témavezető: Tóth Zsolt
helyszín (magyar oldal): Debreceni Egyetem Informatikai Tudományok Doktori Iskola helyszín rövidítés: ITDI
A kutatási téma leírása:
Project Description
Artificial Neural Networks gains popularity with the widespread of Deep Learning. Due to the ever growing computational power, bigger and more complex
artificial neural networks can be trained to solve more difficult problems. Artificial Neural Networks are applied in a wide range of disciplines such as economy,
finance, engineering, healthcare, agriculture or IT security. The most common
tasks are regression, modeling, classification and pattern recognition. Development of more and more intelligent software and applications just increases the
need for these systems. So the researchers are focusing on the training bigger
artificial neural networks which can solve more complex tasks and software engineers are seeking the market gaps and application areas. Even the pruning
of artificial networks is popular topic these days because it allows their usage
in embedded systems where the computational capacity is limited. Although
training, application and even pruning of artificial neural networks are actively
researched and popular topics their inversion is also a vital and important topic.
Inversion of artificial neural networks can be considered as a parameter
searching problem where the model is defined by the neural network. The
input parameters are searched which yields the expected output. Inversion can
be applied for any regression and modeling task where feed-forward artificial
neural networks are applied. For example, in a production process the configuration of the machines affects on the product. So inversion of artificial neural
network which models the production process allows us to find some settings
which results products with expected properties. Finding a single input or a set
of possible solutions is also a task of inversion.
This research project has three major goals. Firstly, it gives an overview
and analyzes the applicability of existing inverting techniques in the field of
deep learning. Secondly, the project investigate the applicability and viability of
inverted artificial networks in modern web applications and distributed systems.
Finally, the theoretical results should be demonstrated via examples.
Tasks
The successful candidate will
• analyze the existing artificial neural network training and inversion techniques.
• implement algorithms for single and multiple parameter search.
• take part in seminars and hold tutorials.
• publish the results in international journals and conferences.
Requirements
The successful candidate has
• a degree in Computer Science, Information Technology, Mathematics or
other related fields.
• a strong analytical skill and familiar with data mining techniques and
metaheuristic optimization.
• proficiency in at least one of the following programming languages: Python,
Java.
Expected Results
• Functions for Artificial Neural Network Inversion in Python or Java.
• Demo application and proof of concept implementation related to different
disciplines.
• Possible cooperation with other research groups where Artificial Neural
Network Inversion can be applied.
• Publications at top venues
Jelentkezési határidő: 2021-11-15
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Bejelentkezés
