Clustering techniques allow to find natural groups inside very large datasets whose internal structure may be forma Partners to further develop the system and/or to establish commercial agreements along with technical cooperation are sought.
The Challenge
Clustering techniques allow to find natural groups inside very large datasets whose internal structure may be formerly unknown, specially when their size makes a manual analysis impractical.
Most clustering techniques have an important limitation, as they can only deal with one form (known as information mode) of the information being analyzed. Our multimodal clustering technique overcomes this limitation, as it can deal with any number of information modes in a single clustering analysis, extracting common patterns from them.
The Technology
It is a natural extension of the spectral clustering algorithm, a state of the art clustering algorithm that allows to cluster similar data not only by distance, but also by contiguity.
This is especially important when analyzing networks of any kind (social, web pages, genes, economic or financial flows, etc.). For example, in an economic study, say company A is related with company B, and B with C. Even if there is no direct connection between A and C, it is important to capture the indirect link between them.
Innovative advantages The algorithm exhibits the following desirable properties:
Scalable: it can be applied to any number of data modes.
Robust: it does not require any adaptation to be applied to new problems.
Mathematically elegant: directly extends spectral clustering algorithm.
Current stage of development
The algorithm and software workflow is developed.
Applications and Target Market
Applications to all kinds of real-life problems: Bioinformatics, Image processing, Social networks analysis, Econometrics, Machine learning Example of application: social networks: Multimedia data, nowadays we face with everywhere, are a good example of multimodal tasks. The information sources for data analysis come from speech, images, tags, social networks, geo-locations etc. For example, we can think of problem on social event detection by looking at photos on http://www.flickr.com with metadata. The challenge is to find all soccer events taking place in Hamburg (Germany) and Madrid (Spain) in the collection (two clusters). Our flexible approach of multimodal spectral clustering llows to aggregate the information from all the sources (photo image, date and location, tags) to perform the task.
Example of application: bioinformatics: Gene to disease relationship is straightforward in exceptional cases. In most cases this relationship is not evident, and therefore it is useful to include as much information as possible in the studies.
In this example two data modes are used. First, a gene interaction network that shows the interactions between genes. Second, a gene expression matrix, that reflects the changes in gene expression inside cells that undergo different experimental conditions. Our technique allows to use both data modes in a single clustering analysis in order to obtain the most useful information from both sources and obtain a set of genes that are both related in terms of genetic interaction and changes in gene expression.
Innovative advantages
The algorithm exhibits the following desirable properties:
Scalable: it can be applied to any number of data modes.
Robust: it does not require any adaptation to be applied to new problems.
Mathematically elegant: directly extends spectral clustering algorithm.
The Universitat Politècnica de Catalunya - BarcelonaTech is a public institution dedicated to higher education and research in the fields of engineering, architecture and science, which contributes its knowledge and expertise in order to increase scientific output, transfer its results to society and provide a network of scientific and technical state-of-the-art facilities and technology valorization services that place us at the leading edge of innovation and economic development.
The UPC has established itself as a driver of innovation and is the technology partner of choice for companies and organizations with which it develops projects and builds partnerships. A role borne out by the numerous agreements and research projects that have been set in motion by groups, organizations and laboratories; the creation of new technology-based companies; the generation and exploitation of patents, and the scientific and technical services UPC makes available to its environment in order to generate progress and employment.
The Technology Transfer Office (SGI) is responsible of Designing, coordinating and implementing research valorisation strategies, carrying out the protection policy of the research results, marketing these results through license contracts and designing and setting up the University's enterprise creation model in order to transfer the results of the research to the market, protect and commercialize these results, promote the culture of entrepreneurship and innovation, and create technology-based companies within the UPC environment.
Create your free account to connect with Universitat Politècnica de Catalunya - UPC and thousands of other innovative organizations and professionals worldwide
Send a request for information
to Universitat Politècnica de Catalunya - UPC
Technology Offers on Innoget are directly posted
and managed by its members as well as evaluation of requests for information. Innoget is the trusted open innovation and science network aimed at directly connect industry needs with professionals online.
Need help requesting additional information or have questions regarding this Technology Offer?
Contact Innoget support
