Bayesia News, September 2010

BayesiaLab 5.0

We are delighted to announce a wide range of major innovations in this new release of BayesiaLab:

• Innovative Knowledge Elicitation Environment: An entirely new module for knowledge elicitation allows to independently collect different experts' domain knowledge via an easy-to-use web-based tool hosted on Bayesia's secure server. The experts individuals assessments can thus be formally assembled to generate a Bayesian network, while entirely avoiding the typical biases caused by group dynamics in brainstorming sessions. The following presentation explains this new feature in greater detail;
• Causal Inference: As an industry-first, Judea Pearl's Do-Calculus is now integrated as a convenient function in BayesiaLab. This unique feature allows analysts to simulate the impact of interventions for formal policy analysis, as opposed to merely observing the states of variables. Furthermore, all of BayesiaLab's comprehensive analysis tools are fully compatible with this new kind of causal inference;
• Multi-Quadrants: This is ideally suited for problem domains that include a "break-out" or "pivot" variable, such as "product type" within a network of "product rating" variables. A SWOT-type chart allows to compare the drivers specific to each state of the break-out variable. The results can be used directly with the Target Dynamic Profile function (TDP), which can find optimal and, more importantly, realistic policies. Please also see a presentation regarding our often-quoted Perfume Analysis with our Probabilistic Structural Equations, in which "product" is the relevant break-out variable;
• Target Optimization Tree: While the Target Dynamic Profile function (TDP) returns the optimal policy, the new Optimization Tree feature describes and visualizes a set of alternative optimization policies, using both hard and soft evidence on observations and interventions;
• BayesiaLab Analyst Edition: This is a "player" version of the BayesiaLab Professional Edition, which allows analysts to share Bayesian networks with a broader audience, including managerial decision makers, so they can gain first-hand familiarity with the dynamics contained in a Bayesian network. The Analyst Edition is thus a great way to promote awareness and understanding of the importance of Bayesian networks for decision making.

Click here for more details about the new features of BayesiaLab 5.0

Download a free 30-day evaluation version of BayesiaLab 5.0

Upcoming Bayesian Belief Networks Training in Nashville, Tennessee, USA

We have scheduled our popular three-day training seminar on knowledge modeling and data analysis with Bayesian networks for November 3-5, 2010, in Nashville, Tennessee.

Click here for details about the training and registration information

INFORMS Annual Meeting 2010 in Austin, Texas, USA
Bayesia and its new American partner,  Conrady Applied Science, LLC , will have a show stand at the INFORMS Annual Meeting in Austin, TX, from November 7 through 10. Please make sure to stop by our booth for a one-on-one demo of the full range capabilities of BayesiaLab 5.0.

You can also find this edition of the newsletter, along with all previous issues, on our website.

Best regards,

Dr. Lionel Jouffe

http://www.bayesia.com

Tel.: +33(0)2 43 49 75 69
Fax: +33(0)2 43 49 75 83