arXiv subjects

Anil Seth

Publications and source records attributed to Anil Seth.

At least 37 records · Page 2Linked to original sources

arXiv2023

The Chemodynamics of the Stellar Populations in M31 from APOGEE Integrated Light Spectroscopy

We present analysis of nearly 1,000 near-infrared, integrated light spectra from APOGEE in the inner $\sim$7 kpc of M31. We utilize full spectrum fitting with A-LIST simple stellar population spectral templates that represent a population of stars with the same age, [M/H], and [$α$/M]. With this, we determine the mean kinematics, metallicities, $α$ abundances, and ages of the stellar populations of M31's bar, bulge, and inner disk ($\sim$4-7 kpc). We find a non-axisymmetric velocity field in M31 resulting from the presence of a bar. The bulge of M31 is metal-poor relative to the disk ([M/H] = $-0.149^{+0.067}_{-0.081}$ dex), features minima in metallicity on either side of the bar ([M/H] $\sim$ -0.2), and is enhanced in $α$ abundance ([$α$/M] = $0.281^{+0.035}_{-0.038}$). The disk of M31 within $\sim$7 kpc is enhanced in both metallicity ([M/H] = $-0.023^{+0.050}_{-0.052}$) and $α$ abundance ([$α$/M] = $0.274^{+0.020}_{-0.025}$). Both of these structural components are uniformly old at $\simeq$ 12 Gyr. We find the metallicity increases with distance from the center of M31, with the steepest gradient along the disk major axis ($0.043\pm0.021$ dex/kpc). This gradient is the result of changing light contributions from the metal-poor bulge and metal-rich disk. The chemodynamics of stellar populations encodes information about a galaxy's chemical enrichment, star formation history, and merger history, allowing us to discuss new constraints on M31's formation. Our results provide a stepping stone between our understanding of the Milky Way and other external galaxies.

astro-ph.GA↗

arXiv2022

Predictive Coding: a Theoretical and Experimental Review

Predictive coding offers a potentially unifying account of cortical function -- postulating that the core function of the brain is to minimize prediction errors with respect to a generative model of the world. The theory is closely related to the Bayesian brain framework and, over the last two decades, has gained substantial influence in the fields of theoretical and cognitive neuroscience. A large body of research has arisen based on both empirically testing improved and extended theoretical and mathematical models of predictive coding, as well as in evaluating their potential biological plausibility for implementation in the brain and the concrete neurophysiological and psychological predictions made by the theory. Despite this enduring popularity, however, no comprehensive review of predictive coding theory, and especially of recent developments in this field, exists. Here, we provide a comprehensive review both of the core mathematical structure and logic of predictive coding, thus complementing recent tutorials in the literature. We also review a wide range of classic and recent work within the framework, ranging from the neurobiologically realistic microcircuits that could implement predictive coding, to the close relationship between predictive coding and the widely-used backpropagation of error algorithm, as well as surveying the close relationships between predictive coding and modern machine learning techniques.

cs.AI↗

arXiv2022

Young, blue, and isolated stellar systems in the Virgo Cluster. II. A new class of stellar system

We discuss five blue stellar systems in the direction of the Virgo cluster, analogous to the enigmatic object SECCO 1 (AGC 226067). These objects were identified based on their optical and UV morphology and followed up with HI observations with the VLA (and GBT), MUSE/VLT optical spectroscopy, and HST imaging. These new data indicate that one system is a distant group of galaxies. The remaining four are extremely low mass ($M_\ast \sim 10^5 \; \mathrm{M_\odot}$), are dominated by young, blue stars, have highly irregular and clumpy morphologies, are only a few kpc across, yet host an abundance of metal-rich, $12 + \log (\mathrm{O/H}) > 8.2$, HII regions. These high metallicities indicate that these stellar systems formed from gas stripped from much more massive galaxies. Despite the young age of their stellar populations, only one system is detected in HI, while the remaining three have minimal (if any) gas reservoirs. Furthermore, two systems are surprisingly isolated and have no plausible parent galaxy within $\sim$30' ($\sim$140 kpc). Although tidal stripping cannot be conclusively excluded as the formation mechanism of these objects, ram pressure stripping more naturally explains their properties, in particular their isolation, owing to the higher velocities, relative to the parent system, that can be achieved. Therefore, we posit that most of these systems formed from ram pressure stripped gas removed from new infalling cluster members, and survived in the intracluster medium long enough to become separated from their parent galaxies by hundreds of kiloparsecs, and that they thus represent a new type of stellar system.

