Skip to main content

Synaptic multi-factor learning rules: from action potentials to behaviour

Final Report Summary - MULTIRULES (Synaptic multi-factor learning rules: from action potentials to behaviour)

The brain consists of a large network containing millions of neurons where each neuron receives inputs from thousands of other neurons. Learning happens in this network by changes in the connections between neurons.The traditional view, inspired by an early theory of Donald Hebb, an American Psychologist who published his monograph over 60 years ago, highlights the fact that changes in the connection strength are caused by the joint activity of the two neurons that are connected to each other. However, it has been clear from the experimental and theoretical research during the last 20 years that other factors also play a role.
One important aspect is that the change of a given connection can only be driven by information that is locally available at the site of the connection point. In that sense, the Hebbian view makes a valid point: information about the activity of the sending neuron and the current state of the receiving neuron are both indeed available at the site of the connections between the two. We call these the two 'local' factors. However, other factors are also available at the site of the connection, such as the concentration of neurotransmitters. Neurotransmitters are signals that are distributed to thousands or even millions of neurons by fine, but wide-spread ramifications of cables covering large portions of the brain. Therefore information of neurotransmitter activity is shared by many neurons and is, in this sense, more 'global' than the local factors mentioned earlier.

In this project we have developed a framework for changes in the connections that are the result of the interaction of the two local factors (activity of sending and receiving neurons) with one or several global factors. Similar in spirit, but going beyond standard paradigms of reward-based learning, we conceived a framework of learning by surprise as well as consolidation by neuromodulators as specific instances of multi-factor learning rules. The work has been presented in international journals as well as international conference and workshops in front of a multi-disiplinary audience of experimental and theoretical neuroscientists.