On the Parameterization and Initialization of Diagonal State Space Models [PDF]
Neural Information Processing Systems, 2022 State space models (SSM) have recently been shown to be very effective as a deep learning layer as a promising alternative to sequence models such as RNNs, CNNs, or Transformers.
Albert Gu +3 more
semanticscholar +1 more source
Reconfigurable Intelligent Surfaces 2.0: Beyond Diagonal Phase Shift Matrices [PDF]
IEEE Communications Magazine, 2023 Reconfigurable intelligent surface (RIS) has been envisioned as a promising technique to enable and enhance future wireless communications due to its potential to engineer the wireless channels in a cost-effective manner. Extensive research attention has
Hongyu Li +3 more
semanticscholar +1 more source
Saddle-to-saddle dynamics in diagonal linear networks [PDF]
Neural Information Processing Systems, 2023 In this paper we fully describe the trajectory of gradient flow over two-layer diagonal linear networks for the regression setting in the limit of vanishing initialisation.
Scott Pesme, Nicolas Flammarion
semanticscholar +1 more source
SNR Maximization in Beyond Diagonal RIS-Assisted Single and Multiple Antenna Links [PDF]
IEEE Signal Processing Letters, 2023 Reconfigurable intelligent surface (RIS) architectures not limited to diagonal phase shift matrices have recently been considered to increase their flexibility in shaping the wireless channel.
I. Santamaría +3 more
semanticscholar +1 more source
Diagonal State Space Augmented Transformers for Speech Recognition [PDF]
IEEE International Conference on Acoustics, Speech, and Signal Processing, 2023 We improve on the popular conformer architecture by replacing the depthwise temporal convolutions with diagonal state space (DSS) models. DSS is a recently introduced variant of linear RNNs obtained by discretizing a linear dynamical system with a ...
G. Saon, Ankit Gupta, Xiaodong Cui
semanticscholar +1 more source
Beyond Diagonal Reconfigurable Intelligent Surfaces: From Transmitting and Reflecting Modes to Single-, Group-, and Fully-Connected Architectures [PDF]
IEEE Transactions on Wireless Communications, 2022 Reconfigurable intelligent surfaces (RISs) are envisioned as a promising technology for future wireless communications. With various hardware realizations, RISs can work under different modes (reflective/transmissive/hybrid) or have different ...
Hongyu Li, Shanpu Shen, B. Clerckx
semanticscholar +1 more source
Nodal band-off-diagonal superconductivity in twisted graphene superlattices [PDF]
Nature Communications, 2023 The superconducting state and mechanism are among the least understood phenomena in twisted graphene systems. Recent tunneling experiments indicate a transition between nodal and gapped pairing with electron filling, which is not naturally understood ...
Maine Christos, S. Sachdev, M. Scheurer
semanticscholar +1 more source
Reconfigurable Intelligent Surfaces Relying on Non-Diagonal Phase Shift Matrices [PDF]
IEEE Transactions on Vehicular Technology, 2022 Reconfigurable intelligent surfaces (RIS) have been actively researched as a potential technique for future wireless communications, which intelligently ameliorate the signal propagation environment.
Qingchao Li +6 more
semanticscholar +1 more source
Closed-Form Global Optimization of Beyond Diagonal Reconfigurable Intelligent Surfaces [PDF]
IEEE Transactions on Wireless Communications, 2022 Reconfigurable intelligent surfaces (RISs) allow controlling the propagation environment in wireless networks by tuning multiple reflecting elements.
Matteo Nerini, Shanpu Shen, B. Clerckx
semanticscholar +1 more source
Beyond Diagonal Reconfigurable Intelligent Surfaces: A Multi-Sector Mode Enabling Highly Directional Full-Space Wireless Coverage [PDF]
IEEE Journal on Selected Areas in Communications, 2022 Reconfigurable intelligent surface (RIS) has gained much traction due to its potential to manipulate the propagation environment via nearly-passive reconfigurable elements. In our previous work, we have analyzed and proposed a beyond diagonal RIS (BD-RIS)
Hongyu Li, Shanpu Shen, B. Clerckx
semanticscholar +1 more source