I have been working on a Deep Learning4J workflow and have got to a logical endpoint where I have decent predictions but was curious if there was any way to visualise the data from the CNN. At the moment I am using the DL4J Model writer but when it writes out it has no extension. One of the best aspects of KNIME is stepping into the outputs of each node to be able to see what needs tweaking on the nodes. Is there any liklihood to see a visualisation of CNN in DeepLearning4J nodes in the future? Otherwise is there any programs I can use where I can easily represent the output model?
unfortunately, visualizing deep learning models is not trivial and an ongoing research topic at the moment. There, are some options available, however it greatly depends on what you exactly want to examine and the specific model architecture. Afaik, there is no general visualization tool for model visualization.
For these reasons, there is no view of the DL4J model and no visualization node.
Regarding the DL4J Model Writer, it just outputs a ZIP file containing the model configuration in JSON format and model parameters.