User profiles for D. Soudry
Daniel SoudryAssociate Professor Verified email at technion.ac.il Cited by 16607 |
Evolution of Tethyan phosphogenesis along the northern edges of the Arabian–African shield during the Cretaceous–Eocene as deduced from temporal variations of …
D Soudry, CR Glenn, Y Nathan, I Segal… - Earth-Science …, 2006 - Elsevier
The evolution of Tethyan phosphogenesis during the Cretaceous–Eocene is examined to try
to explain fluctuations of phosphogenesis through time, and whether or not they reflect long-…
to explain fluctuations of phosphogenesis through time, and whether or not they reflect long-…
Binarized neural networks: Training deep neural networks with weights and activations constrained to+ 1 or-1
We introduce a method to train Binarized Neural Networks (BNNs) - neural networks with
binary weights and activations at run-time. At training-time the binary weights and activations …
binary weights and activations at run-time. At training-time the binary weights and activations …
Binarized neural networks
We introduce a method to train Binarized Neural Networks (BNNs)-neural networks with
binary weights and activations at run-time. At train-time the binary weights and activations are …
binary weights and activations at run-time. At train-time the binary weights and activations are …
Quantized neural networks: Training neural networks with low precision weights and activations
The principal submatrix localization problem deals with recovering a K × K principal
submatrix of elevated mean µ in a large n × n symmetric matrix subject to additive standard …
submatrix of elevated mean µ in a large n × n symmetric matrix subject to additive standard …
[PDF][PDF] Simultaneous denoising, deconvolution, and demixing of calcium imaging data
… d We present a new method for analyzing large-scale calcium imaging datasets … Panel C
shows the inferred spiking signals for both methods and panel D displays the recovered traces …
shows the inferred spiking signals for both methods and panel D displays the recovered traces …
Train longer, generalize better: closing the generalization gap in large batch training of neural networks
… During the initial training phase, to reach a minima of "width" d the weight vector wt has to
travel at least a distance d, and this takes a long time – about exp(d) iterations. Thus, to reach …
travel at least a distance d, and this takes a long time – about exp(d) iterations. Thus, to reach …
The implicit bias of gradient descent on separable data
… Lastly, we define P1 ∈ Rd×d as the orthogonal projection matrix5 to the subspace spanned
by the support vectors (the columns of XS), and P1 = I − P1 as the complementary projection …
by the support vectors (the columns of XS), and P1 = I − P1 as the complementary projection …
Post training 4-bit quantization of convolutional networks for rapid-deployment
…, Y Nahshan, D Soudry - Advances in Neural …, 2019 - proceedings.neurips.cc
Convolutional neural networks require significant memory bandwidth and storage for
intermediate computations, apart from substantial computing resources. Neural network …
intermediate computations, apart from substantial computing resources. Neural network …
Implicit bias of gradient descent on linear convolutional networks
We show that gradient descent on full-width linear convolutional networks of depth $ L $
converges to a linear predictor related to the $\ell_ {2/L} $ bridge penalty in the frequency …
converges to a linear predictor related to the $\ell_ {2/L} $ bridge penalty in the frequency …
Characterizing implicit bias in terms of optimization geometry
We study the bias of generic optimization methods, including Mirror Descent, Natural
Gradient Descent and Steepest Descent with respect to different potentials and norms, when …
Gradient Descent and Steepest Descent with respect to different potentials and norms, when …