astro-ph.GA↗

arXiv2022

AGC 226178 and NGVS 3543: Two deceptive dwarfs towards Virgo

The two sources AGC 226178 and NGVS 3543, an extremely faint, clumpy, blue stellar system and a low surface brightness dwarf spheroidal, are adjacent systems in the direction of the Virgo cluster. Both have been studied in detail previously, with it being suggested that they are unrelated normal dwarf galaxies or that NGVS 3543 recently lost its gas through ram pressure stripping, and that AGC 226178 formed from this stripped gas. However, with HST ACS imaging we demonstrate that the stellar population of NGVS 3543 is inconsistent with being at the distance of the Virgo cluster, and that it is likely a foreground object at approximately 10 Mpc. Whereas the stellar population of AGC 226178 is consistent with it being a very young (10-100 Myr) object in the Virgo cluster. Through a re-analysis of the original ALFALFA HI detection we show that AGC 226178 likely formed from gas stripped from the nearby dwarf galaxy VCC 2034, a hypothesis strengthened by the high metallicity measured with MUSE VLT observations. However, it is unclear whether ram pressure or a tidal interaction is responsible for stripping the gas. AGC 226178 is one of at least five similar objects now known towards Virgo. These objects are all young and unlikely to remain visible for over ~500 Myr, suggesting that they are continually produced in the cluster.

astro-ph.GA↗

arXiv2022

A continuity of Markov blanket interpretations under the Free Energy Principle

Bruineberg and colleagues helpfully distinguish between instrumental and ontological interpretations of Markov blankets, exposing the dangers of using the former to make claims about the latter. However, proposing a sharp distinction neglects the value of recognising a continuum spanning from instrumental to ontological. This value extends to the related distinction between being and having a model.

q-bio.NC↗

arXiv2021

Understanding the Origin of Information-Seeking Exploration in Probabilistic Objectives for Control

The exploration-exploitation trade-off is central to the description of adaptive behaviour in fields ranging from machine learning, to biology, to economics. While many approaches have been taken, one approach to solving this trade-off has been to equip or propose that agents possess an intrinsic 'exploratory drive' which is often implemented in terms of maximizing the agents information gain about the world -- an approach which has been widely studied in machine learning and cognitive science. In this paper we mathematically investigate the nature and meaning of such approaches and demonstrate that this combination of utility maximizing and information-seeking behaviour arises from the minimization of an entirely difference class of objectives we call divergence objectives. We propose a dichotomy in the objective functions underlying adaptive behaviour between \emph{evidence} objectives, which correspond to well-known reward or utility maximizing objectives in the literature, and \emph{divergence} objectives which instead seek to minimize the divergence between the agent's expected and desired futures, and argue that this new class of divergence objectives could form the mathematical foundation for a much richer understanding of the exploratory components of adaptive and intelligent action, beyond simply greedy utility maximization.

cs.LG↗

arXiv2021

Detection of a $\sim$100,000 M$_\odot$ black hole in M31's most massive globular cluster: A tidally stripped nucleus

