Engineering topics

Tegmark, Max

Publications and source records attributed to Tegmark, Max.

4 records

DOE OSTI2021

Symbolic pregression: Discovering physical laws from distorted video

In this work, we present a method for unsupervised learning of equations of motion for objects in raw and optionally distorted unlabeled synthetic video (or, more generally, for discovering and modeling predictable features in time-series data). We first train an autoencoder that maps each video frame into a low-dimensional latent space where the laws of motion are as simple as possible, by minimizing a combination of nonlinearity, acceleration, and prediction error. Differential equations describing the motion are then discovered using Pareto-optimal symbolic regression. We find that our pre-regression (“pregression”) step is able to rediscover Cartesian coordinates of unlabeled moving objects even when the video is distorted by a generalized lens. Using intuition from multidimensional knot theory, we find that the pregression step is facilitated by first adding extra latent space dimensions to avoid topological problems during training and then removing these extra dimensions via principal component analysis. An inertial frame is autodiscovered by minimizing the combined equation complexity for multiple experiments.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

NASA NTRS1996

The Angular Power Spectrum of BATSE 3B Gamma-Ray Bursts

We compute the angular power spectrum C(sub l) from the BATSE 3B catalog of 1122 gamma-ray bursts and find no evidence for clustering on any scale. These constraints bridge the entire range from small scales (which probe source clustering and burst repetition) to the largest scales (which constrain possible anisotropics from the Galactic halo or from nearby cosmological large-scale structures). We develop an analysis technique that takes the angular position errors into account. For specific clustering or repetition models, strong upper limits can be obtained down to scales l approx. equal to 30, corresponding to a couple of degrees on the sky. The minimum-variance burst weighting that we employ is visualized graphically as an all-sky map in which each burst is smeared out by an amount corresponding to its position uncertainty. We also present separate bandpass-filtered sky maps for the quadrupole term and for the multipole ranges l = 3-10 and l = 11-30, so that the fluctuations on different angular scales can be inspected separately for visual features such as localized 'hot spots' or structures aligned with the Galactic plane. These filtered maps reveal no apparent deviations from isotropy.

Tegmark, Max↗

NASA NTRS1996

Improved Limits on Gamma-Ray Burst Repetition from BATSE

We tighten previous upper limits on gamma-ray burst repetition by analyzing the angular power spectrum of the BATSE 3B catalog of 1122 bursts. At 95% confidence, we find that no more than 2% of all observed bursts can be labeled as repeaters, even if no sources are observed to repeat more than once. If a fraction f of all observed bursts can be labeled as repeaters that are observed to burst upsilon times each, then all models with (upsilon - 1)f greater than or equal to 0.05 are ruled out at 99% confidence, as compared to the best previous 99% limit (upsilon - 1)f greater than or equal to 0.27. At 95% confidence, our new limit is (upsilon - 1)f greater than or equal to 0.02. Thus, even a cluster of six events from a single source would have caused excess power above that present in the 3B catalog. We conclude that the current BATSE data are consistent with no repetition of classical gamma-ray bursts and that any repeater model is severely constrained by the near-perfect isotropy of their angular distribution.

Tegmark, Max↗

NASA NTRS1996

Improved Limits on Gamma-Ray Burst Repetition from BATSE

We tighten previous upper limits on gamma-ray burst repetition by analyzing the angular power spectrum of the BATSE 3B catalog of 1122 bursts. At 95% confidence, we find that no more than 2% of all observed bursts can be labeled as repeaters, even if no sources are observed to repeat more than once. If a fraction f of all observed bursts can be labeled as repeaters that are observed to burst nu times each, then all models with (nu - 1)f greater than or equal to 0.05 are ruled out at 99% confidence, as compared to the best previous 99% limit (nu - 1)f greater than or equal to 0.27. At 95% confidence, our new limit is (nu - 1)f greater than or equal to 0.02. Thus, even a cluster of six events from a single source would have caused excess power above that present in the 3B catalog. We conclude that the current BATSE data are consistent with no repetition of classical gamma-ray bursts and that any repeater model is severely constrained by the near-perfect isotropy of their angular dis- tribution.

Tegmark, Max↗