Oracle Java SE is a programming language and gives customers enterprise features that minimize the costs of deployment and maintenance of their Java-based IT environment.
N/A
TensorFlow
Score 8.9 out of 10
N/A
TensorFlow is an open-source machine learning software library for numerical computation using data flow graphs. It was originally developed by Google.
Oracle Java SE is well suited to long-running applications (e.g. servers). Java Swing (UI toolkit) is now rather outdated, lacking support for modern UI features. JavaFX, the potential replacement for Swing, has now been separated out of Java core. Ideally, there would be a path to migrate a large application incrementally from Swing to JavaFX, but due to different threading models and other aspects, it is difficult. At this point, it is probably better to use an embedded web browser (e.g. JxBrowser) to provide a modern UI in HTML/Javascript and keep just the business logic in Java.
TensorFlow is great for most deep learning purposes. This is especially true in two domains: 1. Computer vision: image classification, object detection and image generation via generative adversarial networks 2. Natural language processing: text classification and generation. The good community support often means that a lot of off-the-shelf models can be used to prove a concept or test an idea quickly. That, and Google's promotion of Colab means that ideas can be shared quite freely. Training, visualizing and debugging models is very easy in TensorFlow, compared to other platforms (especially the good old Caffe days). In terms of productionizing, it's a bit of a mixed bag. In our case, most of our feature building is performed via Apache Spark. This means having to convert Parquet (columnar optimized) files to a TensorFlow friendly format i.e., protobufs. The lack of good JVM bindings mean that our projects end up being a mix of Python and Scala. This makes it hard to reuse some of the tooling and support we wrote in Scala. This is where MXNet shines better (though its Scala API could do with more work).
Commercial Licensing in 2019. Oracle will charge commercial organizations using Java SE for upgrading to the latest bug fixes and updates. Organizations will now need to either limit their implementation of Java SE or may need to drop it altogether.
Slow Performance. Due to the all of the abstraction of the JVM, Java SE programs take much more resources to compile and run compared to Python.
Poor UI appearance on all of the major GUI libraries (Swing, SWT, etc.). Through Android Studio, it is easy to get a native look/feel for Java apps, but when it comes to desktops, the UI is far from acceptable (does not mimic the native OS's look/feel at all).
Theano is perhaps a bit faster and eats up less memory than TensorFlow on a given GPU, perhaps due to element-wise ops. Tensorflow wins for multi-GPU and “compilation” time.
The language is fluent and has good support from a number of open source and commercial IDEs. Language features are added every 6 months, although long-term service releases are only available every 3 years. It would be nice if some of the older APIs were depreciated with more pressure to move to the new replacement APIs (e.g. File vs. Path), but transitions to new features are generally well implemented.
Java is such a mature product at this point that there is little support from the vendor that is needed. Various sources on the internet, and especially StackOverflow, provide a wealth of knowledge and advice. Areas that may benefit from support is when dealing with complex multithreading issues and security libraries.
Community support for TensorFlow is great. There's a huge community that truly loves the platform and there are many examples of development in TensorFlow. Often, when a new good technique is published, there will be a TensorFlow implementation not long after. This makes it quick to ally the latest techniques from academia straight to production-grade systems. Tooling around TensorFlow is also good. TensorBoard has been such a useful tool, I can't imagine how hard it would be to debug a deep neural network gone wrong without TensorBoard.
Chose to go with Java instead of Python or C++ due to the expertise on the ground with the technology, for its ease of integration with our heterogeneous setup of production servers, and for the third party library support which we've found was able to address some challenging aspects of our business problem.
Keras is built on top of TensorFlow, but it is much simpler to use and more Python style friendly, so if you don't want to focus on too many details or control and not focus on some advanced features, Keras is one of the best options, but as far as if you want to dig into more, for sure TensorFlow is the right choice
The different versions make it harder to work with other companies where some use newer versions while some use older versions, costing time to make them compatible.
Licenses are getting to be costly, forcing us to consider OpenJDK as an alternative.
New features take time to learn. When someone starts using them, everyone has to take time to learn.
