arXiv subjects

Felipe Torres

Publications and source records attributed to Felipe Torres.

18 recordsLinked to original sources

arXiv2024

CA-Stream: Attention-based pooling for interpretable image recognition

Explanations obtained from transformer-based architectures in the form of raw attention, can be seen as a class-agnostic saliency map. Additionally, attention-based pooling serves as a form of masking the in feature space. Motivated by this observation, we design an attention-based pooling mechanism intended to replace Global Average Pooling (GAP) at inference. This mechanism, called Cross-Attention Stream (CA-Stream), comprises a stream of cross attention blocks interacting with features at different network depths. CA-Stream enhances interpretability in models, while preserving recognition performance.

cs.CV↗

arXiv2024

Steering spin fluctuations in lattice systems via two-tone Floquet engineering

We report on the control of spin pair fluctuations using two-tone Floquet engineering. We consider a one-dimensional spin-1/2 lattice with periodically modulated spin exchanges using parametric resonances. The stroboscopic dynamics generated from distributed spin exchange modulations lead to spin pair fluctuations reaching quasi-maximally correlated states and a subharmonic response in local observables, breaking the discrete-time translational symmetry. We present a protocol to control the interacting many-body dynamics, producing spatial and temporal localization of correlated spin pairs via dynamically breaking correlated spin pairs from the edges towards the center of the lattice. Our result reveals how spin fluctuations distribute in a heterogeneous lattice depending on parametric resonances. This may open new routes for exploring distinct nonequilibrium states of matter and the conduction of quasiparticles in quantum materials.

quant-ph↗

arXiv2023

Disentangling transport mechanisms in a correlated oxide by photoinduced charge injection

We present a novel heterostructured approach to disentangle the mechanism of electrical transport of the strongly correlated PrNiO3, by placing the nickelate under the photoconductor CdS. This enables the injection of carriers into PrNiO3 in a controlled way, which can be used to interrogate its intrinsic transport mechanism. We find a non-volatile resistance decrease when illuminating the system at temperatures below the PrNiO3 metal-insulator transition. The photoinduced change becomes more volatile as the temperature increases. These data help understand the intrinsic transport properties of the nickelate-CdS bilayer. Together with data from a bare PrNiO3 film, we find that the transport mechanism includes a combination of mechanisms including both thermal activation and variable range hopping. At low temperatures without photoinduced carriers the transport is governed by hopping, while at higher temperatures and intense illumination the activation mechanism becomes relevant. This work shows a new way to optically control the low-temperature resistance of PrNiO3.

cond-mat.str-el↗

arXiv2023

Light induced decoupling of electronic and magnetic properties in manganites

The strongly correlated material La0.7Sr0.3MnO3 (LSMO) exhibits metal-to-insulator and magnetic transition near room temperature. Although the physical properties of LSMO can be manipulated by strain, chemical doping, temperature, or magnetic field, they often require large external stimuli. To include additional flexibility and tunability, we developed a hybrid optoelectronic heterostructure that uses photocarrier injection from cadmium sulfide (CdS) to an LSMO layer to change its electrical conductivity. LSMO exhibits no significant optical response, however, the CdS/LSMO heterostructures show an enhanced conductivity, with ~ 37 % resistance drop, at the transition temperature under light stimuli. This enhanced conductivity in response to light is comparable to the effect of a 9 T magnetic field in pure LSMO. Surprisingly, the optical and magnetic responses of CdS/LSMO heterostructures are decoupled and exhibit different effects when both stimuli are applied. This unexpected behavior shows that heterostructuring strongly correlated oxides may require a new understanding of the coupling of physical properties across the transitions and provide the means to implement new functionalities.

cond-mat.mtrl-sci↗

arXiv2023

Opti-CAM: Optimizing saliency maps for interpretability

Methods based on class activation maps (CAM) provide a simple mechanism to interpret predictions of convolutional neural networks by using linear combinations of feature maps as saliency maps. By contrast, masking-based methods optimize a saliency map directly in the image space or learn it by training another network on additional data. In this work we introduce Opti-CAM, combining ideas from CAM-based and masking-based approaches. Our saliency map is a linear combination of feature maps, where weights are optimized per image such that the logit of the masked image for a given class is maximized. We also fix a fundamental flaw in two of the most common evaluation metrics of attribution methods. On several datasets, Opti-CAM largely outperforms other CAM-based approaches according to the most relevant classification metrics. We provide empirical evidence supporting that localization and classifier interpretability are not necessarily aligned.