We investigate the presence of a central black hole (BH) in B023-G078, M31's most massive globular cluster. We present high-resolution, adaptive-optics assisted, integral-field spectroscopic kinematics from Gemini/NIFS that shows a strong rotation ($\sim$20 km/s) and a velocity dispersion rise towards the center (37 km/s). We combine the kinematic data with a mass model based on a two-component fit to $HST$ ACS/HRC data of the cluster to estimate the mass of a putative BH. Our dynamical modeling suggests a $>$3$σ$ detection of a BH component of 9.1$^{+2.6}_{-2.8}\times$10$^4$ M$_\odot$ (1$σ$ uncertainties). The inferred stellar mass of the cluster is 6.22$^{+0.03}_{-0.05}\times$10$^6$ M$_\odot$, consistent with previous estimates, thus the BH makes up 1.5% of its mass. We examine whether the observed kinematics are caused by a collection of stellar mass BHs by modeling an extended dark mass as a Plummer profile. The upper limit on the size scale of the extended mass is 0.56 pc (95% confidence), which does not rule out an extended mass. There is compelling evidence that B023-G078 is the tidally stripped nucleus of a galaxy with a stellar mass $>$10$^9$ M$_{\odot}$, including its high mass, two-component luminosity profile, color, metallicity gradient, and spread in metallicity. Given the emerging evidence that the central BH occupation fraction of $>$10$^9$ M$_{\odot}$ galaxies is high, the most plausible interpretation of the kinematic data is that B023-G078 hosts a central BH. This makes it the strongest BH detection in a lower mass ($<$10$^7$ M$_{\odot}$) stripped nucleus, and one of the few dynamically detected intermediate-mass BHs.

astro-ph.GA↗

arXiv2021

The TREX Survey: Kinematical Complexity Throughout M33's Stellar Disk and Evidence for a Stellar Halo

We present initial results from a large spectroscopic survey of stars throughout M33's stellar disk. We analyze a sample of 1667 red giant branch (RGB) stars extending to projected distances of $\sim 11$ kpc from M33's center ($\sim 18$ kpc, or $\sim 10$ scale lengths, in the plane of the disk). The line-of-sight velocities of RGB stars show the presence of two kinematical components. One component is consistent with rotation in the plane of M33's HI disk and has a velocity dispersion ($\sim 19$ km s$^{-1}$) consistent with that observed in a comparison sample of younger stars, while the second component has a significantly higher velocity dispersion. A two-component fit to the RGB velocity distribution finds that the high dispersion component has a velocity dispersion of $59.3^{+2.6}_{-2.5}$ km s$^{-1}$ and rotates very slowly in the plane of the disk (consistent with no rotation at the $<1.5σ$ level), which favors interpreting it as a stellar halo rather than a thick disk population. A spatial analysis indicates that the fraction of RGB stars in the high-velocity-dispersion component decreases with increasing radius over the range covered by the spectroscopic sample. Our spectroscopic sample establishes that a significant high-velocity-dispersion component is present in M33's RGB population from near M33's center to at least the radius where M33's HI disk begins to warp at 30$'$ ($\sim 7.5$ kpc) in the plane of the disk. This is the first detection and spatial characterization of a kinematically hot stellar component throughout M33's inner regions.

astro-ph.GA↗

arXiv2021

Resolved nuclear kinematics link the formation and growth of nuclear star clusters with the evolution of their early and late-type hosts

We present parsec-scale kinematics of eleven nearby galactic nuclei, derived from adaptive-optics assisted integral-field spectroscopy at (near-infrared) CO band-head wavelengths. We focus our analysis on the balance between ordered rotation and random motions, which can provide insights into the dominant formation mechanism of nuclear star clusters (NSCs). We divide our target sample into late- and early-type galaxies, and discuss the nuclear kinematics of the two sub-samples, aiming at probing any link between NSC formation and host galaxy evolution. The results suggest that the dominant formation mechanism of NSCs is indeed affected by the different evolutionary paths of their hosts across the Hubble sequence. More specifically, nuclear regions in late-type galaxies are on average more rotation dominated, and the formation of nuclear stellar structures is potentially linked to the presence of gas funnelled to the center. Early-type galaxies, in contrast, tend to display slower-rotating NSCs with lower ellipticity. However, some exceptions suggest that in specific cases, early-type hosts can form NSCs in a way similar to spirals.

