Frequently Asked Questions
Not at the moment, but there is an ongoing development to add GPU support.
Not directly through its own deep learning library, it will be able to support it through third-party frameworks.
It can be slow for large amounts of data for complex problems (like video), but in many cases it can provide comparable performance and to be more suitable than GPU/distributed solutions in both performance and cost.
|Framework / Library
|JSR 381 support
|Commercial support available
|TensorFlow for Java
|High performance, GPU support, direct use of models created in Python.
|Java API experimental, not actively supported. Not so easy-to-use. Difficult for beginners, even experienced dev, very unintuitive for Java developers.
|Pure Java, intuitive API, easy to use for beginners, RI for standard Java API JSR381
|No GPU or Distributed support at the moment is a limitation for large scale data sets.
|GPLv2 with CPE
|Supports many types of architectures, GPU, distributed training with Spark.
|Not very friendly API, hard to follow and understand for non-experts and beginners.
|Framework agnostic, support for standard Java API JSR381.
|Mainly a wrapper for other existing libraries (MXNet, but TensorFlow and PyTorch support coming).
The Deep Netts is a perfect fit if you require: – pure java deep learning solution – easy to integrate, distribute and maintain deep learning solution – smaller and mid-scale machine learning problems (in terms of data) with large scale deployment – easy to use for users with less experience with machine learning – friendly tools for users with software development backgroundt vivamus ut a condimentum neque ac.
Technically yes, the Deep Netts development tool is able to support any 3rd party deep learning framework. The Tensorflow support is under development, and if you’re interested in commercial development for support for some other framework please let us know.
Impossible to say since it depends of the problem complexity and size of the model. Best way to check this is to try it using Free Development License