How to add a BatchNorm layer after a 1D CNN layer?

DL4J
1

How can I add a BatchNormalization layer after a 1D CNN layer? The BatchNormalization layer is designed for 2D CNNs by default, and adding it directly after a 1D CNN layer throws an error complaining about incompatible input shapes.

2

@cqiaoYc which api are you using? computation graph or samediff? You could try using snapshots.

Try:

``
<repositories>
<repository>
<id>central-portal-snapshots</id>
<name>Central Portal Snapshots</name>
<url> Repository - Sonatype Nexus Repository </url>
<releases>
<enabled>false</enabled>
</releases>
<snapshots>
<enabled>true</enabled>
</snapshots>
</repository>
</repositories>
```

Use org.eclipse.deeplearning4j for the groupId imports should still be the same. I remember fixing a BUNCH of bugs with keras import and 1d cnns.

3

@agibsonccc I am using the NeuralNetConfiguration class, Version:

      <dependency>
            <groupId>org.deeplearning4j</groupId>
            <artifactId>deeplearning4j-core</artifactId>
            <version>1.0.0-M2.1</version>
      </dependency>
4

@cqiaoYc yeah I’m aware. I’m saying I fixed a bunch of bugs in snapshots and you can give that a try. The namespace will change in the next release to org.eclipse.deeplearning4j though.

The imports won’t change for now though. I’m making a transition release before doing that refactoring.