Narrow Results

We develop a new approach to the design of neural networks, which utilizes a collaborative framework of knowledge-driven experience. In contrast to the "standard" way of developing neural networks, which explicitly exploits experimental data, this approach incorporates a mechanism of knowledge-driven experience. The essence of the proposed scheme of learning is to take advantage of the parameters (connections) of neural networks built in the past for the same phenomenon (which might also exhibit some variability over time or space) for which are interested to construct the network on a basis of currently available data. We establish a conceptual and algorithmic framework to reconcile these two essential sources of information (data and knowledge) in the process of the development of the network. To make a presentation more focused and come up with a detailed quantification of the resulting architecture, we concentrate on the experience-based design of radial basis function neural networks (RBFNNs). We introduce several performance indexes to quantify an effect of utilization of the knowledge residing within the connections of the networks and establish an optimal level of their use. Experimental results are presented for low-dimensional synthetic data and selected datasets available at the Machine Learning Repository.

Recent studies have shown that it is feasible to record simultaneously intracerebral EEG (icEEG) and functional magnetic resonance imaging (fMRI) in patients with epilepsy. While it has mainly been used to explore the hemodynamic changes associated with epileptic spikes, this approach could also provide new insight into human cognition. However, the first step is to ensure that cognitive EEG components, that have lower amplitudes than epileptic spikes, can be appropriately detected under fMRI. We compared the high frequency activities (HFA, 50–150Hz) elicited by a reading task in icEEG-only and subsequent icEEG-fMRI in the same patients ($n=3$), implanted with depth electrodes. Comparable responses were obtained, with 71% of the recording sites that responded during the icEEG-only session also responding during the icEEG-fMRI session. For all the remaining sites, nearby clusters (distant of 7mm or less) also demonstrated significant HFA increase during the icEEG-fMRI session. Significant HFA increases were also observable at the single-trial level in icEEG-fMRI recordings. Our results show that low-amplitude icEEG signal components such as cognitive-induced HFAs can be reliably recorded with simultaneous fMRI. This paves the way for the use of icEEG-fMRI to address various fundamental and clinical issues, notably the identification of the neural correlates of the BOLD signal.