arXiv subjects

R. Venturini

Publications and source records attributed to R. Venturini.

3 recordsLinked to original sources

arXiv2024

Room-temperature memristive switching between charge density wave states

Control over charge density wave (CDW) states is technologically promising for the development of ultra-efficient memory devices. However, using electrical pulses for non-volatile resistance switching involving CDW states has so far been limited to cryogenic temperatures. In this work, we investigate a recently discovered layered semiconductor EuTe4, which exhibits the coexistence of distinct CDW orders. We report that electrical pulses can be used for excitation to non-equilibrium, yet stable electronic states across a broad temperature range from 6 K to 400 K. We find that switching occurs through a non-thermal pathway and is reversible via a thermal erase procedure. The resistance of the new electronic state is tunable by the pulse voltage, so the device acts as a memristor. Calculations show fixed bilayer CDW, whereas CDW in single layers shows bistability due to weak Eu-Te links. Low-voltage, fast, and energy-efficient CDW switching holds potential for memristor applications.

cond-mat.mtrl-sci↗

arXiv2024

From domain walls and the stripe phase to full suppression of charge density wave in the superconducting 1T-Ti$_{1-\text{x}}$Ta$_\text{x}$Se$_2$

1T-TiSe$_2$ hosts a $2 \times 2 \times 2$ charge density wave (CDW) that is known to form the state with localized domains separated by the domain walls upon Cu intercalation. The CDW state with the domain wall network has attracted significant interest due to its coexistence with superconductivity. Here we present a scanning tunneling microscopy, transport and magnetic susceptibility study of 1T-Ti$_{1-\text{x}}$Ta$_\text{x}$Se$_2$. Ta substitution for Ti atoms allows us to perform experiments over the wide range of doping ($ 0 \leqslant \text{x} \leqslant 0.2$), providing access to a significantly broader phase diagram than Cu intercalation experiments. At x = 0.02, we observe a complex network of domains and domain walls. We identify two distinct types of domain walls and show their structure with atomic resolution. Additionally, an elusive symmetry-breaking stripe CDW is found at the light substitution of x = 0.02. We also measure highly substituted x = 0.2 crystals that are superconducting despite the full collapse of the CDW order. Our results uncover rich CDW physics in Ta-substituted 1T-TiSe2 crystals and illuminate the interplay between the CDW and superconductivity.

cond-mat.str-el↗

arXiv2022

Distilled Neural Networks for Efficient Learning to Rank

Recent studies in Learning to Rank have shown the possibility to effectively distill a neural network from an ensemble of regression trees. This result leads neural networks to become a natural competitor of tree-based ensembles on the ranking task. Nevertheless, ensembles of regression trees outperform neural models both in terms of efficiency and effectiveness, particularly when scoring on CPU. In this paper, we propose an approach for speeding up neural scoring time by applying a combination of Distillation, Pruning and Fast Matrix multiplication. We employ knowledge distillation to learn shallow neural networks from an ensemble of regression trees. Then, we exploit an efficiency-oriented pruning technique that performs a sparsification of the most computationally-intensive layers of the neural network that is then scored with optimized sparse matrix multiplication. Moreover, by studying both dense and sparse high performance matrix multiplication, we develop a scoring time prediction model which helps in devising neural network architectures that match the desired efficiency requirements. Comprehensive experiments on two public learning-to-rank datasets show that neural networks produced with our novel approach are competitive at any point of the effectiveness-efficiency trade-off when compared with tree-based ensembles, providing up to 4x scoring time speed-up without affecting the ranking quality.

cs.LG↗