2
The information can be encoded in terms of time of release, concentration, or molecule
type [4]. Nano network, a future framework, is designed with the inter-connection of
Nano-machines. Modeling of Nano networks has been proposed in biomedical,
environmental, and military applications [5].
The Internet of Bio-Nano Things (IoBNT) is presented as a network of biological
components and nano-machines that communicate with one another and the internet to
empower different applications such as healthcare-monitoring. Additionally, MC has
emerged as the most capable nano-scale communication prototypical for IoBNT [6].
Moreover, it has been identified as a potential upcoming application in the health
field, and we can see this in disease monitoring/therapy/diagnosis [7] as well as target
drug-delivery (TDD) [8].
Detection techniques are very important for the trustworthy retrieval of the signal
directed from a communication source, in which the concentration encoded signal is
calculated from the degraded variety detected at the decoder [9]. Detection technique
design and analysis have traditionally relied on computational models that describe the
transmission process, signal and receiver noise propagation, and many other
components of end-to-end signal transmission and reception [10].
MC relies on chemical signal and doesn’t use electromagnetic (EM) signals, and the
consistent source model may be unidentified or mathematically irreconcilable. One
solution to this problem is using detection techniques inspired by ML [11] which can
be used to design detection techniques that can learn directly from data.
It is clear that there are no research creativities in the field of MC-based signal
detection of the received signal until time using automated ML. In a production-ready
environment, model automation is frequently disregarded. As a result, we investigate
automation in the proposed work to make the model accessible to the end-user.
Whenever there is a requirement to run computationally complex ML models,
automation is usually required. For automation, one viable solution to examine is ML
as a service (MLaaS) [12]. Automated ML implies that scientists do not need to create
a logistic regression or a neural network model from zero but may utilize the benefits
of automated algorithmic components offered on MLaaS [13]. As a proof of concept,
the work of automation is accomplished utilizing the MLaaS model. Typically, the data
which ML programs rely on it is challenging to process and store. It affords a robust
and effective crossing point that allows ML specialists to carry out experiments with
their datasets rather than worrying about storage, compute, or networking issues [14].
When the difficulty of data grows, ML poses numerous difficulties. As a result,
automating an ML activity becomes critical.
Hyper-Parameter Optimization (HPO) is without a doubt one of the most important
aspects of automated ML [15], which is regularly challenging to attain. As a result, in
the proposed scheme, we used Azure ML, a production-ready tool, to ease designing
ML infrastructure. Essentially, it is critical to recognize a significant variance between
implementing algorithms on a device and software as a service portal. However, it is
worth noting that all of these algorithms are performed on an MLaaS prototype, which
allows for extensive fully automated tuning of hyper-parameters, significantly lowering
human work and time. Azure ML, when combined with ML Operations, aids in the
improvement of workflow performance. Traditional ML frameworks' credibility in a
real-time scenario involving massive networking data is highly dubious. However, if
the focus is given on the use of MLaaS settings such as BigML, Algorithmic, Data