }
+  }
+}
+
+object GenRegressionMetricsTF {
+
+  protected def calculate(preds: Tensor, targets: Tensor): (Tensor, Tensor, Tensor, Tensor) = {
+    val error = targets.subtract(preds)
+
+    println("Shape of error tensor: "+error.shape.toString()+"\n")
+
+    val num_instances = error.shape(0)
+    val rmse = error.square.mean(axes = 0).sqrt
+
+    val mae = error.abs.mean(axes = 0)
+
+    val corr = {
+
+      val mean_preds = preds.mean(axes = 0)
+
+      val mean_targets = targets.mean(axes = 0)
+
+      val preds_c = preds.subtract(mean_preds)
+
+      val targets_c = targets.subtract(mean_targets)
+
+      val (sigma_t, sigma_p) = (targets_c.square.mean(axes = 0).sqrt, preds_c.square.mean(axes = 0).sqrt)
+
+      preds_c.multiply(targets_c).mean(axes = 0).divide(sigma_t.multiply(sigma_p))
+    }
+
+    (error, rmse, mae, corr)
+  }
+}