Supervisors info:
Χάρης Μεσαριτάκης, Αναπληρωτής Καθηγητής, Τμήμα Μηχανικών Πληροφοριακών και Επικοινωνιακών Συστημάτων, Πανεπιστήμιο Αιγαίου
Summary:
This thesis studies the quantum dot lasers from the scope of neuromorphic processors.
Neuromorphic processors are a new generation of processors which contrary to
conventional devices, its architecture does not consist of two separate systems for
processing and memory, but it has a different form that assimilates the neural system of
living beings. The neural system in human and other animals is significantly efficient
when it performs operations related to cognition and decision making. Such functionality
is extremely important for artificial intelligence, with neuromorphic processors being the
ideal paradigm to support future intelligent systems.
Although photonics is of fundamental importance in telecommunication systems, since
Internet is based on technologies such as the laser and optical fibers, this area has not
achieved similar remarkable results with respect to information processing, which is
based exclusively on electronics. Due to increased necessity for neuromorphic
computers, the similarity between the laser and the biological neurons has become the
object of many new studies. Lasers have properties similar to biological neurons but in a
different time scale that is 1000000 times smaller, which renders them ideal for
cognitive processing of high bandwidth signals.
In this work, a special laser that is based on the technology of quantum dots is studied.
Quantum dots contrary to other active materials can lead to lasing silicon devices by
creating silicon substrates carefully doped with quantum dots. Additionally, quantum dot
lasers have lower energy consumption compared to other lasers and they have the
ability to generate coherent light from multiple quantum states.
For these reasons, in this work a detailed mathematical study of the quantum dot laser
under optical injection takes place from the viewpoint of non-linear systems theory, in
order to highlight their neural properties. Under this mathematical prism, the similarity
between the quantum dot lasers and the biological neurons is examined. The
mathematical analysis focuses on the ability of the quantum dot laser for multi-
waveband emission and provides useful conclusions for the deployment of the quantum
dot laser neuron in a large neural system consisting of multiple neuromorphic nodes for
cognitive processing at the speed of light.
