Magnus Johnsson

Cognitive Scientist, Computer Scientist





Referee Assignments



Action Recognition


PhD Students

AI Research AB



I'm a cross-disciplinary scientist. Though I have degrees I'm also to a large extent an autodidact. My research interests are wide and include computational modeling, computational intelligence, artificial neural networks, artificial intelligence, computational cognitive neuroscience, cognitive robotics, data science, cognitive architectures, and the philosophy of mind. I have two main goals, to try to understand how brains and cognition work from a systems perspective by computational modelling, and to develop artificial intelligence methods and artificial cognitive architectures. I'm interested in applications of artificial intelligence to solve practical problems, for example medical diagnosis, forecasting and control of robots, and in entrepreneurship/commercialization related to such applications. I'm also curious about consciousness and qualia, as well as in epistemological and other philosophical questions.

I do research on artificial neural networks (ANNs) and other kinds of machine learning, as well as machine learning applications and other applications, e.g. medical diagnosis and data mining. I have a particularly strong interest in self-organizing neural networks, and I have invented some novel variants of the self-organizing map (SOM). One of these, the associative self-organizing map (A-SOM) was invented to be used as a building block (roughly modelling a cortical area in the brain) in artificial cognitive architectures. An A-SOM can learn to associate its self-organized representation of input with arbitrarily many sets of additional input. For example, it can learn to associate its activity with the activity in another A-SOM or with its own activity at one or more earlier moments in time. This enables cross-modal expectations (as for example expecting thunder after seeing a lightning), and what could be seen as mental imagination in an artificial cognitive architecture by a mechanism called internal simulation. The internal simulation hypothesis is related to the mirror neuron theory and was proposed by Germund Hesslow. It proposes that when we imagine that we perceive or act, similar neural activity patterns are elicited in the same brain areas as if we actually had those perceptions or carried out those actions we imagined. The neural activity patterns produced by imagination thus correspond to those elicited had the stimuli/response sequences actually taken place.

I'm very interested in modelling in general for all kinds of applications, in the industry, in medicine, in finance etc. For example, I have been involved in simulating the re-organization of the somatosensory cortex after nerve injuries between the hand and the brain, and in modelling urological dysfunction. I have also done much work on bio-inspired touch perception in robots, and I have designed and built some (now outdated) robots for this aim.