搜索结果：destruction

This paper explores the challenges and benefits of a trainable destruction process in diffusion samplers -- diffusion-based generative models trained to sample an unnormalised density without access to data samples. Contrary to the majority of work that views diffusion samplers as approximations to an underlying continuous-time model, we view diffusion models as discrete-time policies trained to produce samples in very few generation steps. We propose to trade some of the elegance of the underlying theory for flexibility in the definition of the generative and destruction policies. In particular, we decouple the generation and destruction variances, enabling both transition kernels to be learned as unconstrained Gaussian densities. We show that, when the number of steps is limited, training both generation and destruction processes results in faster convergence and improved sampling quality on various benchmarks. Through a robust ablation study, we investigate the design choices necessary to facilitate stable training. Finally, we show the scalability of our approach through experiments on GAN latent space sampling for conditional image generation.

Context. While supernova remnants (SNRs) are observed to produce up to 1 M$_\odot$ of dust, the amount of dust destroyed by the forward shock (FS) is poorly constrained, raising the question whether they are net dust producers or destroyers. Aims. We aim to estimate the dust destruction efficiency of SNR FSs in a realistically turbulent interstellar medium (ISM) during their most destructive phase, and assess dust shielding by high density filaments during this period. Methods. We run 3D turbulence simulations for different turbulent Mach numbers (0-3) and average ISM densities (1-100 cm$^{-3}$) to resemble observations of the turbulent ISM. We then set off a supernova to trace its 3D magnetohydrodynamical evolution for 10 kyr. Finally, we run post-processing simulations to study the dust transport and destruction by the SNR FS, considering gas and plasma drag, kinetic and thermal sputtering, and grain-grain collisions, and either silicate or carbonaceous dust. Results. The dust destruction rate of the FS strongly depends on the average ISM density and turbulence strength, varying between 27-92% (0.85-11.0 M$_\odot$) in the studied 10 kyr. Overall, dust is less efficiently destroye

Strange metal behavior has been observed in an expanding list of quantum materials, with heavy fermion metals serving as a prototype setting. Among the intriguing questions is the nature of charge carriers; there is an increasing recognition that the quasiparticles are lost, as captured by Kondo destruction quantum criticality. Among the recent experimental advances is the measurement of shot noise in a heavy-fermion strange metal. We are thus motivated to study current fluctuations by advancing a minimal Bose-Fermi Kondo lattice model, which admits a well-defined large-$N$ limit. Showing that the model in equilibrium captures the essential physics of Kondo destruction, we proceed to derive quantum kinetic equations and compute shot noise to the leading nontrivial order in $1/N$. Our results reveal a strong suppression of the shot noise at the Kondo destruction quantum critical point, thereby providing the understanding of the striking experiment. Broader implications of our results are discussed.

Dust in the interstellar medium (ISM) is critical to the absorption and intensity of emission profiles used widely in astronomical observations, and necessary for star and planet formation. Supernovae (SNe) both produce and destroy ISM dust. In particular the destruction rate is difficult to assess. Theory and prior simulations of dust processing by SNe in a uniform ISM predict quite high rates of dust destruction, potentially higher than the supernova dust production rate in some cases. Here we show simulations of supernova-induced dust processing with realistic ISM dynamics including magnetic field effects and demonstrate how ISM inhomogeneity and magnetic fields inhibit dust destruction. Compared to the non-magnetic homogeneous case, the dust mass destroyed within 1 Myr per SNe is reduced by more than a factor of two, which can have a great impact on the ISM dust budget.

I present diffusion models as part of a family of machine learning techniques that withhold information from a model's input and train it to guess the withheld information. I argue that diffusion's destroying approach to withholding is more flexible than typical hand-crafted information withholding techniques, providing a rich training playground that could be advantageous in some settings, notably data-scarce ones. I then address subtle issues that may arise when porting reinforcement learning techniques to the diffusion context, and wonder how such exploration problems could be addressed in more diffusion-native ways. I do not have definitive answers, but I do point my fingers in directions I deem interesting. A tutorial follows this thesis, expanding on the destroy-then-generate perspective. A novel kind of probabilistic graphical models is introduced to facilitate the tutorial's exposition.

Earth is deficient in carbon and nitrogen by up to ${\sim}4$ orders of magnitude compared with the Sun. Destruction of (carbon- and nitrogen-rich) refractory organics in the high-temperature planet forming regions could explain this deficiency. Assuming a refractory cometary composition for these grains, their destruction enhances nitrogen-containing oxygen-poor molecules in the hot gas ($\gtrsim 300$K) after the initial formation and sublimation of these molecules from oxygen-rich ices in the warm gas (${\sim}150$K). Using observations of $37$ high-mass protostars with ALMA, we find that oxygen-containing molecules (CH$_3$OH and HNCO) systematically show no enhancement in their hot component. In contrast, nitrogen-containing, oxygen-poor molecules (CH$_3$CN and C$_2$H$_3$CN) systematically show an enhancement of a factor ${\sim} 5$ in their hot component, pointing to additional production of these molecules in the hot gas. Assuming only thermal excitation conditions, we interpret these results as a signature of destruction of refractory organics, consistent with the cometary composition. This destruction implies a higher C/O and N/O in the hot gas than the warm gas, while, the exa

A leading candidate for the heating source of chondrules and igneous rims is shock waves. This mechanism generates high relative velocities between chondrules and dust particles. We have investigated the possibility of the chondrule destruction in collisions with dust particles behind a shock wave using a semianalytical treatment. We find that the chondrules are destroyed during melting in collisions. We derive the conditions for the destruction of chondrules and show that the typical size of the observed chondrules satisfies the condition. We suggest that the chondrule formation and rim accretion are different events if they are heated by shock waves.

