Multiscale analysis: all levels are important
We have reported earlier that our partner college DevEducation and the leading U.S. research center for complex systems NECSI, headquartered in Boston, USA, have also become partners. As you may recall, the partnership agreement involves, among other things, the following types of actions:
- — mutual partner visits by academic and technical experts;
- — sharing educational programs, materials, research, publications, and academic information;
- — organization of joint research and publications;
- — joint organization of conferences, seminars, and other academic and educational events;
- — joint conduction of technology training sessions and administrative programs.
As a company that welcomes IT talent from DevEducation, we are very excited about this partnership with one of the world’s leaders in new knowledge. All the more so because WizardsDev also makes extensive use of NECSI developments. In November, the President of NECSI, the world-renowned scientist Yaneer Bar-Yam, spoke at a conference organized by WizardsDev in collaboration with DevEducation about complex systems and the interconnectedness of their parts.
Here is another handy text from our partner NECSI — about multiscale analysis.
Complex system sciences are changing the way we think about science and its role in society. This goes beyond the traditional approach of focusing on parts of the system, integrating a network of relationships within and between the systems. These relationships produce the “emergent” behavior that we observe in all physical, biological, social, economic, and technological systems. Such an approach allows researchers to address questions that were once considered inaccessible to science, including human behavior, social interactions, and the consequences of our society’s politics and decisions.
One way to look at a system in its integrity is through multiscale analysis which gathers important information from big data.
Shifting from sickness to health and from economic instability to growth are among the challenges we face today. How can we apply the massive amount of data that is increasingly available to addressing these pressing issues? The data contain a lot of detail but generally do not have tags advising which pieces of information are important in determining successful interventions in the process. The questions we must answer after such an analysis deal with the properties of complex systems: human physiology, global economy. Solving questions about such systems requires unraveling the tangled dependencies and myriad causes and effects of behavior, and recognizing that behavior ranges in scale from the microscopic to the global.
For example, a war can be interpreted as a collision of complex adaptive systems, to describe which the nonlinearity paradigm is used. In complex conflicts and hybrid warfare, the classical analysis of the combat capabilities of armed forces based on an assessment of the fire capabilities of large-scale formations is insufficient and must be supplemented by an assessment of the complexity of both the armed forces themselves and the operational environment as a whole. The analysis is complicated by the fact that scale and complexity are not independent parameters.
Or here is an example from another area. Multiscale maps are dynamic maps that display data differently across a range of scales. They are different from static maps, which are designed to be viewed and displayed at the same scale. You can zoom in and out of the map, but multiscale maps are created in a way that provides visual integrity and effectively communicates information to the user. That is, multiscale maps are the most efficient way to display data virtually continuously across different scales.
The key to solving these problems is to focus on how behavior at different scales is interrelated and how dependencies within the system lead to large-scale behaviors that can be characterized directly without displaying all the complex details. The approach is based on understanding how to aggregate behavior of components to identify larger scale behaviors. In this structure, the information itself has a scale, and the larger scale information is the most critical information you need to know, and the finer scale information is only important to provide details when necessary. This analysis focuses on the information that provides insight on how to affect the behavior of the system on the largest scale. The specific causes and effects to be studied are just a few that underlie the behavior of the system at all levels.
Thus, multiscale analysis is actually a huge simplification compared to the prevalence of traditional approaches. And, if applied correctly, the result is a clear guide on how to intervene and solve key problems in order to get a result that justifies the effort. Successful examples demonstrate that it can be applied to a wide range of scientific and real-world matters, although many aspects of how to proceed more generally are yet to be worked out.
This approach complements many other strategies that are useful for complex systems, including not only big data but also network models, agent-based models, game theory, system dynamics, machine learning, stochastic modeling, coupled differential equations, and other structures in which the starting point is a particular representational structure. In spirit, it is closer to fractals and chaos with its emphasis on the role of scales, but — as with other approaches — it does not adopt their particular representational strategies. In multiscale analysis, the strategy is to describe the largest scale behavior with the smallest but in precise representation, and each of the various representative strategies or combinations thereof can be used as needed.