Machine learning approaches to spatiotemporal physical systems have primarily focused on next-frame prediction, with the goal of learning an accurate emulator for the system's evolution in time. However, these emulators are computationally expensive to train and are subject to performance pitfalls, such as compounding errors during autoregressive rollout. In this work, we take a different perspective and look at scientific tasks further downstream of predicting the next frame, such as estimation of a system's governing physical parameters. Accuracy on these tasks offers a uniquely quantifiable glimpse into the physical relevance of the representations of these models. We evaluate the effectiveness of general-purpose self-supervised methods in learning physics-grounded representations that are useful for downstream scientific tasks. Surprisingly, we find that not all methods designed for physical modeling outperform generic self-supervised learning methods on these tasks, and methods that learn in the latent space (e.g., joint embedding predictive architectures, or JEPAs) outperform those optimizing pixel-level prediction objectives. Code is available at https://github.com/helenqu/physical-representation-learning.
Materials with interesting physical properties are often designed based on our understanding of the target physical effects. The physical properties can be either explicitly observed ("apparent") or concealed by the perceived symmetry ("hidden") but still exist. Both are enabled by specific symmetries and induced by certain physical interactions. Using the underlying approach of condensed matter theory of real materials (rather than schematic model Hamiltonians), we discuss apparent and hidden physics in real materials focusing on the properties of spin splitting and spin polarization. Depending on the enabling symmetries and underlying physical interactions, we classify spin effects into four categories with each having two subtypes; representative materials are pointed out. We then discuss the electric tunability and switch of apparent and hidden spin splitting and polarization in antiferromagnets. Finally, we extend "hidden effects" to views that are farsighted in the sense of resolving the correct atomistic and reciprocal symmetry and replaced by the incorrect higher symmetry. This framework could guide and enable systematic discovery of such intriguing effects.
Recently we have witnessed progress in hiding road vehicles against object detectors through adversarial camouflage in the digital world. The extension of this technique to the physical world is crucial for testing the robustness of autonomous driving systems. However, existing methods do not show good performances when applied to the physical world. This is partly due to insufficient photorealism in training examples, and lack of proper physical realization methods for camouflage. To generate a robust adversarial camouflage suitable for real vehicles, we propose a novel method called PAV-Camou. We propose to adjust the mapping from the coordinates in the 2D map to those of corresponding 3D model. This process is critical for mitigating texture distortion and ensuring the camouflage's effectiveness when applied in the real world. Then we combine two renderers with different characteristics to obtain adversarial examples that are photorealistic that closely mimic real-world lighting and texture properties. The method ensures that the generated textures remain effective under diverse environmental conditions. Our adversarial camouflage can be optimized and printed in the form of 2D patterns, allowing for direct application on real vehicles. Extensive experiments demonstrated that our proposed method achieved good performance in both the digital world and the physical world.
Virtual Reality (VR) moves away from well-controlled laboratory environments into public and personal spaces. As users are visually disconnected from the physical environment, interacting in an uncontrolled space frequently leads to collisions and raises safety concerns. In my thesis, I investigate this phenomenon which I define as the physical breakdown in VR. The goal is to understand the reasons for physical breakdowns, provide solutions, and explore future mechanisms that could perpetuate safety risks. First, I explored the reasons for physical breakdowns by investigating how people interact with the current VR safety mechanism (e.g., Oculus Guardian). Results show one reason for breaking out of the safety boundary is when interacting with large motions (e.g., swinging arms), the user does not have enough time to react although they see the safety boundary. I proposed a solution, FingerMapper, that maps small-scale finger motions onto virtual arms and hands to enable whole-body virtual arm motions in VR to avoid physical breakdowns. To demonstrate future safety risks, I explored the malicious use of perceptual manipulations (e.g., redirection techniques) in VR, which could deliberately create physical breakdowns without users noticing. Results indicate further open challenges about the cognitive process of how users comprehend their physical environment when they are blindfolded in VR.
This is the first of a series of papers that we intend to publish about the epistemology of fundamental physics in its current state. One of the main objectives of these papers is to improve our understanding of fundamental physics (and modern physics in particular) from an epistemological and interpretative perspective (i.e. versus formal perspective). Another main objective is to investigate and assess the merit of searching for a unified physical theory (the so-called ``theory of everything'') considering the fact that contemporary physics is a collection of theories created and developed by different individuals and groups of scientists in different eras of history reflecting different levels of scientific, philosophical and epistemological development and dealing with largely separate physical phenomena and hence such unification may mean ``stitching together'' an inhomogeneous collection of theoretical structures which may be clumsy (if not impossible) at least from an epistemological viewpoint.
This is a comment on both Gruber et al (2022) and Bunamano and Rovelli (2022), which discuss the relation between physical time and human time. I claim here, contrary to many views discussed there, that there is no foundational conflict between the way physics views the passage of time and the way the mind/brain perceives it. The problem rather resides in a number of misconceptions leading to the representation of spacetime as a timeless Block Universe. The physical expanding universe is in fact an Evolving Block Universe with a time-dependent future boundary. This establishes a global direction of time that determines local arrows of time. Furthermore time passes when quantum wave function collapse takes place; during this process, information is lost. The mind/brain acts as an imperfect clock, which coarse-grains the physical passage of time along a world line to determine the experienced passage of time, because neuronal processes take time to occur. This happens in a contextual way, so experienced time is not linearly related to physical time in general. Finally I point out that the Universe is never infinitely old: its future endpoint always lies infinitely faraway in the future
Symmetry lies at the heart of todays theoretical study of particle physics. Our manuscript is a tutorial introducing foundational mathematics for understanding physical symmetries. We start from basic group theory and representation theory. We then introduce Lie Groups and Lie Algebra and their properties. We next discuss with detail two important Lie Groups in physics Special Unitary and Lorentz Group, with an emphasis on their applications to particle physics. Finally, we introduce field theory and its version of the Noether Theorem. We believe that the materials cover here will prepare undergraduates for future studies in mathematical physics.