cs.CV↗

arXiv2021

Polariton-based quantum memristors

Information processing and storing by the same physical system is emerging as a promising alternative to traditional computing platforms. In turn, this requires the realization of elementary units whose memory content can be easily tuned and controlled. Here, we introduce a polariton-based quantum memristor where the memristive nature arises from the inter-cavity polariton exchange and is controlled by a time-varying atom-cavity detuning. A dynamical hysteresis is characterized by the fluctuations in the instantaneous polariton number, where the history information is encoded into a dynamical phase. Using a Lindblad master equation approach, we find that features of the quantum memristor dynamics, such as the area and circulation of the hysteresis loop, showcase a kind of "plasticity" controlled by quantum state initialization. This makes this quantum memristor very versatile for a wide range of applications

quant-ph↗

arXiv2020

Steering Interchange of Polariton Branches via Coherent and Incoherent Dynamics

Controlling light-matter based quantum systems in the strong coupling regime allows for exploring quantum simulation of many-body physics in nowadays architectures. For instance, the atom-field interaction in a cavity QED network provides control and scalability for quantum information processing. Here, we propose the control of single- and two-body Jaynes-Cummings systems in a non-equilibrium scenario, which allows us to establish conditions for the coherent and incoherent interchange of polariton branches. Our findings provide a systematic approach to manipulate polaritons interchange, that we apply to reveal new insights in the transition between Mott Insulator- and Superfluid-like states. Furthermore, we study the asymmetry in the absorption spectrum by triggering the cavity and atomic losses as a function of the atom-cavity detuning and photon's hopping.

quant-ph↗

arXiv2020

SIMBA: Specific Identity Markers for Bone Age Assessment

Bone Age Assessment (BAA) is a task performed by radiologists to diagnose abnormal growth in a child. In manual approaches, radiologists take into account different identity markers when calculating bone age, i.e., chronological age and gender. However, the current automated Bone Age Assessment methods do not completely exploit the information present in the patient's metadata. With this lack of available methods as motivation, we present SIMBA: Specific Identity Markers for Bone Age Assessment. SIMBA is a novel approach for the task of BAA based on the use of identity markers. For this purpose, we build upon the state-of-the-art model, fusing the information present in the identity markers with the visual features created from the original hand radiograph. We then use this robust representation to estimate the patient's relative bone age: the difference between chronological age and bone age. We validate SIMBA on the Radiological Hand Pose Estimation dataset and find that it outperforms previous state-of-the-art methods. SIMBA sets a trend of a new wave of Computer-aided Diagnosis methods that incorporate all of the data that is available regarding a patient. To promote further research in this area and ensure reproducibility we will provide the source code as well as the pre-trained models of SIMBA.

eess.IV↗

arXiv2019

Dynamical dimerization phase in Jaynes-Cummings lattices

We report on an emergent dynamical phase of a strongly-correlated light-matter system, which is governed by dimerization processes due to short-range and long-range two-body interactions. The dynamical phase is characterized by the spontaneous symmetry breaking of the translational invariance and appears in an intermediate regime of light-matter interaction between the resonant and dispersive cases. We describe the quench dynamics from an initial state with integer filling factor of a finite-sized array of coupled resonators, each doped with a two-level system, in a closed and open scenario. The closed system dynamics has an effective Hilbert space description that allows us to demonstrate and characterize the emergent dynamical phase via time-averaged quantities, such as fluctuations in the number of polaritons per site and linear entropy. We prove that the dynamical phase is governed by intrinsic two-body interactions and the lattice topological structure. In the open system dynamics, we show evidence about the robustness of dynamical dimerization processes under loss mechanisms. Our findings can be used to determine the light-matter detuning range, where the dimerized phase emerges.

quant-ph↗

arXiv2018

Nucleation of superfluid-light domains in a quenched dynamics

Strong correlation effects emerge from light-matter interactions in coupled resonator arrays, such as the Mott-insulator to superfluid phase transition of atom-photon excitations. We demonstrate that the quenched dynamics of a finite-sized complex array of coupled resonators induces a first-order like phase transition. The latter is accompanied by domain nucleation that can be used to manipulate the photonic transport properties of the emerging superfluid phase; this in turn leads to an empirical scaling law. This universal behavior emerges from the light-matter interaction and the topology of the array. The validity of our results over a wide range of complex architectures might lead to to a promising device for use in scaled quantum simulations.

quant-ph↗

arXiv2015

Network complexity and topological phase transitions

A new type of collective excitations, due exclusively to the topology of a complex random network that can be characterized by a fractal dimension $D_F$, is investigated. We show analytically that these excitations generate phase transitions due to the non-periodic topology of the $D_F>1$ complex network. An Ising system, with long range interactions over such a network, is studied in detail to support the claim. The analytic treatment is possible because the evaluation of the partition function can be decomposed into closed factor loops, in spite of the architectural complexity. This way we compute the magnetization distribution, magnetization loops, and the two point correlation function; and relate them to the network topology. In summary, the removal of the infrared divergences leads to an unconventional phase transition, where spin correlations are robust against thermal fluctuations.

cond-mat.stat-mech↗

arXiv2014

Dzyaloshinskii-Moriya interaction and magnetic ordering in 1D and 2D at nonzero $T$

The inclusion of a Dzyaloshinskii-Moriya short range antisymmetric interaction induces spontaneous magnetization, at nonzero temperatures, in one and two dimensions. It is shown that quantum fluctuations are reduced by the Dzyaloshinskii-Moriya interaction, but short range correlations are increased, thereby allowing to obtain long range magnetic order in these low dimensional systems.

