Does Machine Learning Bring a Novel Epistemology in Communications ?

[…]  Significant progress in the analysis and design of communication systems has been rendered possible by the milestone work of C. Shannon that provided a methodological plant to attack the challenge of reliably transmitting information through a given communication mean. Shannon’s mathematical approach suggests to represent the system as a chain of blocks mathematically modeled, namely, the transmitter, the channel, and the receiver. […] Three generations of scientists and engineers grew up with this mathematical mind set which provided tools to acquire domain knowledge and use it to build a model for each block so that the overall behavior becomes known. Such a framework intrinsically has the advantage that each block can be individually studied and optimized. We would refer to this approach as physical and bottom-up.

From an epistemological point of view, the mathematical theory of communications is based on knowledge coming from a priori justifications and relying on intuitions and the nature of these intuitions, which is intrinsically what mathematics does. On the contrary, a posteriori knowledge is created by what is known from experience, therefore generated afterwards with an empirical and top-down approach. The immense contributions of I. Kant with his philosophy of transcendental aesthetics and logic, made a step further into the understanding and the definition of knowledge: knowledge of the structure of time and space and their relationships, is a priory knowledge; knowledge acquired from observations is a posteriori knowledge; most of our knowledge comes from the process of learning and observing phenomena, and without a priori knowledge is impossible to reach the true knowledge. In this respect, the broad field of machine learning (ML) can be considered an implementation by humans of techniques in machines to acquire knowledge from a posteriori observations of realizations of natural phenomena. The origin of success of machine learning relies on its ability to derive relations among phenomena and potentially discover the hidden (latent) state of a system, i.e, potentially provide an intrinsic true knowledge of the system. System identification and model based design through the aid of machine learning constitute a first step to find undiscovered system properties via a mixed a priori – a posteriori learning approach, which, retrospectively, follows Kant’s philosophical plant.

Machine learning is indeed bringing new lymph in the domain of communication systems modeling, design, optimization, and management. It provides a paradigm shift: rather than concentrating on a physical bottom-up description of the communication scheme, ML aims to learn and capture information from a collection of data, to derive the input-output relations of the system or of a given task in the system. We would argue that a miraculous solution of communications challenges with ML does not yet exist, and that what is learned via ML is not necessarily representative of the physical reality, rather it can generate miss believes on existing relations and dependencies, which requires validating results through a probabilistic framework. But the path has been mapped out: ML offers a great deal of opportunities to research and design communication systems. […]

The above argumentation on the epistemological sense of ML is an excerpt from a recent paper contribution [1] where we have discussed what the domains of application of ML in communications, and in particular power line communications (PLC), are. The paper provides a vision of what ML can do in PLC.  A unified formulation of ML tools within a probabilistic framework that are relevant to communications is offered. Then, ML applications in PLC for each layer, namely, for characterization and modeling, for physical layer algorithms, for media access control and networking algorithms. are considered. Finally, other applications of PLC that can benefit from the usage of ML, as grid diagnostics, are analyzed. Illustrative numerical examples are reported to serve the purpose of validating the ideas and motivate future research endeavors in this stimulating signal/data processing field.


Further reading

A. M. Tonello, N. A. Letizia, D. Righini, F. Marcuzzi, “Machine Learning Tips and Tricks for Power Line Communications,” arXiv:1904.11949, April 2019 (selected  as part of best readings by the Research Library of the IEEE Machine Learning For Communications Emerging Technologies Initiative).


Recommended Posts