Search for: All records

Federated Learning (FL) has pioneered the idea of share wisdom not raw data to enable collaborative learning over decentralized data. FL achieves this goal by averaging model parameters instead of centralizing data. However, representing wisdom in the form of model parameters has its own limitations including the requirement for uniform model architectures across clients and communication overhead proportional to model size.In this work we introduce Co-Dream a framework for representing wisdom in data space instead of model parameters. Here, clients collaboratively optimize random inputs based on their locally trained models and aggregate gradients of their inputs. Our proposed approach overcomes the aforementioned limitations and comes with additional benefits such as adaptive optimization and interpretable representation of knowledge. We empirically demonstrate the effectiveness of Co-Dream and compare its performance with existing techniques. 

Neural radiance fields (NeRFs) show potential for transforming images captured worldwide into immersive 3D visual experiences. However, most of this captured visual data remains siloed in our camera rolls as these images contain personal details. Even if made public, the problem of learning 3D representations of billions of scenes captured daily in a centralized manner is computationally intractable. Our approach, DecentNeRF, is the first attempt at decentralized, crowd-sourced NeRFs that require less server computing for a scene than a centralized approach. Instead of sending the raw data, our approach requires users to send a 3D representation, distributing the high computation cost of training centralized NeRFs between the users. It learns photorealistic scene representations by decomposing users’ 3D views into personal and global NeRFs and a novel optimally weighted aggregation of only the latter. We validate the advantage of our approach to learn NeRFs with photorealism and minimal server computation cost on structured synthetic and real-world photo tourism datasets. We further analyze how secure aggregation of global NeRFs in DecentNeRF minimizes the undesired reconstruction of personal content by the server. 

The microstructure, phase behavior, mechanical properties, and corrosion properties of a series of Al10Cr15(Fe3Mn)75−x(Ni)x medium-entropy alloys (MEAs) spanning 0–20 at% Ni were studied to elucidate the chemistry-structure-property relationship of this system as a function of Ni content. This work demonstrates that from an initial BCC phase Al10Cr15(Fe3Mn)75 MEA, Ni additions of 5 and 10 at% result in the formation of ordered B2-phase precipitates due to interaction of Ni with Al, resulting in high hardness (∼475 HV). Further Ni addition to 15 at% leads to a dual-phase FCC+BCC structure, with B2 phase precipitates distributed in the BCC matrix relatively rich in Al and Ni but depleted in Cr. This dual-phase structure has a high yield strength (YS) of 600 MPa with a total elongation of 15%. Additionally, the B2 precipitates in BCC phase serve as preferential sites for corrosion in 0.6 M NaCl. Increasing Ni content to 20 at% results in lower YS of 300 MPa, but a significant improvement in ductility and corrosion resistance due to the increased FCC phase fraction. 

Although there have been a few reports of macromammals from the Siwalik site of Dunera, no micromammals have yet been described. Recently, a diverse micromammal fossil assemblage represented by isolated teeth has been recovered from Dunera. The specimens are identified as a murine similar to Progonomys hussaini (Progonomys cf. hussaini), the ctenodactylid Sayimys sivalensis, the cricetid Democricetodon fejfari, and a sciurid, cf. Tamias urialis. These rodents are documented for the first time from this region of the Indian Siwaliks. Based on the biostratigraphic ranges of these rodents from well-dated localities in the Siwaliks of Pakistan along with previously collected magnetostratigraphic data, Dunera locality best correlates to between ~11-10 Ma (early Late Miocene), approximately equivalent to the lower half of the Nagri Formation. Previous magneto-stratigraphic data suggests that an age closer to ~11 Ma is more likely, perhaps sampling a temporal gap in the known Siwaliks micromammal record and extending the age ranges of P. hussaini and T. urialis by a few hundred thousand years.