From 45bbf12694850d1e70ae463d97ff633221ed0bbc Mon Sep 17 00:00:00 2001 From: NA-V10 Date: Tue, 18 Nov 2025 21:09:22 +0530 Subject: [PATCH 1/3] Improve and expand README for NDArray examples --- nd4j-ndarray-examples/README.md | 160 ++++++++++++++++++++++---------- 1 file changed, 111 insertions(+), 49 deletions(-) diff --git a/nd4j-ndarray-examples/README.md b/nd4j-ndarray-examples/README.md index 03be227cf9..12521cab25 100644 --- a/nd4j-ndarray-examples/README.md +++ b/nd4j-ndarray-examples/README.md @@ -1,50 +1,112 @@ -## Eclipse Deeplearning4j: ND4J NDArray Examples - -This project contains a set of examples that demonstrate how to manipulate NDArrays. The functionality of ND4J demonstrated here can be likened to NumPy. - -The examples in this project along with a short summary are listed below. This is also the recommended order to explore them in. - -[Go back](../README.md) to the main repository page to explore other features/functionality of the **Eclipse Deeplearning4J** ecosystem. File an issue [here](https://github.com/eclipse/deeplearning4j-examples/issues) to request new features. - -## Quickstart -* [Nd4jEx0_INDArrayBasics.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx0_INDArrayBasics.java) -Creating, concatenating, stacking, padding NDArrays -* [Nd4jEx1_INDArrayBasics.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx1_INDArrayBasics.java) -Printing, querying shape properties, modifying elements, doing operations -* [Nd4jEx2_CreatingINDArrays.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx2_CreatingINDArrays.java) -Different ways to create NDArrays -* [Nd4jEx3_GettingAndSettingSubsets.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx3_GettingAndSettingSubsets.java) -Querying and modifying subset of an NDArray -* [Nd4jEx4_Ops.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx4_Ops.java) -More ops on NDArrays -* [Nd4jEx5_Accumulations.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx5_Accumulations.java) -Reduction Operations -* [Nd4jEx6_BooleanIndexing.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx6_BooleanIndexing.java) -Simple conditional element wise operations -* [Nd4jEx7_MatrixOperation.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx7_MatrixOperation.java) -Matrix multiplication -* [Nd4jEx8_ReshapeOperation.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx8_ReshapeOperation.java) -Reshaping -* [Nd4jEx9_Functions.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx9_Functions.java) -Transforming NDArrays -* [Nd4jEx10_ElementWiseOperation.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx10_ElementWiseOperation.java) -Even more operations like add row/col etc -* [NumpyCheatSheet.java](NumpyCheatSheet.java) (FIXME???) -Examples to help NumPy users get acquainted with ND4J - -## Advanced -* [MultiClassLogitExample.java](./src/main/java/org/nd4j/examples/advanced/lowlevelmodeling/MultiClassLogitExample.java) -Multiclass logistic regression from scratch with ND4J -* [WorkspacesExample.java](./src/main/java/org/nd4j/examples/advanced/memoryoptimization/WorkspacesExample.java) -For cyclic workloads like training a neural net the DL4J ecosystem does not rely on garbage collection. Instead a chunk of memory is resued avoiding the performance hits from expensive pauses for GC. Workspaces are used by default when calling .fit on a neural network etc. This example demonstrates the concepts behind it for advanced users if they need to go beyond what is available by default in the library in their particular use case. -* [Nd4jEx11_Axpy.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx11_Axpy.java) -Use the ND4J blas wrapper to call the AXPY operation -* [Nd4jEx12_LargeMatrices.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx12_LargeMatrices.java) -Operations with a 10000000 element NDarray and its transpose -* [Nd4jEx13_Serialization.