• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site
National Research University Higher School of EconomicsStudent ThesesPredictive Modeling in Big Data: Learning Multimodal Enviroment with Deep Reinforcement Learning

Student
Title
Supervisor
Faculty
Educational Programme
Final Grade
Year of Graduation
Tatiana Grunina
Predictive Modeling in Big Data: Learning Multimodal Enviroment with Deep Reinforcement Learning
Big Data Systems
(Master’s programme)
2017
Currently neural networks are known to be one of the most powerful supervised learning methods in machine learning, as they show high performance on threshold, ordering and probability metrics among very different datasets, and often beat methods from other families . They are not programmed in a well-known sense, instead they are learning. Their ability to learn being the main advantage in comparison to traditional predictive methods is a task of finding coefficients of relations between the neurons. This gives the neural network the power to reveal the complex relations between the inputs and the outputs, as well as executing generalization. That means that in a case of successful training a neural network can return the right result based on the data which was absent in the training dataset, or was partially biased, noised.

The method is widely used in business applications, including logistics, business planning, prediction of clients behaviour, forecasting, marketing research solutions and even for the optimization of data center cooling infrastructures. Neural networks with online learning also can be applied by financial institutions that deal with KPI tracking or with company inner data consisting of millions of events per second. Analytics capabilities such as live-stream processing and visual analytics have enabled the management boards to act on a constantly updating environment. This can allow to display insights and analysis gained from that data in an approachable to human eye and brain visual format. This fastens and facilitates data-driven decision making.

Deep reinforcement learning is known to be one of the most accurate method to train a neural network solving problems of live-stream processing and visual analytics. Papers on new approaches appears weekly if not daily in the field. Thus we wanted to explore its advantages in details as well as to evaluate the up-to-date methods and check the heuristics.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses