network models 

IPR products and licences

​

​​

Machine learning software

 

• Multi-Layer, Deep Learning Neural Network software (MLDL Neural Network) 

​

• Optimally partitioned, multi dimensional state space for advanced machine learning. Software with applications in financial forecasting, image processing, medical diagnoses and prognosis.

​

 

Applications in Finance

 

A number of factors have - in the light of recent events - emerged as important driving forces in financial markets world-wide. Known collectively as market imperfections, they include: ”extreme” events, non-linear dependence, credit & liquidity risk, regime-change, etc. These factors are incorporated in the software implementations of our suit of IPR products for:

 

• Modelling asset prices and volatility

• Forecasting

• Trading signal generation

• Asset/Liability & Risk management

• Portfolio optimization & fund management, and

• Real-options valuation of projects & companies

• Credit rating

 

Technical advantages of Network Models methodology in finance

We use an improved version of the EM algorithm to replace missing data in any given time series with values that replicate properties of the original series.
We use both 'traditional' stochastic differential equations and newer data mining techniques to model asset dynamics.
We can model complex asset dynamics with: asymmetric and 'fat-tailed' return distributions, sudden price jumps and stochastic volatility.
We can model multiple assets with time and/or level-dependent correlations which can be deterministic or stochastic.
Our model maintains arbitrage-free multidimensional asset dynamics for realism.
We can include transaction costs (fixed and/or proportional), liquidity constraints and bid-offer spreads for portfolio (or hedge) trading.
Uses proprietary low-discrepancy sequences that are clearly superior to Sobol sequences, for Monte-Carlo simulations.
Our model uses the Bionomic Algorithm for better calibration of option pricing / hedging models to market prices and the volatility smile.
We use the same, consistent methodology for a variety of problems: portfolio and risk management (with user-defined risk measures), hedge-fund management, option pricing/hedging and company valuation.

 

This methodology, based on Dynamic Programming, duality and Statespace Relaxation, was developed by Network Models. The resulting consistency, enables an analyst to compare assets across different, diverse, classes.

 

 

Network Models grants licences to its intellectual property for the above applications.

​

​