artemmalykh commented on a change in pull request #5767: [ML] IGNITE-10573: 
Consistent API for Ensemble training
URL: https://github.com/apache/ignite/pull/5767#discussion_r247874720
 
 

 ##########
 File path: 
modules/ml/src/main/java/org/apache/ignite/ml/composition/combinators/sequential/TrainersSequentialComposition.java
 ##########
 @@ -0,0 +1,125 @@
+/*
+ * 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 org.apache.ignite.ml.composition.combinators.sequential;
+
+import org.apache.ignite.ml.IgniteModel;
+import org.apache.ignite.ml.composition.CompositionUtils;
+import org.apache.ignite.ml.composition.DatasetMapping;
+import org.apache.ignite.ml.dataset.DatasetBuilder;
+import org.apache.ignite.ml.math.functions.IgniteBiFunction;
+import org.apache.ignite.ml.math.functions.IgniteFunction;
+import org.apache.ignite.ml.math.primitives.vector.Vector;
+import org.apache.ignite.ml.math.primitives.vector.VectorUtils;
+import org.apache.ignite.ml.trainers.DatasetTrainer;
+import sun.reflect.generics.reflectiveObjects.NotImplementedException;
+
+/**
+ * Sequential composition of trainers.
+ * Sequential composition of trainers is itself trainer which produces {@link 
ModelsSequentialComposition}.
+ * Training is done in following fashion:
+ * <pre>
+ *     1. First trainer is trained and `mdl1` is produced.
+ *     2. From `mdl1` {@link DatasetMapping} is constructed. This mapping 
`dsM` encapsulates dependency between first
+ *     training result and second trainer.
+ *     3. Second trainer is trained using dataset aquired from application 
`dsM` to original dataset; `mdl2` is produced.
+ *     4. `mdl1` and `mdl2` are composed into {@link 
ModelsSequentialComposition}.
+ * </pre>
+ *
+ * @param <I> Type of input of model produced by first trainer.
+ * @param <O1> Type of output of model produced by first trainer.
+ * @param <O2> Type of output of model produced by second trainer.
+ * @param <L> Type of labels.
+ */
+public class TrainersSequentialComposition<I, O1, O2, L> extends 
DatasetTrainer<ModelsSequentialComposition<I, O1, O2>, L> {
+    /** First trainer. */
+    private DatasetTrainer<IgniteModel<I, O1>, L> tr1;
+
+    /** Second trainer. */
+    private DatasetTrainer<IgniteModel<O1, O2>, L> tr2;
+
+    /** Dataset mapping. */
+    private IgniteFunction<? super IgniteModel<I, O1>, DatasetMapping<L, L>> 
datasetMapping;
+
+    /**
+     * Construct sequential composition of given two trainers.
+     *
+     * @param tr1 First trainer.
+     * @param tr2 Second trainer.
+     * @param datasetMapping Dataset mapping.
+     */
+    public TrainersSequentialComposition(DatasetTrainer<? extends 
IgniteModel<I, O1>, L> tr1,
+        DatasetTrainer<? extends IgniteModel<O1, O2>, L> tr2,
+        IgniteFunction<? super IgniteModel<I, O1>, DatasetMapping<L, L>> 
datasetMapping) {
+        this.tr1 = CompositionUtils.unsafeCoerce(tr1);
+        this.tr2 = CompositionUtils.unsafeCoerce(tr2);
+        this.datasetMapping = datasetMapping;
+    }
+
+    /** {@inheritDoc} */
+    @Override public <K, V> ModelsSequentialComposition<I, O1, O2> 
fit(DatasetBuilder<K, V> datasetBuilder,
+        IgniteBiFunction<K, V, Vector> featureExtractor, IgniteBiFunction<K, 
V, L> lbExtractor) {
+
+        IgniteModel<I, O1> mdl1 = tr1.fit(datasetBuilder, featureExtractor, 
lbExtractor);
+        DatasetMapping<L, L> mapping = datasetMapping.apply(mdl1);
+
+        IgniteModel<O1, O2> mdl2 = tr2.fit(datasetBuilder,
+            featureExtractor.andThen(mapping::mapFeatures),
+            lbExtractor.andThen(mapping::mapLabels));
+
+        return new ModelsSequentialComposition<>(mdl1, mdl2);
+    }
+
+    /** {@inheritDoc} */
+    @Override public <K, V> ModelsSequentialComposition<I, O1, O2> update(
+        ModelsSequentialComposition<I, O1, O2> mdl, DatasetBuilder<K, V> 
datasetBuilder,
+        IgniteBiFunction<K, V, Vector> featureExtractor, IgniteBiFunction<K, 
V, L> lbExtractor) {
+
+        IgniteModel<I, O1> firstUpdated = tr1.update(mdl.firstModel(), 
datasetBuilder, featureExtractor, lbExtractor);
+        DatasetMapping<L, L> mapping = datasetMapping.apply(firstUpdated);
+
+        IgniteModel<O1, O2> secondUpdated = tr2.update(mdl.secondModel(),
+            datasetBuilder,
+            featureExtractor.andThen(mapping::mapFeatures),
+            lbExtractor.andThen(mapping::mapLabels));
+
+        return new ModelsSequentialComposition<>(firstUpdated, secondUpdated);
+    }
+
+    /** {@inheritDoc} */
+    @Override protected boolean checkState(ModelsSequentialComposition<I, O1, 
O2> mdl) {
+        // Never called.
+        throw new IllegalStateException();
+    }
+
+    /** {@inheritDoc} */
+    @Override protected <K, V> ModelsSequentialComposition<I, O1, O2> 
updateModel(
 
 Review comment:
   For the moment, no.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

Reply via email to