Deep learning methods (DL) have given groundbreaking results on several difficult machine learning tasks, notably complex image recognition and machine translation. We use DL to predict future behavior of prices on financial markets. Such prediction is a very difficult machine learning task. The difficulty stems from the fact that very little is known about structure of such data and the structure itself is changing in time. We need sophisticated visualization methods to gain basic understanding of data and, in particular, to extract their important features.

The aim of this thesis is to develop a web-based interactive environment in which financial time series can be visualized. The environment should allow:

1. Display of time series, moving and zooming along the time axis.
2. Visualization of values of output neurons in different market situations (so that networks can be diagnosed).

In addition, the environment is supposed to have the following properties:

1. It should allow multiple users, including personalized settings and basic security measures (password-secure access).
2. It should be optimized to work with large time series.

The environment should be developed in .NET framework, preferably in C#.