On 11th of June, Blerim Emruli will present his licentiate thesis addressing a set of simple computational principles with cognitive qualities, which can enable computers to function more like brains.
Today, more data is produced than can be physically stored. Additionally, only a small percentage of the information being produced is “structured” and therefore easy to process with the help of computers. The rest is so called “unstructured” information such as text, voice, video, and complex sensor data.
From this perspective, Blerim Emruli’s licentiate thesis raises a number of key challenges and opportunities. For example, how can computers:
- process and archive useful information, while discarding useless data?
- process text and other non-numeric information in a meaningful way?
- process and exploit the information streaming from large heterogeneous sensor systems on the internet?
- detect and predict events quickly and accurately?
Towards more intelligent computers
In order to find answers to these challenging questions, Blerim Emruli is studying and developing mathematical models of cognitive functions. These models can enable programmers to represent and process information in a new way, using data structures and operators that are 10 000 bits instead of the 32 or 64 bits used today.
– The human brain handles most of the above questions effortlessly. This is the main motivation behind my work, says Blerim Emruli.
– Furthermore, we want computers to understand and function more like the human brain. For example, when we Google for something, we do not want Google to simply process keywords and link us to matching web pages, but rather we would like Google to process the “meaning” of the query and answer in a meaningful way. And if needed to suggest additional information which we did not address in our initial query.
Better results with new principles
In his thesis, Blerim Emruli presents a set of simple computational principles with cognitive qualities, which are computationally efficient, distributable and scalable; and thus suitable for networked sensor systems. These systems are often resource-constrained in terms of battery, computing and communication capacity.
Blerim Emruli is a Ph.D. student at EISLAB within the Department of Computer Science, Electrical and Space Engineering at the Luleå University of Technology. The name of his licentiate thesis is Simple principles of cognitive computation with distributed representations.
Contact Blerim Emruli