SPHS20
Home Thesis
Sinusodial House of the Rising Sun Mackey-Glass Equations Mid-price Forecasting
Cartpole(ESN) Mid-price Trading
Links Log

MT ⋅ HS21

Thesis Page

Contents

Thesis
Thesis. Contains the up-to-date results.
Reservoir Computing Examples
  • Sinusoidal:
    Features RC-RNN Comparison on Sine functions. RC displays superiority over RNN in terms of speed and accuracy.
  • House of the Rising Sun:
    Features RC on learning melodies using output feedback method. Showcases the problem with periodic inputs and how to tackle it. Includes RC-RNN comparison.
  • Mackey-Glass Equations:
    Features RC-RNN Comparison on Mackey-Glass equations. RC displays superiority over RNN in terms of speed, accuracy, generativity and robustness to smaller dataset.
  • Mid-price Forecasting:
    Features a Reservoir Computing approach to the work done in SPHS20.
Reinforcement Learning Examples
  • Cartpole(ESN):
    Features Echo State Network (ESN) trained on the cartpole environment.
  • Mid-price Trading:
    Features reinforcement learning algorithms trading on mid-prices using high frequency limit order book data. The models used are fully trained feedforward networks, Extreme Learning Machines and Echo State Networks.
Links
  • Github:
    Contains the pdf of the thesis and the programming resources, which can be used to reproduce the results shown in the thesis.
  • Thesis:
    Thesis on Github.
  • EchoStateNetwork:
    Documentation of the framework written for Echo State Network applications.
Log
Progress updates and changes made to the website.