Lepton provides expertise in the modeling of data including: bioinformatics, time series prediction  and financial data analysis.  While our services include traditional linear modeling we also can provide more sophisticated analysis based on cutting edge techniques such as Neural Networks, Chaotic Attractor Reconstruction, Hidden Markov Models and Stochastic Grammars.  We thrive on messy and difficult problems with insufficient information and non-linear relationships. 

Lepton's Advantage

Capabilities

Classical Linear Models

At Lepton, we can begin with standard techniques such as: Orthogonal Basis Linear Regression, Markov Modeling, Principal Components Analysis and Generalized AutoRegressive Conditional Heteroskedasticity modeling. These models often serve as a good baseline to evaluate the efficacy of more advanced methods.

 

Advanced Nonlinear Models

When the data is more intricate, more complex techniques can provide deeper insights into relationships within data. Lepton can generate non-linear models based on Hidden Markov Models, Multi-layer Neural Networks, Chaotic Attractor Reconstruction, and Stochastic Context Free Grammars.  Lepton can also generate custom models that encompass your expert understanding of the problem domain through the use of Bayesian Inference. For the most difficult problems, where the process domain is not well understood, Lepton can also implement Genetic Co-evolutionary Systems to search very large process model spaces.

 

Applications of Modeling

Lepton can also provide expertise in the use of effective models.  Strongly descriptive models can often be leveraged into predictive models.  Predictive models can be employed for data compression and utility base decision management support.

[ Home ] [ Services ] [ Contents ] [ Previous Web Site ] [ Contacts ]

Send mail to WebServices@LeptonCorp.com with questions or comments about this web site.
Copyright © 2002 Lepton Incorporated
Last modified: April 15, 2002