Zobrazeno 1 - 10
of 2 645
pro vyhledávání: '"Filippas, A."'
Structured sparsity is an efficient way to prune the complexity of modern Machine Learning (ML) applications and to simplify the handling of sparse data in hardware. In such cases, the acceleration of structured-sparse ML models is handled by sparse
Externí odkaz:
http://arxiv.org/abs/2402.10850
We prove a local unique continuation result for Schr\''odinger operators with time independent Lipschitz metrics and lower order terms which are Gevrey 2 in time and bounded in space. This implies global unique continuation from any open set in a con
Externí odkaz:
http://arxiv.org/abs/2401.14820
The widespread proliferation of deep learning applications has triggered the need to accelerate them directly in hardware. General Matrix Multiplication (GEMM) kernels are elemental deep-learning constructs and they inherently map onto Systolic Array
Externí odkaz:
http://arxiv.org/abs/2309.02969
Systolic Array (SA) architectures are well suited for accelerating matrix multiplications through the use of a pipelined array of Processing Elements (PEs) communicating with local connections and pre-orchestrated data movements. Even though most of
Externí odkaz:
http://arxiv.org/abs/2304.12691
Autor:
Filippas, Stathis, Tersenov, Alkis
We obtain estimates of all components of the velocity of a 3D rigid body moving in a viscous incompressible fluid without any symmetry restriction on the shape of the rigid body or the container. The estimates are in terms of suitable norms of the ve
Externí odkaz:
http://arxiv.org/abs/2304.11887
The acceleration of deep-learning kernels in hardware relies on matrix multiplications that are executed efficiently on Systolic Arrays (SA). To effectively trade off deep-learning training/inference quality with hardware cost, SA accelerators employ
Externí odkaz:
http://arxiv.org/abs/2304.01668
Autor:
Filippas, Spyridon
We construct a specific family of eigenfunctions for a Laplace operator with coefficients having a jump across an interface. These eigenfunctions have an exponential concentration arbitrarily close to the interface, and therefore could be considered
Externí odkaz:
http://arxiv.org/abs/2302.05215
Autor:
Filippas, Kostas
A new duality is proposed in four-dimensional flat space, which exchanges between spin and orbital degrees of freedom. This is motivated by a Hodge decomposition of the angular-momentum bivector for massive fields, along which spin and orbital angula
Externí odkaz:
http://arxiv.org/abs/2212.11340
Convolutional Neural Networks (CNNs) are the state-of-the-art solution for many deep learning applications. For maximum scalability, their computation should combine high performance and energy efficiency. In practice, the convolutions of each CNN la
Externí odkaz:
http://arxiv.org/abs/2211.12600
Autor:
Filippas, Spyridon
In this article we prove quantitative unique continuation results for wave operators of the form $\partial$ 2 t -- div(c(x)$\nabla$$\bullet$) where the scalar coefficient c is discontinuous across an interface of codimension one in a bounded domain o
Externí odkaz:
http://arxiv.org/abs/2210.04634