Reservoir Computing: Humboldt Fellow researches image processing technologies for improved crop yields

Even when I was at school, my family said: 'You're good at math, you should become a mathematician. But, I felt myself better in physics as I also got the first prize in physics competition in school

says Jimmi H. Talla Mbé: 

I like being in contact with equipment, things I can see and touch and that have a social benefit.

He first came into contact with the Alexander von Humboldt Foundation shortly after studying physics: 

During my studies, my supervisor was the main source of inspiration. He was also a Humboldtian with a great passion for research and science. That made me want to go into research myself in order to make a difference to society.

In 2009, the physicist was awarded the 1st prize of the Alexander von Humboldt College in Cameroon for his academic achievements. After completing his doctorate at the University of Yaoundé I in 2012, his academic career took him to the renowned FEMTO-ST Institute in Besançon, France, as a postdoctoral researcher. He then worked as a teacher at a secondary school before being recruited by the University of Dschang in 2016 as a Lecturer. 

Since 2022, Jimmi H. Talla Mbé has been Associate Professor of Physics at the University of Dschang, where he previously taught and researched as a lecturer and assistant professor for six years. Prof. Talla Mbé dedicates his theoretical and experimental work to the phenomenon of non-linear dynamics - a field of research that connects him with theoretical physicist Prof. Kathy Lüdge: 

This means we investigate systems in which even small changes in the input can lead to large, unexpected changes in the output. Examples of this are systems such as the weather, the human heart or lasers with optical feedback, which, in addition to stable light emission, can also flicker or pulsate chaotically. These seemingly irregular or complex temporal patterns are not ‘errors’, but are often very rich in information," explains Prof. Katy Lüdge. Like Prof. Talla Mbé, she has set herself the task of understanding and using these non-linear dynamics: “It is precisely because non-linear systems react in such a complex way that they can process patterns in data very well.

Computing without computers, clouds, or chatGPT

Prof. Talla Mbé's research focuses on optical and photonic systems such as special electronic and optoelectronic oscillators or semiconductor lasers. By deliberately putting these systems into a state of complex, non-linear dynamics, they can process patterns in data streams and serve as “reservoirs” for so-called reservoir computing – a field of research that also fascinates Prof. Kathy Lüdge:

Reservoir computing is a type of analog computing in which a task is not taught to a digital computer, for example by direct programming, but instead uses the response of a physical system, for example the light intensity of a laser or the deflection of a micromechanical oscillator. For training, the system is optically or electrically perturbed by input data and optimal readout weights are determined. All this can be done in just one step compared to the time-consuming training of neural networks. Complex input data can then be analyzed and classified directly on site. If the data are time series of weather measurement data, the trained physical system, for example the laser, can predict future developments practically instantaneously within certain limits.

Prof. Talla Mbé adds: “The systems thus become a kind of artificial intelligence that can be trained in a similar way to neural networks, but without computers, clouds, or ChatGPT.” Only the output layer is trained, while the actual reservoir - the dynamic system - is used to map structures in the input data:

I find this approach of reservoir computing very fascinating, because instead of using huge computers and clusters, we can perhaps end up building a small physical system that ‘calculates’ naturally and does many times the work of a large computer - and with significantly fewer computing resources and energy.

Prof. Talla Mbé already knows exactly what he would like to use the systems he want so develop for: 

If we manage to build these systems so small, robust and energy-efficient that they could also be used in rural regions without large data centers, we could teach them to monitor the status of agricultural land and plant leaves, for example. Ideally, they would be able to make decisions based on image data that is optically recorded and processed with the help of the energy-efficient system we have trained.

Thus, for example, plant diseases or pest infestations could be detected at an early stage and crop losses prevented. Irrigation systems could also be optimized by predicting water requirements in order to make better use of resources, especially in regions where water is scarce. The research area thus has strong links with the Ilmenau School of Green Electronics (ISGE), which also focuses on resource-saving computing in many projects. 

At TU Ilmenau, Prof. Talla Mbé will be working on a very specific form of reservoir computing, known as delay-based reservoir computing - a method that uses integrated feedback loops to create a kind of memory so that a system can also react to temporal changes in images or analyze video sequences.

Ambassador and mentor for young people 

With this research, Prof. Talla Mbé would also like to serve as an ambassador and communicator between his home country of Cameroon and Germany in the spirit of the Alexander von Humboldt Foundation and promote cross-border exchange between these countries. During the annual conference of the German Society for Applied Optics (DGaO) in June, which he attended together with researchers from TU Ilmenau, he was already able to make new international contacts. 

Above all he is keen to inspire young people himself to address the technological, social and ecological challenges facing our society and to use science to shape a more sustainable and fairer world in line with the Sustainable Development Goals.

He is not only a founder and Fellow of the Cameroon Academy of Young Scientists (CAYS), but also Advisor of the IEEE-Cameroon Student Chapter and founder and former President of the Optica (formerly Optical Society of America) Student Chapter in Cameroon. He has received several awards for his research and his commitment to young scientists, including the African-German Network of Excellence in Science (AGNES) Young Scientist Grant 2012 and the Edmund Optics Educational Award 2014. 

 It's always about role models. You always need someone to light the spark so that an idea can ignite.