Class incremental learning (CIL) aims to incrementally update a trained model with the new classes of samples (plasticity) while retaining previously learned ability (stability). To address the most challenging issue in this goal, i.e., catastrophic forgetting, the mainstream paradigm is memory-replay CIL, which consolidates old knowledge by replaying a small number of old classes of samples saved in the memory. Despite effectiveness, the inherent destruction-reconstruction dynamics in memory-replay CIL are an intrinsic limitation: if the old knowledge is severely destructed, it will be quite hard to reconstruct the lossless counterpart. Our theoretical analysis shows that the destruction of old knowledge can be effectively alleviated by balancing the contribution of samples from the current phase and those saved in the memory. Motivated by this theoretical finding, we propose a novel Balanced Destruction-Reconstruction module (BDR) for memory-replay CIL, which can achieve better knowledge reconstruction by reducing the degree of maximal destruction of old knowledge. Specifically, to achieve a better balance between old knowledge and new classes, the proposed BDR module takes into

Using a set of clusters in dark matter only cosmological simulations, we study the consequences of merging of clusters and groups of galaxies (with mass ratio larger than 5:1) to investigate the tidal impact of mergers on the satellite halos. We compare our results to a control sample of clusters that have had no major mergers over the same time period. Clusters that undergo major mergers are found to have a significant enhancement in destruction of their subhalos of ~10-30%, depending on how major the merger is. Those with mass ratios less than 7:1 showed no significant enhancement. The number of destroyed subhalos are measured for the cluster members that were inside the virial radius of clusters before the merger begins. This means preprocessed galaxies brought in by the merger are deliberately excluded, allowing us to clearly see the enhanced destruction purely as a result of the distorted and disturbed tidal field of the cluster during the merger. We also consider secondary parameters affecting the destruction of those satellites but find that the major mergers are the dominant factor. These results highlight how major mergers can significantly impact the cluster population, w

Supernova generated shock waves are responsible for most of the destruction of dust grains in the interstellar medium (ISM). Calculations of the dust destruction timescale have so far been carried out using plane parallel steady shocks, however that approximation breaks down when the destruction timescale becomes longer than that for the evolution of the supernova remnant (SNR) shock. In this paper we present new calculations of grain destruction in evolving, radiative SNRs. To facilitate comparison with the previous study by Jones et al. (1996), we adopt the same dust properties as in that paper. We find that the efficiencies of grain destruction are most divergent from those for a steady shock when the thermal history of a shocked gas parcel in the SNR differs significantly from that behind a steady shock. This occurs in shocks with velocities >~ 200 km/s for which the remnant is just beginning to go radiative. Assuming SNRs evolve in a warm phase dominated ISM, we find dust destruction timescales are increased by a factor of ~2 compared to those of Jones et al. (1996), who assumed a hot gas dominated ISM. Recent estimates of supernova rates and ISM mass lead to another factor

Path of Destruction (PoD) is a self-supervised method for learning iterative generators. The core idea is to produce a training set by destroying a set of artifacts, and for each destructive step create a training instance based on the corresponding repair action. A generator trained on this dataset can then generate new artifacts by repairing from arbitrary states. The PoD method is very data-efficient in terms of original training examples and well-suited to functional artifacts composed of categorical data, such as game levels and discrete 3D structures. In this paper, we extend the Path of Destruction method to allow designer control over aspects of the generated artifacts. Controllability is introduced by adding conditional inputs to the state-action pairs that make up the repair trajectories. We test the controllable PoD method in a 2D dungeon setting, as well as in the domain of small 3D Lego cars.

Theories which have fundamental information destruction or decoherence are motivated by the black hole information paradox where one appears to have pure states evolving into mixed states. However such theories have either violated conservation laws, or are highly non-local. Here, we show that the tension between conservation laws and locality can be circumvented by constructing a relational theory of information destruction. In terms of conservation laws, we derive a generalisation of Noether's theorem for general theories, and show that symmetries imply a strong restriction on the type of evolution permissable. With respect to locality, we distinguish violations of causality from the creation or destruction of space-like seperated correlations. We show that violations of causality need not occur in a relational framework, although one can have situations where correlations decay faster than one might otherwise expect or can be created over spatial distances. This creation or destruction of correlations cannot be used to signal superluminally, and thus no violation of causality occurs. We prove that theories with information destruction can be made time-symmetric, thus impossing n

This short note serves an addendum to the article "Destruction of very simple trees" by Fill, Kapur and Panholzer (2004). Therein, the limit law of one-sided tree destruction with a toll function was determined by its moment sequence. We add an identification of the limit law, using recent results of Bertoin (2022), in terms of the local time of a noise reinforced Bessel process.

