Search

The strengthening of synapses between coactivating neurons is believed to be an important underlying mechanism for learning and memory. Hebbian learning of this type has been observed in the brain, with the degree of synaptic strength change dependent on the relative timing of pre-spike arrival and post-spike emission called spike timing dependent plasticity (STDP). Another important feature of learning and memory is the existence of neural spike-timing patterns. Early work by Izhikevich (2006) argued that STDP spontaneously produces structures known as polychronous groups, defined by the network connectivity, that can produce such patterns. However, studies involving STDP face two important issues: how the STDP rule distributes synaptic weights and not knowing what STDP rule is used in the brain. This highlights the importance of understanding the fundamental properties of different STDP rules to determine their effect on the outcome of computational studies. This study focuses on the comparison of two STDP rules, one used by Izhikevich (2006), add-STDP, that produces a bimodal weight distribution, and log-STDP which produces a lognormal weight distribution. The comparison made is between the number of polychronous groups produced and the number of spike-timing patterns, or cell ensembles, found with another detection method that is applicable to experimental data. The number of polychronous groups found with add-STDP was significantly larger as were their sizes and durations. In contrast, the number of cell ensembles found in log-STDP was considerably larger, however, sizes and lifetimes were comparable. Lastly, the activity of cell ensembles in the log-STDP simulations has a non-trivial relationship with the dynamics of synaptic weights in the network, whereas no relationship was found for add-STDP.