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx13_Serialization.java) -Examples for binary and text serialization. -* [Nd4jEx14_Normalizers.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx14_Normalizers.java) -Create and fit a normalizer, and save and restore it. -* [CustomOpsExamples.java](./src/main/java/org/nd4j/examples/advanced/operations/CustomOpsExamples.java) -**Only relevant to the 1.0.0-beta6 release**. There are some operations that were implemented in C++ that had not been mapped to Java. This example demonstrates how to access them using ND4J's DynamicCustomOp. As of the beta7 release all maps have corresponding Java mappings. +# ND4J NDArray Examples + +This module contains simple examples that demonstrate how to use **ND4J** +(N-dimensional arrays for Java) ā€” the core numerical computing library behind Deeplearning4j. + +ND4J provides fast tensor operations similar to NumPy, backed by optimized BLAS +implementations. + +These examples help new users understand: + +- What an `INDArray` is +- How to create arrays +- How to perform basic math +- How to reshape, slice, and combine arrays +- How to run the examples using Maven + +--- + +## šŸ“Œ What is an NDArray? + +An **NDArray** (N-dimensional array) is ND4Jā€™s fundamental data structure. +It represents: + +- Scalars (0D) +- Vectors (1D) +- Matrices (2D) +- Higher dimensional tensors (3D, 4D, ā€¦) + +ND4J operations are *vectorized* and run efficiently on CPU or GPU (via CUDA backend). + +Example: + +```java +INDArray arr = Nd4j.create(new float[]{1, 2, 3}); +System.out.println(arr); // [1.00, 2.00, 3.00] +šŸ“˜ Basic NDArray Operations +Create Arrays +java + +INDArray zeros = Nd4j.zeros(3, 3); +INDArray ones = Nd4j.ones(2, 2); +INDArray random = Nd4j.rand(1, 5); +INDArray arange = Nd4j.arange(1, 10); +Math Operations +java + +INDArray a = Nd4j.create(new float[]{1, 2, 3}); +INDArray b = Nd4j.create(new float[]{4, 5, 6}); + +System.out.println(a.add(b)); // elementwise add +System.out.println(a.mul(b)); // elementwise multiply +System.out.println(a.mmul(b.T())); // matrix multiplication +Reshape and Shape Ops +java +Copy code +INDArray m = Nd4j.linspace(1, 9, 9).reshape(3, 3); +System.out.println(m); + +System.out.println(m.transpose()); +System.out.println(m.permute(1, 0)); +ā–¶ļø How to Run These Examples +Make sure you have: + +Java 8 or later + +Maven installed + +Then run: + + +mvn -q exec:java -Dexec.mainClass="org.nd4j.examples.quickstart.BasicNDArrayExample" +Or run any other example under: + + + +src/main/java/org/nd4j/examples/ +Example: + + +mvn -q exec:java -Dexec.mainClass="org.nd4j.examples.advanced.SlicingExample" +šŸ“¤ Expected Output (Sample) +Running a basic NDArray creation example: + + +Copy code +Zeros array: +[[0.00, 0.00, 0.00], + [0.00, 0.00, 0.00], + [0.00, 0.00, 0.00]] + +Random array: +[0.42, 0.91, 0.12, 0.55, 0.33] + +Vector add: +[5.00, 7.00, 9.00] +šŸ“ Project Structure + +nd4j-ndarray-examples/ + ā”œā”€ā”€ src/ + ā”‚ ā””ā”€ā”€ main/java/org/nd4j/examples/ + ā”‚ ā”œā”€ā”€ advanced/ + ā”‚ ā”œā”€ā”€ quickstart/ + ā”‚ ā”œā”€ā”€ utils/ + ā”‚ ā””ā”€ā”€ resources/ + ā”œā”€ā”€ pom.xml + ā””ā”€ā”€ README.md +šŸ“„ License +This project is part of the Eclipse Deeplearning4j examples collection +and is licensed under the Apache License 2.0. + +Happy coding with ND4J tensors! šŸš€ From 0280b2b1ebf545702d8bd1f56879d00151a8dc13 Mon Sep 17 00:00:00 2001 From: NA-V10 Date: Wed, 19 Nov 2025 06:39:24 +0530 Subject: [PATCH 2/3] Refine README: restore original context + improve explanations --- nd4j-ndarray-examples/README.md | 55 +++++++++++++++++++++++++++++++-- 1 file changed, 53 insertions(+), 2 deletions(-) diff --git a/nd4j-ndarray-examples/README.md b/nd4j-ndarray-examples/README.md index 12521cab25..382a244419 100644 --- a/nd4j-ndarray-examples/README.md +++ b/nd4j-ndarray-examples/README.md @@ -1,3 +1,55 @@ + + +## Eclipse Deeplearning4j: ND4J NDArray Examples + +This project contains a set of examples that demonstrate how to manipulate NDArrays. The functionality of ND4J demonstrated here can be likened to NumPy. + +The examples in this project along with a short summary are listed below. This is also the recommended order to explore them in. + +[Go back](../README.md) to the main repository page to explore other features/functionality of the **Eclipse Deeplearning4J** ecosystem. File an issue [here](https://github.com/eclipse/deeplearning4j-examples/issues) to request new features. + +## Quickstart +* [Nd4jEx0_INDArrayBasics.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx0_INDArrayBasics.java) +Creating, concatenating, stacking, padding NDArrays +* [Nd4jEx1_INDArrayBasics.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx1_INDArrayBasics.java) +Printing, querying shape properties, modifying elements, doing operations +* [Nd4jEx2_CreatingINDArrays.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx2_CreatingINDArrays.java) +Different ways to create NDArrays +* [Nd4jEx3_GettingAndSettingSubsets.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx3_GettingAndSettingSubsets.java) +Querying and modifying subset of an NDArray +* [Nd4jEx4_Ops.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx4_Ops.java) +More ops on NDArrays +* [Nd4jEx5_Accumulations.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx5_Accumulations.java) +Reduction Operations +* [Nd4jEx6_BooleanIndexing.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx6_BooleanIndexing.java) +Simple conditional element wise operations +* [Nd4jEx7_MatrixOperation.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx7_MatrixOperation.java) +Matrix multiplication +* [Nd4jEx8_ReshapeOperation.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx8_ReshapeOperation.java) +Reshaping +* [Nd4jEx9_Functions.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx9_Functions.java) +Transforming NDArrays +* [Nd4jEx10_ElementWiseOperation.java](./src/main/java/org/nd4j/examples/quickstart/Nd4jEx10_ElementWiseOperation.java) +Even more operations like add row/col etc +* [NumpyCheatSheet.java](NumpyCheatSheet.java) (FIXME???) +Examples to help NumPy users get acquainted with ND4J + +## Advanced +* [MultiClassLogitExample.java](./src/main/java/org/nd4j/examples/advanced/lowlevelmodeling/MultiClassLogitExample.java) +Multiclass logistic regression from scratch with ND4J +* [WorkspacesExample.java](./src/main/java/org/nd4j/examples/advanced/memoryoptimization/WorkspacesExample.java) +For cyclic workloads like training a neural net the DL4J ecosystem does not rely on garbage collection. Instead a chunk of memory is resued avoiding the performance hits from expensive pauses for GC. Workspaces are used by default when calling .fit on a neural network etc. This example demonstrates the concepts behind it for advanced users if they need to go beyond what is available by default in the library in their particular use case. +* [Nd4jEx11_Axpy.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx11_Axpy.java) +Use the ND4J blas wrapper to call the AXPY operation +* [Nd4jEx12_LargeMatrices.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx12_LargeMatrices.java) +Operations with a 10000000 element NDarray and its transpose +* [Nd4jEx13_Serialization.