• Aucun résultat trouvé

Biologically-plausible learning algorithms can scale to large datasets

N/A
N/A
Protected

Academic year: 2021

Partager "Biologically-plausible learning algorithms can scale to large datasets"

Copied!
10
0
0

Texte intégral

Loading

Figure

Figure 1: a, Top-1 and b, top-5 validation error on ImageNet for ResNet-18 and AlexNet trained with different learning algorithms
Table 1: ImageNet 1-crop validation accuracy of networks trained with different algorithms, all 50 epochs
Figure 3: a, During training with sign-symmetry, alignment angles between feedforward weights W and feedback weights sign(W ) decreased in the last 3 layers but increased in early layers, whereas during training with backpropagation, the analogous alignmen
Figure 4: The specific wiring required for sign-symmetric feedback can be achieved using axonal guidance by specific receptor-ligand recognition

Références

Documents relatifs

HTMRL can explore rarely selected actions through the boosting mechanism also present in core HTM theory, which amplifies incoming signals of rarely activated output bits..

In this paper, we present the GeneTegra Alignment Tool (GT-Align), a practical implementation of the ASMOV ontology alignment algorithm within a Web-based

Furthermore, we apply Assani’s result to prove that the M¨ obius and Liouville functions are a good weight for the ho- mogeneous ergodic multilinear averages if the restriction of

D‟autres pays européens ont suivi et, en France en 1997, la conférence nationale de santé a insisté sur la nécessité d‟offrir aux malades cancéreux une

As we mentioned in the introduction, the lower bound can be obtained with the help of the coarea formula, like in [ABO03], and, before that, as seen in [ABL88].. I am indebted

Indeed, we prove that on any given Riemannian manifold, there exists families of densities such that the associated Cheeger constants are as small as desired while the

The EOM has been designed to calculate the position on the 3D space, yet the information is displayed to the Learner as an augmented path in 2D space; a 2D line that

Very recently, the author provided a simultaneously simple proof of Birkhoff ergodic theorem and Bourgain homogeneous ergodic bilinear theorem with an extension to polynomials