A model is presented to calculate projectile core destruction in knockout reactions. It incorporates physics arguments similar to the formulation of the state of the art theory to calculate stripping and diffraction dissociation cross sections in heavy ion collisions with bombarding energies around 100 MeV/nucleon and larger. It is shown that secondary collisions between the incoming and struck nucleons and the projectile core decrease the core survival probability by as much as 9.5\%. However, no clear evidence is found for reduction of the cross section with increasing binding energy of the removed nucleon.

We investigate the destruction of dust grains by sputtering in the high-velocity interstellar shocks driven by supernovae (SNe) in the early universe to reveal the dependence of the time-scale of dust destruction on the gas density $n_{{\rm H}, 0}$ in the interstellar medium (ISM) as well as on the progenitor mass $M_{\rm pr}$ and explosion energy $E_{\rm 51}$ of SN. The sputtering yields for the combinations of dust and ion species of interest to us are evaluated by applying the so-called universal relation with a slight modification. The dynamics of dust grains and their destruction by sputtering in shock are calculated by taking into account the size distribution of each dust species, together with the time evolution of temperature and density of gas in spherically symmetric shocks. The results of calculations show that the efficiency of dust destruction depends not only on the sputtering yield but also on the initial size distribution of each grain species. The efficiency of dust destruction increases with increasing $E_{\rm 51}$ and/or increasing $n_{{\rm H}, 0}$, but is almost independent of $M_{\rm pr}$ as long as $E_{\rm 51}$ is the same. The mass of gas swept up by shock i

We calculate orbits, tidal radii, and bulge-bar and disk shocking destruction rates for 63 globular clusters in our Galaxy. Orbits are integrated in both an axisymmetric and a non-axisymmetric Galactic potential that includes a bar and a 3D model for the spiral arms. With the use of a Monte Carlo scheme, we consider in our simulations observational uncertainties in the kinematical data of the clusters. In the analysis of destruction rates due to the bulge-bar, we consider the rigorous treatment of using the real Galactic cluster orbit, instead of the usual linear trajectory employed in previous studies. We compare results in both treatments. We find that the theoretical tidal radius computed in the nonaxisymmetric Galactic potential compares better with the observed tidal radius than that obtained in the axisymmetric potential. In both Galactic potentials, bulge-shocking destruction rates computed with a linear trajectory of a cluster at its perigalacticons give a good approximation to the result obtained with the real trajectory of the cluster. Bulge-shocking destruction rates for clusters with perigalacticons in the inner Galactic region are smaller in the non-axisymmetric potent

The destruction of planets by migration into the star will release significant amounts of energy and material, which will present opportunities to observational study planets in new ways. To observe planet destruction, it is important to understand the processes of how this energy and material is released as planets are destroyed. It is not known how fast the large amounts of energy and material are released, making it difficult to predict how observable planet destruction will be. There is a huge amount of energy made available by falling deep into the star's potential well: Simple calculations show that many of the currently known "hot Jupiters" can potentially produce events as luminous as a small nova if the energy is released fast enough. To observe these events, the important questions are how will this energy be released, and whether the energy will be released rapidly enough to create an event luminous enough to be found by transient surveys. Alternatively, if planet destruction is slowed by the inward migration alternating with periods of outward migration caused by Roche lobe overflow (RLOF), then the primary signature may be the effects of the release of large amounts of

The size distribution of dust particles in nuclear fusion devices is close to the power function. A function of this kind can be the result of brittle destruction. From the similarity assumption it follows that the size distribution obeys the power law with the exponent between -4 and -1. The model of destruction has much in common with the fractal theory. The power exponent can be expressed in terms of the fractal dimension. Reasonable assumptions on the shape of fragments concretize the power exponent, and vice versa possible destruction laws can be inferred on the basis of measured size distributions.

Small, pebble-sized objects and large bodies of planetesimal size both play important roles in planet formation. They form the evolutionary steps of dust growth in their own respect. However, at later times, they are also thought to provide background populations of mass that larger bodies might feed upon. What we suggest in this work is that starting at times of viscous stirring, planetesimals on eccentric orbits could simply explode as they become supersonic in comparison to small, porous planetary bodies entering Earth's atmosphere. We present a toy model of planetesimal motion and destruction to show the key aspects of this process. The consequences are quite severe. At all times, it is shown that only planetesimals on more or less circular orbits exist in the inner disk. After the destruction of a planetesimal, the remaining matter is continuously redistributed to the pebble reservoir of the protoplanetary disk. Since destruction typically occurs at small stellar distances due to supersonic speeds, it is expected to boost pebble accretion in the inner protoplanetary disk as one of its main effects.