cond-mat.other↗

arXiv2013

On hitting times for simple random walk on dense Erd\"os-R\'enyi random graphs

Let $G(N,p)=(V,E)$ be an Erd\"os-R\'enyi random graph and $(X_n)_{n \in \mathbb{N}}$ be a simple random walk on it. We study the the order of magnitude of $\sum_{i \in V} \pi_ih_{ij} $ where $\pi_i=d_i / 2|E|$ for $d_i$ the number of neighbors of node $i$ and $h_{ij}$ the hitting time for $(X_n)_{n \in \mathbb{N}}$ between nodes $i$ and $j$, in a regime of $p=p(N)$ such that $G(N,p)$ is almost surely connected as $N\to\infty$. Our main result is that $\sum_{i \in V} \pi_ih_{ij} $ is almost surely of order $N(1+o(1))$ as $N\to \infty$, which coincides with previous results in the physics literature \cite{sood}, though our techniques are based on large deviations bounds on the number of neighbors of a typical node and the number of edges in $G(N,p)$ together with recent work on bounds on the spectrum of the (random) adjacency matrix of $G(N,p)$.

math.PR↗

arXiv2012

General approach to the fluctuations problem in random sequence comparison

We present a general approach to the problem of determining the asymptotic order of the variance of the optimal score between two independent random sequences defined over an arbitrary finite alphabet. Our general approach is based on identifying random variables driving the fluctuations of the optimal score and conveniently choosing functions of them which exhibit certain monotonicity properties. We show how our general approach establishes a common theoretical background for the techniques used by Matzinger et al. in a series of previous articles [6, 8, 20, 24, 26, 37] studying the same problem in especial cases. Additionally, we explicitely apply our general approach to study the fluctuations of the optimal score between two random sequences over a finite alphabet (closing the study as initiated in [26]) and of the length of the longest common subsequences between two random sequences with a certain block structure (generalizing part of [37]).

math.PR↗

arXiv2011

Large deviations principle for Curie-Weiss models with random fields

In this article we consider an extension of the classical Curie-Weiss model in which the global and deterministic external magnetic field is replaced by local and random external fields which interact with each spin of the system. We prove a Large Deviations Principle for the so-called {\it magnetization per spin} $S_n/n$ with respect to the associated Gibbs measure, where $S_n/n$ is the scaled partial sum of spins. In particular, we obtain an explicit expression for the LDP rate function, which enables an extensive study of the phase diagram in some examples. It is worth mentioning that the model considered in this article covers, in particular, both the case of i.\,i.\,d.\ random external fields (also known under the name of random field Curie-Weiss models) and the case of dependent random external fields generated by e.\,g.\ Markov chains or dynamical systems.

math.PR↗

arXiv2010

Random modification effect in the size of the fluctuation of the LCS of two sequences of i.i.d. blocks

The problem of the order of the fluctuation of the Longest Common Subsequence (LCS) of two independent sequences has been open for decades. There exist contradicting conjectures on the topic, due to Chvatal - Sankoff in 1975 and Waterman in 1994. In the present article, we consider a special model of i.i.d. sequences made out of blocks. A block is a contiguous substring consisting only of one type of symbol. Our model allows only three possible block lengths, each been equiprobable picked up. In this context, we introduce a random operation (random modification) on the blocks of one of the sequences. In the present article, we develop the techniques to prove the following: if we suppose that the random modification increases the length of the LCS with high probability, then the order of the fluctuation of the LCS is as conjectured by Waterman. This result is a key technical part in the study of the size of the fluctuation of the LCS for sequences of i.i.d. blocks, developed by Matzinger and Torres.

math.PR↗

arXiv2010

The rate of the convergence of the mean score in random sequence comparison

We consider a general class of super-additive scores measuring the similarity of two independent sequences of $n$ i.i.d. letters from a finite alphabet. Our object of interest is the mean score by letter $l_n$. By the subadditivity $l_n$ is nondecreasing and converges to a limit $l$. We give a simple method of bounding the difference $l-l_n$ and obtaining the rate of convergence. Our result generalizes a previous result of Alexander, where only the special case of the longest common subsequence is considered.

math.PR↗

arXiv2010

Fluctuations of the Longest Common Subsequence for Sequences of Independent Blocks

The problem of the fluctuation of the Longest Common Subsequence (LCS) of two i.i.d. sequences of length $n>0$ has been open for decades. There exist contradicting conjectures on the topic. Chvatal and Sankoff conjectured in 1975 that asymptotically the order should be $n^{2/3}$, while Waterman conjectured in 1994 that asymptotically the order should be $n$. A contiguous substring consisting only of one type of symbol is called a block. In the present work, we determine the order of the fluctuation of the LCS for a special model of sequences consisting of i.i.d. blocks whose lengths are uniformly distributed on the set $\{l-1,l,l+1\}$, with $l$ a given positive integer. We showed that the fluctuation in this model is asymptotically of order $n$, which confirm Waterman's conjecture. For achieving this goal, we developed a new method which allows us to reformulate the problem of the order of the variance as a (relatively) low dimensional optimization problem.

math.PR↗