arXiv subjects

Akshat Pandey

Publications and source records attributed to Akshat Pandey.

27 records · Page 2Linked to original sources

arXiv2023

Random geometry at an infinite-randomness fixed point

We study the low-energy physics of the critical (2+1)-dimensional random transverse-field Ising model. The one-dimensional version of the model is a paradigmatic example of a system governed by an infinite-randomness fixed point, for which many results on the distributions of observables are known via an asymptotically exact renormalization group (RG) approach. In two dimensions, the same RG rules have been implemented numerically, and demonstrate a flow to infinite randomness. However, analytical understanding of the structure of this RG has remained elusive due to the development of geometrical structure in the graph of interacting spins. To understand the character of the fixed point, we consider the RG flow acting on a joint ensemble of graphs and couplings. We propose that the RG effectively occurs in two stages: (1) randomization of the interaction graph until it belongs to a certain ensemble of random triangulations of the plane, and (2) a flow of the distributions of couplings to infinite randomness while the graph ensemble remains invariant. This picture is substantiated by a numerical RG in which one obtains a steady-state graph degree distribution and subsequently infinite-randomness scaling distributions of the couplings. Both of these aspects of the RG flow can be approximately reproduced in simplified analytical models.

cond-mat.stat-mech↗

arXiv2023

A stability bound on the $T$-linear resistivity of conventional metals

Perturbative considerations account for the properties of conventional metals, including the range of temperatures where the transport scattering rate is $1/τ_\text{tr} = 2πλT$, where $λ$ is a dimensionless strength of the electron-phonon coupling. The fact that measured values satisfy $λ\lesssim 1$ has been noted in the context of a possible "Planckian" bound on transport. However, since the electron-phonon scattering is quasi-elastic in this regime, no such Planckian considerations can be relevant. We present and analyze Monte Carlo results on the Holstein model which show that a different sort of bound is at play: a "stability" bound on $λ$ consistent with metallic transport. We conjecture that a qualitatively similar bound on the strength of residual interactions, which is often stronger than Planckian, may apply to metals more generally.

cond-mat.str-el↗

arXiv2022

What the DAAM: Interpreting Stable Diffusion Using Cross Attention

Large-scale diffusion neural networks represent a substantial milestone in text-to-image generation, but they remain poorly understood, lacking interpretability analyses. In this paper, we perform a text-image attribution analysis on Stable Diffusion, a recently open-sourced model. To produce pixel-level attribution maps, we upscale and aggregate cross-attention word-pixel scores in the denoising subnetwork, naming our method DAAM. We evaluate its correctness by testing its semantic segmentation ability on nouns, as well as its generalized attribution quality on all parts of speech, rated by humans. We then apply DAAM to study the role of syntax in the pixel space, characterizing head--dependent heat map interaction patterns for ten common dependency relations. Finally, we study several semantic phenomena using DAAM, with a focus on feature entanglement, where we find that cohyponyms worsen generation quality and descriptive adjectives attend too broadly. To our knowledge, we are the first to interpret large diffusion models from a visuolinguistic perspective, which enables future lines of research. Our code is at https://github.com/castorini/daam.

cs.CV↗

arXiv2022

Strange-metal behavior in a Fermi liquid with strange scatterers

A variety of low-temperature, normal-state properties of optimally and overdoped cuprate superconductors, including the DC and optical transport responses, are sufficiently anomalous that they might seem to be inconsistent with any quasiparticle description. However, we show by explicit construction that the most salient phenomena can be accounted for in a system with localized "strange scatterers" embedded into a Fermi liquid with a conventional quasiparticle description. Such scatterers could originate from "two-level systems" associated with an electronic glassy state with short-range charge-order correlations.

cond-mat.str-el↗

arXiv2022

SpeechNet: Weakly Supervised, End-to-End Speech Recognition at Industrial Scale

End-to-end automatic speech recognition systems represent the state of the art, but they rely on thousands of hours of manually annotated speech for training, as well as heavyweight computation for inference. Of course, this impedes commercialization since most companies lack vast human and computational resources. In this paper, we explore training and deploying an ASR system in the label-scarce, compute-limited setting. To reduce human labor, we use a third-party ASR system as a weak supervision source, supplemented with labeling functions derived from implicit user feedback. To accelerate inference, we propose to route production-time queries across a pool of CUDA graphs of varying input lengths, the distribution of which best matches the traffic's. Compared to our third-party ASR, we achieve a relative improvement in word-error rate of 8% and a speedup of 600%. Our system, called SpeechNet, currently serves 12 million queries per day on our voice-enabled smart television. To our knowledge, this is the first time a large-scale, Wav2vec-based deployment has been described in the academic literature.

