Page 77 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

Chapter 3 Application, algorithm, tools directly related to deep learning 65

Deep neural network includes up to 36,000 nodes. Tensor
Board collapses these nodes in high-level blocks and then high-
lights entire identical structures. This allows clear analysis of
graph focusing on the primary sections of the computation
graph. The Tensor Board visualization is seemed to be interactive
where a user can zoom and expand the nodes to display the de-
tails [2].
The algorithms collapse nodes into high-level blocks and
outstand the specific groups with identical structures and distin-
guish high-degree nodes. The Tensor Board thus created is very
much useful and is used for tuning a machine learning model.
This visualization tool is designed for the configuration log file,
which contains entire summary information and details that
need to be displayed.

2.2 Keras
Keras is an open-source neural network library that can be
written in Python that runs mostly on top of Theano or Tensor-
Flow. It is modular, fast, and very easy to use. It was made by
François Chollet, a Google engineer.
Keras low-level computation is not handled in Keras. More-
over, it uses another library to do the functions, called the “back-
end.” Keras is high-level API wrapper for all low-level API, capable
of running TensorFlow, CNTK, and Theano [3].
Keras high-level API handles the way we make models,
defining inputeoutput models [4]. Keras also compiles our model
with different loss and optimizer functions, training process, and
fit function. Keras does not handle low-level API such as compu-
tational graph, making tensors or other variables because it has
handled only by the “backend” engine [4]. The difference between
keras and tensorflow is shown in Table 3.1[5,28].

2.2.1 Backend in Keras
Backend Keras performs all low-level computation, namely,
Tensor products, convolutions, and many other things with the
help of other libraries such as TensorFlow or Theano [5].
One more backend engine for Keras is the Microsoft Cognitive
Toolkit or CNTK. It is an open-source deep learning framework
that was developed by Microsoft team. It can be executed on
multi-GPUs or multimachine for training deep learning model
on a massive scale. In some cases, CNTK was reported faster
than other frameworks such as TensorFlow or Theano.