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx13_Serialization.java) +Examples for binary and text serialization. +* [Nd4jEx14_Normalizers.java](./src/main/java/org/nd4j/examples/advanced/operations/Nd4jEx14_Normalizers.java) +Create and fit a normalizer, and save and restore it. +* [CustomOpsExamples.java](./src/main/java/org/nd4j/examples/advanced/operations/CustomOpsExamples.java) +**Only relevant to the 1.0.0-beta6 release**. There are some operations that were implemented in C++ that had not been mapped to Java. This example demonstrates how to access them using ND4J's DynamicCustomOp. As of the beta7 release all maps have corresponding Java mappings. + # ND4J NDArray Examples This module contains simple examples that demonstrate how to use **ND4J** @@ -51,8 +103,7 @@ System.out.println(a.add(b)); // elementwise add System.out.println(a.mul(b)); // elementwise multiply System.out.println(a.mmul(b.T())); // matrix multiplication Reshape and Shape Ops -java -Copy code + INDArray m = Nd4j.linspace(1, 9, 9).reshape(3, 3); System.out.println(m); From d47dd4787bc064d43fb1617d201a998c935e07a6 Mon Sep 17 00:00:00 2001 From: NA-V10 Date: Sun, 29 Mar 2026 09:48:48 +0530 Subject: [PATCH 3/3] Improve charmodelling README documentation --- .../charmodelling/melodl4j/README.md | 36 +++++++++++++++++++ 1 file changed, 36 insertions(+) diff --git a/dl4j-examples/src/main/java/org/deeplearning4j/examples/advanced/modelling/charmodelling/melodl4j/README.md b/dl4j-examples/src/main/java/org/deeplearning4j/examples/advanced/modelling/charmodelling/melodl4j/README.md index e6887b62d1..8b00d8553f 100644 --- a/dl4j-examples/src/main/java/org/deeplearning4j/examples/advanced/modelling/charmodelling/melodl4j/README.md +++ b/dl4j-examples/src/main/java/org/deeplearning4j/examples/advanced/modelling/charmodelling/melodl4j/README.md @@ -27,6 +27,33 @@ To help judge the extent to which generated melodies mimic existing melodies in See http://www.asimovinstitute.org/analyzing-deep-learning-tools-music/, and https://github.com/tensorflow/magenta for some previous work in music generation via deep learning. I found out recently that MELODL4J is similar to magenta: both extract monophonic melodies from MIDI -- and both have a class named NoteSequence! -- but MELODL4J was developed independently and uses different techniques. * * * +## Prerequisites + +- Java 8 or later +- Maven installed +- Internet connection to download MIDI datasets +- Audio output enabled for melody playback +- Enough memory for training larger LSTM models + +## Quick Start + +1. Run `MidiMelodyExtractor.java` to extract melody strings from MIDI files. +2. Run `MelodyModelingExample.java` to train the LSTM network. +3. Listen to generated melodies after each epoch. +4. Review the generated melody output file after training. + +## Main Classes + +- `MidiMelodyExtractor.java` ā€“ Extracts symbolic melodies from MIDI files +- `MelodyModelingExample.java` ā€“ Trains the LSTM model +- `PlayMelodyStrings.java` ā€“ Converts melody strings back into MIDI and plays them +- `NoteSequence.java` ā€“ Represents note sequences internally + +## How to Run + +```bash +mvn -q exec:java -Dexec.mainClass="org.deeplearning4j.examples.advanced.modelling.charmodelling.melodl4j.MelodyModelingExample" + ### Overview of methodology, simplifications, and tricks 1. MidiMelodyExtractor.java parses MIDI files and outputs melodies in symbolic form. @@ -55,6 +82,15 @@ See http://www.asimovinstitute.org/analyzing-deep-learning-tools-music/, and htt * * * +## Expected Output + +During training, you should see: +- MIDI files being downloaded and extracted +- Melody extraction statistics +- Network training progress +- Generated melodies after each epoch +- Output file containing generated melodies + ### Possible directions for improvement 1. Represent durations numerically, not symbolically. Currently, both pitches and durations are represented symbolically (as characters), not numerically. This makes sense for pitches probably, since the qualitative feel of a C followed by a G is quite different from the qualitative feel of a C followed by a G#. Likewise, following a C, a G is more similar to a E than to a G#; both E and G are notes in a C major chord. But for tempos, a 32d note is more similar to a 16th note than to a quarter note, etc.