cs.CL↗

arXiv2022

Interstitial-induced ferromagnetism in a two-dimensional Wigner crystal

The two-dimensional Wigner crystal (WC) occurs in the strongly interacting regime ($r_s \gg 1$) of the two-dimensional electron gas (2DEG). The magnetism of a pure WC is determined by tunneling processes that induce multi-spin ring-exchange interactions, resulting in fully polarized ferromagnetism for large enough $r_s$. Recently, Hossain et al. [PNAS 117 (51) 32244-32250] reported the occurrence of a fully polarized ferromagnetic insulator at $r_s \gtrsim 35$ in an AlAs quantum well, but at temperatures orders of magnitude larger than the predicted exchange energies for the pure WC. Here, we analyze the large $r_s$ dynamics of an interstitial defect in the WC, and show that it produces local ferromagnetism with much higher energy scales. Three hopping processes are dominant, which favor a large, fully polarized ferromagnetic polaron. Based on the above results, we speculate concerning the phenomenology of the magnetism near the metal-insulator transition of the 2DEG.

cond-mat.str-el↗

arXiv2021

Disparate Impact of Artificial Intelligence Bias in Ridehailing Economy's Price Discrimination Algorithms

Ridehailing applications that collect mobility data from individuals to inform smart city planning predict each trip's fare pricing with automated algorithms that rely on artificial intelligence (AI). This type of AI algorithm, namely a price discrimination algorithm, is widely used in the industry's black box systems for dynamic individualized pricing. Lacking transparency, studying such AI systems for fairness and disparate impact has not been possible without access to data used in generating the outcomes of price discrimination algorithms. Recently, in an effort to enhance transparency in city planning, the city of Chicago regulation mandated that transportation providers publish anonymized data on ridehailing. As a result, we present the first large-scale measurement of the disparate impact of price discrimination algorithms used by ridehailing applications. The application of random effects models from the meta-analysis literature combines the city-level effects of AI bias on fare pricing from census tract attributes, aggregated from the American Community Survey. An analysis of 100 million ridehailing samples from the city of Chicago indicates a significant disparate impact in fare pricing of neighborhoods due to AI bias learned from ridehailing utilization patterns associated with demographic attributes. Neighborhoods with larger non-white populations, higher poverty levels, younger residents, and high education levels are significantly associated with higher fare prices, with combined effect sizes, measured in Cohen's d, of -0.32, -0.28, 0.69, and 0.24 for each demographic, respectively. Further, our methods hold promise for identifying and addressing the sources of disparate impact in AI algorithms learning from datasets that contain U.S. geolocations.

cs.CY↗

arXiv2020

Automatically Characterizing Targeted Information Operations Through Biases Present in Discourse on Twitter

This paper considers the problem of automatically characterizing overall attitudes and biases that may be associated with emerging information operations via artificial intelligence. Accurate analysis of these emerging topics usually requires laborious, manual analysis by experts to annotate millions of tweets to identify biases in new topics. We introduce extensions of the Word Embedding Association Test from Caliskan et al. to a new domain (Caliskan, 2017). Our practical and unsupervised method is used to quantify biases promoted in information operations. We validate our method using known information operation-related tweets from Twitter's Transparency Report. We perform a case study on the COVID-19 pandemic to evaluate our method's performance on non-labeled Twitter data, demonstrating its usability in emerging domains.

cs.CY↗

arXiv2020

Analytical theory of pyrochlore cooperative paramagnets

The pyrochlore lattice is associated with several potential and actual spin liquid phases as a result of its strong geometric frustration. At finite temperature, these can exhibit an unusually broad cross-over regime to a conventional paramagnet. Here, we study this regime analytically by showing how a single-tetrahedron Hamiltonian can extrapolate beyond the first term of a high-temperature expansion and yield insights into the build-up of correlations. We discuss how this unusual behaviour is brought about by the structure of the eigenspaces of the coupling matrix. Further interesting behaviour can appear for parameter values located near phase transitions: we find coexistence of $(111)$ rods and $(220)$ peaks in the structure factor, as observed in neutron scattering experiments on Yb$_2$Ti$_2$O$_7$.

cond-mat.str-el↗