Search for: All records

Abstract Novel stimulation protocols for neuromodulation with magnetic fields are explored in clinical and laboratory settings. Recent evidence suggests that the activation state of the nervous system plays a significant role in the outcome of magnetic stimulation, but the underlying cellular and molecular mechanisms of state-dependency have not been completely investigated. We recently reported that high frequency magnetic stimulation could inhibit neural activity when the neuron was in a low active state. In this paper, we investigate state-dependent neural modulation by applying a magnetic field to single neurons, using the novel micro-coil technology. High frequency magnetic stimulation suppressed single neuron activity in a state-dependent manner. It inhibited neurons in slow-firing states, but spared neurons from fast-firing states, when the same magnetic stimuli were applied. Using a multi-compartment NEURON model, we found that dynamics of voltage-dependent sodium and potassium channels were significantly altered by the magnetic stimulation in the slow-firing neurons, but not in the fast-firing neurons. Variability in neural activity should be monitored and explored to optimize the outcome of magnetic stimulation in basic laboratory research and clinical practice. If selective stimulation can be programmed to match the appropriate neural state, prosthetic implants and brain-machine interfaces can be designed based on these concepts to achieve optimal results. 

Not all people are equally easy to identify: color statistics might be enough for some cases while others might re- quire careful reasoning about high- and low-level details. However, prevailing person re-identification(re-ID) meth- ods use one-size-fits-all high-level embeddings from deep convolutional networks for all cases. This might limit their accuracy on difficult examples or makes them needlessly ex- pensive for the easy ones. To remedy this, we present a new person re-ID model that combines effective embeddings built on multiple convolutional network layers, trained with deep-supervision. On traditional re-ID benchmarks, our method improves substantially over the previous state-of- the-art results on all five datasets that we evaluate on. We then propose two new formulations of the person re- ID problem under resource-constraints, and show how our model can be used to effectively trade off accuracy and computation in the presence of resource constraints.