- Scrambling is a process by which the state of a quantum system is effectively randomized. Scrambling exhibits different complexities depending on the degree of randomness it produces. For example, the complete randomization of a pure quantum state (Haar scrambling) implies the inability to retrieve information of the initial state by measuring only parts of the system (Page/information scrambling), but the converse is not necessarily the case. Here, we formally relate scrambling complexities to the degree of randomness, by studying the behaviors of generalized entanglement entropies -- in particular Rényi entropies -- and their relationship to designs, ensembles of states or unitaries that match the completely random states or unitaries (drawn from the Haar measure) up to certain moments. The main result is that the Rényi-$\alpha$ entanglement entropies, averaged over $\alpha$-designs, are almost maximal. The result generalizes Page's theorem for the von Neumann entropies of small subsystems of random states. For designs of low orders, the average Rényi entanglement entropies can be non-maximal: we exhibit a projective 2-design such that all higher order Rényi entanglement entropies are bounded away from the maximum. However, we show that the Rényi entanglement entropies of all orders are almost maximal for state or unitary designs of order logarithmic in the dimension of the system. That is, such designs are indistinguishable from Haar-random by the entanglement spectrum. Our results establish a formal correspondence between generalized entropies and designs of the same order.
- Mar 24 2017 cond-mat.stat-mech hep-th arXiv:1703.08054v1The flow of the low energy eigenstates of a $U_q[sl(2|1)]$ superspin chain with alternating fundamental ($3$) and dual ($\bar{3}$) representations is studied as function of a twist angle determining the boundary conditions. The finite size spectrum is characterized in terms of scaling dimensions and quasi momenta representing the two families of commuting transfer matrices for the model which are even and odd under the interchange $3\leftrightarrow \bar{3}$, respectively. Varying boundary conditions from periodic to antiperiodic for the fermionic degrees of freedom levels from the continuous part of the finite size spectrum are found to flow into discrete levels and vice versa.
- Synapses in real neural circuits can take discrete values, including zero (silent or potential) synapses. The computational role of zero synapses in unsupervised feature learning of unlabeled noisy data is still unclear, yet important to understand how the sparseness of synaptic activity is shaped during learning and its relationship with receptive field formation. Here, we formulate this kind of sparse feature learning by statistical mechanics approach. We find that learning decreases the fraction of zero synapses, and when the fraction decreases rapidly around a critical data size, the intrinsically structured receptive field starts to develop. Further increasing the data size refines the receptive field, while a very small fraction of zero synapses remain to act as contour detectors. This feature is discovered not only in a handwritten digits dataset, but also in retinal neural activity measured in a natural movie stimuli experiment.
- Mar 24 2017 hep-th cond-mat.stat-mech arXiv:1703.08155v1
- Mar 24 2017 cond-mat.stat-mech arXiv:1703.08126v1
- Mar 24 2017 cond-mat.stat-mech arXiv:1703.07990v1
- Mar 24 2017 cond-mat.stat-mech arXiv:1703.07813v1