astro-ph.GA↗

arXiv2021

A Mathematical Walkthrough and Discussion of the Free Energy Principle

The Free-Energy-Principle (FEP) is an influential and controversial theory which postulates a deep and powerful connection between the stochastic thermodynamics of self-organization and learning through variational inference. Specifically, it claims that any self-organizing system which can be statistically separated from its environment, and which maintains itself at a non-equilibrium steady state, can be construed as minimizing an information-theoretic functional -- the variational free energy -- and thus performing variational Bayesian inference to infer the hidden state of its environment. This principle has also been applied extensively in neuroscience, and is beginning to make inroads in machine learning by spurring the construction of novel and powerful algorithms by which action, perception, and learning can all be unified under a single objective. While its expansive and often grandiose claims have spurred significant debates in both philosophy and theoretical neuroscience, the mathematical depth and lack of accessible introductions and tutorials for the core claims of the theory have often precluded a deep understanding within the literature. Here, we aim to provide a mathematically detailed, yet intuitive walk-through of the formulation and central claims of the FEP while also providing a discussion of the assumptions necessary and potential limitations of the theory. Additionally, since the FEP is a still a living theory, subject to internal controversy, change, and revision, we also present a detailed appendix highlighting and condensing current perspectives as well as controversies about the nature, applicability, and the mathematical assumptions and formalisms underlying the FEP.

cs.AI↗

arXiv2021

Neural Kalman Filtering

The Kalman filter is a fundamental filtering algorithm that fuses noisy sensory data, a previous state estimate, and a dynamics model to produce a principled estimate of the current state. It assumes, and is optimal for, linear models and white Gaussian noise. Due to its relative simplicity and general effectiveness, the Kalman filter is widely used in engineering applications. Since many sensory problems the brain faces are, at their core, filtering problems, it is possible that the brain possesses neural circuitry that implements equivalent computations to the Kalman filter. The standard approach to Kalman filtering requires complex matrix computations that are unlikely to be directly implementable in neural circuits. In this paper, we show that a gradient-descent approximation to the Kalman filter requires only local computations with variance weighted prediction errors. Moreover, we show that it is possible under the same scheme to adaptively learn the dynamics model with a learning rule that corresponds directly to Hebbian plasticity. We demonstrate the performance of our method on a simple Kalman filtering task, and propose a neural implementation of the required equations.

cs.NE↗

arXiv2021

Relics of Supermassive Black Hole Seeds: The Discovery of an Accreting Black Hole in an Optically Normal, Low Metallicity Dwarf Galaxy

The detection and characterization of supermassive black holes (SMBHs) in local low mass galaxies is crucial to our understanding of the origins of SMBHs. This statement assumes that low mass galaxies have had a relatively quiet cosmic history, so that their black holes have not undergone significant growth and therefore can be treated as relics of the original SMBH seeds. While recent studies have found optical signatures of active galactic nuclei (AGNs) in a growing population of dwarf galaxies, these studies are biased against low metallicity and relatively merger-free galaxies, thus missing precisely the demographic in which to search for the relics of SMBH seeds. Here, we report the detection of the [\ion{Si}{6}]1.963~$μ$m coronal line (CL), a robust indicator of an AGN in the galaxy SDSS~J160135.95+311353.7, a nearby ($z=0.031$) low metallicity galaxy with a stellar mass approximately an order of magnitude lower than the LMC ($M_*\approx10^{8.56}$~M$_\odot$) and no optical evidence for an AGN. The AGN bolometric luminosity implied by the CL detection is $\approx10^{42}$~erg~s$^{-1}$, precisely what is predicted from its near-infrared continuum emission based on well-studied AGNs. Our results are consistent with a black hole of mass $\approx~10^5$~M$_\odot$, in line with expectations based on its stellar mass. This is the first time a near-infrared CL has been detected in a low mass, low metallicity galaxy with no optical evidence for AGN activity, providing confirmation of the utility of infrared CLs in finding AGNs in low mass galaxies when optical diagnostics fail. These observations highlight a powerful avenue of investigation to hunt for low mass black holes in the JWST era.

