More Like this

1. Perceptual Compression for Video Storage and Processing Systems

   https://doi.org/10.1145/3357223.3362725  

   Mazumdar, Amrita; Haynes, Brandon; Balazinska, Magda; Ceze, Luis; Cheung, Alvin; Oskin, Mark (November 2019, Proceedings of the ACM Symposium on Cloud Computing)

   null (Ed.)

   Compressed videos constitute 70% of Internet traffic, and video upload growth rates far outpace compute and storage improvement trends. Past work in leveraging perceptual cues like saliency, i.e., regions where viewers focus their perceptual attention, reduces compressed video size while maintaining perceptual quality, but requires significant changes to video codecs and ignores the data management of this perceptual information. In this paper, we propose Vignette, a compression technique and storage manager for perception-based video compression in the cloud. Vignette complements off-the-shelf compression software and hardware codec implementations. Vignette's compression technique uses a neural network to predict saliency information used during transcoding, and its storage manager integrates perceptual information into the video storage system. Our results demonstrate the benefit of embedding information about the human visual system into the architecture of cloud video storage systems. 

   more » « less
2. Primordial Helium-3 Redux: The Helium Isotope Ratio of the Orion Nebula*

   https://doi.org/10.3847/1538-4357/ac6503  

   Cooke, Ryan J.; Noterdaeme, Pasquier; Johnson, James W.; Pettini, Max; Welsh, Louise; Peroux, Celine; Murphy, Michael T.; Weinberg, David H. (June 2022, The Astrophysical Journal)

   Abstract We report the first direct measurement of the helium isotope ratio, 3 He/ 4 He, outside of the Local Interstellar Cloud, as part of science-verification observations with the upgraded CRyogenic InfraRed Echelle Spectrograph. Our determination of 3 He/ 4 He is based on metastable He i * absorption along the line of sight toward Θ 2 A Ori in the Orion Nebula. We measure a value 3 He/ 4 He = (1.77 ± 0.13) × 10 −4 , which is just ∼40% above the primordial relative abundance of these isotopes, assuming the Standard Model of particle physics and cosmology, ( 3 He/ 4 He) p = (1.257 ± 0.017) × 10 −4 . We calculate a suite of galactic chemical evolution simulations to study the Galactic build up of these isotopes, using the yields from Limongi & Chieffi for stars in the mass range M = 8–100 M ⊙ and Lagarde et al. for M = 0.8–8 M ⊙ . We find that these simulations simultaneously reproduce the Orion and protosolar 3 He/ 4 He values if the calculations are initialized with a primordial ratio 3 He / 4 He p = ( 1.043 ± 0.089 ) × 10 − 4 . Even though the quoted error does not include the model uncertainty, this determination agrees with the Standard Model value to within ∼2 σ . We also use the present-day Galactic abundance of deuterium (D/H), helium (He/H), and 3 He/ 4 He to infer an empirical limit on the primordial 3 He abundance, 3 He / H p ≤ ( 1.09 ± 0.18 ) × 10 − 5 , which also agrees with the Standard Model value. We point out that it is becoming increasingly difficult to explain the discrepant primordial 7 Li/H abundance with nonstandard physics, without breaking the remarkable simultaneous agreement of three primordial element ratios (D/H, 4 He/H, and 3 He/ 4 He) with the Standard Model values. 

   more » « less
3. Simultaneous perception–action design via invariant finite belief sets

   https://doi.org/10.1016/j.automatica.2023.111140  

   Hibbard, Michael; Tanaka, Takashi; Topcu, Ufuk (September 2023, Automatica)

   Although perception is an increasingly dominant portion of the overall computational cost for autonomous systems, only a fraction of the information perceived is likely to be relevant to the current task. To alleviate these perception costs, we develop a novel simultaneous perception–action design framework wherein an agent senses only the task-relevant information. This formulation differs from that of a partially observable Markov decision process, since the agent is free to synthesize not only its policy for action selection but also its belief-dependent observation function. The method enables the agent to balance its perception costs with those incurred by operating in its environment. To obtain a computationally tractable solution, we approximate the value function using a novel method of invariant finite belief sets, wherein the agent acts exclusively on a finite subset of the continuous belief space. We solve the approximate problem through value iteration in which a linear program is solved individually for each belief state in the set, in each iteration. Finally, we prove that the value functions, under an assumption on their structure, converge to their continuous state-space values as the sample density increases. 

   more » « less
4. Simultaneous perception–action design via invariant finite belief sets

   Hibbard, M. (July 2023, Automatica)

   Although perception is an increasingly dominant portion of the overall computational cost for autonomous systems, only a fraction of the information perceived is likely to be relevant to the current task. To alleviate these perception costs, we develop a novel simultaneous perception–action design framework wherein an agent senses only the task-relevant information. This formulation differs from that of a partially observable Markov decision process, since the agent is free to synthesize not only its policy for action selection but also its belief-dependent observation function. The method enables the agent to balance its perception costs with those incurred by operating in its environment. To obtain a computationally tractable solution, we approximate the value function using a novel method of invariant finite belief sets, wherein the agent acts exclusively on a finite subset of the continuous belief space. We solve the approximate problem through value iteration in which a linear program is solved individually for each belief state in the set, in each iteration. Finally, we prove that the value functions, under an assumption on their structure, converge to their continuous state-space values as the sample density increases. 

   more » « less
5. Simultaneous Perception-Action Design via Invariant Finite Belief Sets

   Hibbard, M; Tanaka, T.; & Topcu, U. (July 2023, Automatica)

   Although perception is an increasingly dominant portion of the overall computational cost for autonomous systems, only a fraction of the information perceived is likely to be relevant to the current task. To alleviate these perception costs, we develop a novel simultaneous perception–action design framework wherein an agent senses only the task-relevant information. This formulation differs from that of a partially observable Markov decision process, since the agent is free to synthesize not only its policy for action selection but also its belief-dependent observation function. The method enables the agent to balance its perception costs with those incurred by operating in its environment. To obtain a computationally tractable solution, we approximate the value function using a novel method of invariant finite belief sets, wherein the agent acts exclusively on a finite subset of the continuous belief space. We solve the approximate problem through value iteration in which a linear program is solved individually for each belief state in the set, in each iteration. Finally, we prove that the value functions, under an assumption on their structure, converge to their continuous state-space values as the sample density increases. 

   more » « less