In glaring contrast to its indisputable century-old experimental success, the ultimate objects and meaning of quantum physics remain a matter of vigorous debate among physicists and philosophers of science. This article attempts to shed new light on the debate. It relies upon two comments by Albert Einstein on his general approach to physical theories. I draw their consequences for the definition of a physical theory's ontology, and next for the ontology of quantum physics - i.e. what it may ultimately be about. The quantum ontology thus derived appears to be strictly limited to evolving experimental contexts and instantaneous measurement outcomes, which are to be understood, respectively, as mere potential measurement outcomes and actual ones. The notions of material body in particular and physical system in general are absent from this ontology, hence the vanishing of Schroedinger's cat and EPR paradoxes, as well as of the quantum measurement problem. Apart from its ability to clear up well-known conundrums, this ontology reveals what quantum physics may fundamentally be - i.e. its possible ultimate meaning - namely a theory of general relativity of experimental context. On this basis, I conclude with a new conception of objectivity in the sciences of nature.
We consider the application of deep generative models in propagating uncertainty through complex physical systems. Specifically, we put forth an implicit variational inference formulation that constrains the generative model output to satisfy given physical laws expressed by partial differential equations. Such physics-informed constraints provide a regularization mechanism for effectively training deep probabilistic models for modeling physical systems in which the cost of data acquisition is high and training data-sets are typically small. This provides a scalable framework for characterizing uncertainty in the outputs of physical systems due to randomness in their inputs or noise in their observations. We demonstrate the effectiveness of our approach through a canonical example in transport dynamics.
The physical constructs underlying the properties of quantum mechanics are explored. Arguments are given that the particle wave function as well as photon and phonon quanta must derive from a more fundamental physical construct that has not yet been identified. An approach to identifying the construct is discussed and a specific construct is proposed and explained.
Event horizons are (generically) not physically observable. In contrast, apparent horizons (and the closely related trapping horizons) are generically physically observable --- in the sense that they can be detected by observers working in finite-size regions of spacetime. Consequently event horizons are inappropriate tools for defining astrophysical black holes, or indeed for defining any notion of evolving}black hole, (evolving either due to accretion or Hawking radiation). The only situation in which an event horizon becomes physically observable is for the very highly idealized stationary or static black holes, when the event horizon is a Killing horizon which is degenerate with the apparent and trapping horizons; and then it is the physical observability of the apparent/trapping horizons that is fundamental --- the event horizon merely comes along for the ride.
Using basic physical arguments, we derive by dimensional and physical analysis the characteristic masses and sizes of important objects in the Universe in terms of just a few fundamental constants. This exercise illustrates the unifying power of physics and the profound connections between the small and the large in the Cosmos we inhabit. We focus on the minimum and maximum masses of normal stars, the corresponding quantities for neutron stars, the maximum mass of a rocky planet, the maximum mass of a white dwarf, and the mass of a typical galaxy. To zeroth order, we show that all these masses can be expressed in terms of either the Planck mass or the Chandrasekar mass, in combination with various dimensionless quantities. With these examples we expose the deep interrelationships imposed by Nature between disparate realms of the Universe and the amazing consequences of the unifying character of physical law.
We present a vision for the next decade of hadron physics in which the central question being addressed is how one might win new physical insight into the way hadronic systems work. The topics addressed include the relevance of model building, the role of spontaneously broken chiral symmetry, spectroscopy, form factors and physics in the deep inelastic regime.
Every algorithm which can be executed on a computer can at least in principle be realized in hardware, i.e. by a discrete physical system. The problem is that up to now there is no programming language by which physical systems can constructively be described. Such tool, however, is essential for the compact description and automatic production of complex systems. This paper introduces a programming language, called Akton-Algebra, which provides the foundation for the complete description of discrete physical systems. The approach originates from the finding that every discrete physical system reduces to a spatiotemporal topological network of nodes, if the functional and metric properties are deleted. A next finding is that there exists a homeomorphism between the topological network and a sequence of symbols representing a program by which the original nodal network can be reconstructed. Providing Akton-Algebra with functionality turns it into a flow-controlled general data processing language, which by introducing clock control and addressing can be further transformed into a classical programming language. Providing Akton-Algebra with metrics, i.e. the shape and size of the components, turns it into a novel hardware system construction language.
The process of cognition is analysed to adjust the set theory to physical description. Postulates and basic definitions are revised. The specific sets of predicates, called presets, corresponding to the physical objects identified by an observer during cognition are introduced. Unlike sets, the presets are free of logical or set-theoretical paradoxes and may be consistently used in physical description. Schemes of cognition based on presets are considered. Being different logical systems, the relativistic and quantum theories, observations in modern cosmology cannot be consistently considered in one `unified physical theory', but they are in frames of introduced schemes of cognition.