astro-ph.GA↗

arXiv2021

NGC 5128 globular cluster candidates out to 150 kpc: a comprehensive catalog from Gaia and ground based data

We present a new catalog of 40502 globular cluster (GC) candidates in NGC 5128 out to a projected radius of $\sim$150 kpc, based on data from the Panoramic Imaging Survey of Centaurus and Sculptor (PISCeS), Gaia Data Release 2, and the NOAO Source Catalog. Ranking these candidates based on the likelihood that they are true GCs, we find that approximately 1900 belong to our top two ranking categories and should be the highest priority for spectroscopic follow-up for confirmation. Taking into account our new data and a vetting of previous GC catalogs, we estimate a total GC population of $1450 \pm 160$ GCs. We show that a substantial number of sources previously argued to be low-velocity GCs are instead foreground stars, reducing the inferred GC velocity dispersion. This work showcases the power of Gaia to identify slightly extended sources at the $\sim 4$ Mpc distance of NGC 5128, enabling accurate identification of GCs throughout the entire extended halo, not just the inner regions that have been the focus of most previous work.

astro-ph.GA↗

arXiv2021

Composite Bulges -- II. Classical Bulges and Nuclear Discs in Barred Galaxies: The Contrasting Cases of NGC 4608 and NGC 4643

We present detailed morphological, photometric, and stellar-kinematic analyses of the central regions of two massive, early-type barred galaxies with nearly identical large-scale morphologies. Both have large, strong bars with prominent inner photometric excesses that we associate with boxy/peanut-shaped (B/P) bulges; the latter constitute ~ 30% of the galaxy light. Inside its B/P bulge, NGC 4608 has a compact, almost circular structure (half-light radius R_e approx. 310 pc, Sérsic n = 2.2) we identify as a classical bulge, amounting to 12.1% of the total light, along with a nuclear star cluster (R_e ~ 4 pc). NGC 4643, in contrast, has a nuclear disc with an unusual broken-exponential surface-brightness profile (13.2% of the light), and a very small spheroidal component (R_e approx. 35 pc, n = 1.6; 0.5% of the light). IFU stellar kinematics support this picture, with NGC 4608's classical bulge slowly rotating and dominated by high velocity dispersion, while NGC 4643's nuclear disc shows a drop to lower dispersion, rapid rotation, V-h3 anticorrelation, and elevated h4. Both galaxies show at least some evidence for V-h3 correlation in the bar (outside the respective classical bulge and nuclear disc), in agreement with model predictions. Standard 2-component (bulge/disc) decompositions yield B/T ~ 0.5-0.7 (and bulge n > 2) for both galaxies. This overestimates the true "spheroid" components by factors of four (NGC 4608) and over 100 (NGC 4643), illustrating the perils of naive bulge-disc decompositions applied to massive barred galaxies.

astro-ph.GA↗

arXiv2021

The Panchromatic Hubble Andromeda Treasury: Triangulum Extended Region (PHATTER) I. Ultraviolet to Infrared Photometry of 22 Million Stars in M33

We present panchromatic resolved stellar photometry for 22 million stars in the Local Group dwarf spiral Triangulum (M33), derived from Hubble Space Telescope (HST) observations with the Advanced Camera for Surveys (ACS) in the optical (F475W, F814W), and the Wide Field Camera 3 (WFC3) in the near ultraviolet (F275W, F336W) and near-infrared (F110W, F160W) bands. The large, contiguous survey area covers $\sim$14 square kpc and extends to 3.5 kpc (14 arcmin, or 1.5-2 scale lengths) from the center of M33. The PHATTER observing strategy and photometry technique closely mimic those of the Panchromatic Hubble Andromeda Treasury (PHAT), but with updated photometry techniques that take full advantage of all overlapping pointings (aligned to within $<$5-10 milliarcseconds) and improved treatment of spatially-varying point spread functions. The photometry reaches a completeness-limited depth of F475W$\sim$28.5 in the lowest surface density regions observed in M33 and F475W$\sim$26.5 in the most crowded regions found near the center of M33. We find the young populations trace several relatively tight arms, while the old populations show a clear, looser two-armed structure. We present extensive analysis of the data quality including artificial star tests to quantify completeness, photometric uncertainties, and flux biases. This stellar catalog is the largest ever produced for M33, and is publicly available for download by the community.

astro-ph.GA↗

arXiv2020

The APOGEE Library of Infrared SSP Templates (A-LIST): High-Resolution Simple Stellar Population Spectral Models in the H-Band

Integrated light spectroscopy from galaxies can be used to study the stellar populations that cannot be resolved into individual stars. This analysis relies on stellar population synthesis (SPS) techniques to study the formation history and structure of galaxies. However, the spectral templates available for SPS are limited, especially in the near-infrared. We present A-LIST (APOGEE Library of Infrared SSP Templates), a new set of high-resolution, near-IR SSP spectral templates spanning a wide range of ages (2-12 Gyr), metallicities ($\rm -2.2 < [M/H] < +0.4$) and $α$ abundances ($\rm -0.2 < [α/M] < +0.4$). This set of SSP templates is the highest resolution ($R\sim22500$) available in the near infrared, and the first such based on an empirical stellar library. Our models are generated using spectra of $\sim$300,000 stars spread across the Milky Way, with a wide range of metallicities and abundances, from the APOGEE survey. We show that our model spectra provide accurate fits to M31 globular cluster spectra taken with APOGEE, with best-fit metallicities agreeing with those of previous estimates to within $\sim$0.1 dex. We also compare these model spectra to lower-resolution E-MILES models and demonstrate that we recover the ages of these models to within $\sim$1.5 Gyr. This library is available in https://github.com/aishashok/ALIST-library.

astro-ph.GA↗

arXiv2020

Investigating the Scalability and Biological Plausibility of the Activation Relaxation Algorithm

The recently proposed Activation Relaxation (AR) algorithm provides a simple and robust approach for approximating the backpropagation of error algorithm using only local learning rules. Unlike competing schemes, it converges to the exact backpropagation gradients, and utilises only a single type of computational unit and a single backwards relaxation phase. We have previously shown that the algorithm can be further simplified and made more biologically plausible by (i) introducing a learnable set of backwards weights, which overcomes the weight-transport problem, and (ii) avoiding the computation of nonlinear derivatives at each neuron. However, tthe efficacy of these simplifications has, so far, only been tested on simple multi-layer-perceptron (MLP) networks. Here, we show that these simplifications still maintain performance using more complex CNN architectures and challenging datasets, which have proven difficult for other biologically-plausible schemes to scale to. We also investigate whether another biologically implausible assumption of the original AR algorithm -- the frozen feedforward pass -- can be relaxed without damaging performance.

cs.AI↗

arXiv2020

Relaxing the Constraints on Predictive Coding Models

Predictive coding is an influential theory of cortical function which posits that the principal computation the brain performs, which underlies both perception and learning, is the minimization of prediction errors. While motivated by high-level notions of variational inference, detailed neurophysiological models of cortical microcircuits which can implements its computations have been developed. Moreover, under certain conditions, predictive coding has been shown to approximate the backpropagation of error algorithm, and thus provides a relatively biologically plausible credit-assignment mechanism for training deep networks. However, standard implementations of the algorithm still involve potentially neurally implausible features such as identical forward and backward weights, backward nonlinear derivatives, and 1-1 error unit connectivity. In this paper, we show that these features are not integral to the algorithm and can be removed either directly or through learning additional sets of parameters with Hebbian update rules without noticeable harm to learning performance. Our work thus relaxes current constraints on potential microcircuit designs and hopefully opens up new regions of the design-space for neuromorphic implementations of predictive coding.

q-bio.NC↗