Top 10 Arxiv Papers Today
4.09 Mikeys
#1. Possible superconductivity in brain
P. Mikheenko
The unprecedented power of the brain suggests that it may process information
quantum-mechanically. Since quantum processing is already achieved in
superconducting quantum computers, it may imply that superconductivity is the
basis of quantum computation in brain too. Superconductivity could also be
responsible for long-term memory. Following these ideas, the paper reviews the
progress in the search for superconductors with high critical temperature and
tries to answer the question about the superconductivity in brain. It focuses
on recent electrical measurements of brain slices, in which graphene was used
as a room-temperature quantum mediator, and argues that these measurements
could be interpreted as providing evidence of superconductivity in the neural
network of mammalian brains. The estimated critical temperature of
superconducting network in brain is rather high: 2063 plus-minus 114 K. A
similar critical temperature was predicted in the Little's model for
one-dimensional organic chains linked to certain molecular complexes....
more |
pdf
| html
Figures
Tweets
S_Kakita:
Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
adhara_mathphys:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
adhara_mathphys:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
47rou:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
47rou:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
silver_thinfilm:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
hoshutaro3:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
RinochiNOeSIS:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
plus7:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
S_Kakita:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
jaialkdanel:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
tjmlab:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
tjmlab:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
hf_and_beyond:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
PhiHamilton:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
dette_iu_san:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
deluwater:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
QM_phys_kyoto:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
QM_phys_kyoto:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
D_Plius:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
CoronzonX:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
superbradyon:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
cultivatetsubo:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
solid_kumaaa:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
Mopepe51:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
ray_yut:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
ray_yut:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
pax_lucis:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
fox_aki310ooooo:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
bemoroid:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
s_hskz:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
NouminEngineer:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
wakame_Kanan:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
HrmsTrsmgsts:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
HrmsTrsmgsts:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
68ruketa:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
h_iwaoki:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
Kotomo1020:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
qptxyz:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
00001V:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
102Dvd:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
mtoiiotm:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
nr11235:
RT @S_Kakita: Fig.3 が強すぎるw
面白 (トンデモ) 超伝導業界に新たな風を吹き込む研究だな……>脳内超伝導
https://t.co/IruM8CNy61 https://t.co/2YlleBwTOR
oknd1_:
RT @tjmlab: Possible superconductivity in brain
脳内室温超伝導の可能性ｷﾀ━━━━(ﾟ∀ﾟ)━━━━!!
https://t.co/N6yfAIrwHG
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 1
Total Words: 7303
Unqiue Words: 2223
3.13 Mikeys
#2. Denoising Weak Lensing Mass Maps with Deep Learning
Masato Shirasaki, Naoki Yoshida, Shiro Ikeda
Weak gravitational lensing is a powerful probe of the large-scale cosmic
matter distribution. Wide-field galaxy surveys allow us to generate the
so-called weak lensing maps, but actual observations suffer from noise due to
imperfect measurement of galaxy shape distortions and to the limited number
density of the source galaxies. In this paper, we explore a deep-learning
approach to reduce the noise. We develop an image-to-image translation method
with conditional adversarial networks (CANs), which learn efficient mapping
from an input noisy weak lensing map to the underlying noise field. We train
the CANs using 30000 image pairs obtained from 1000 ray-tracing simulations of
weak gravitational lensing. We show that the trained CANs reproduce the true
one-point probability distribution function of the noiseless lensing map with a
bias less than $1\sigma$ on average, where $\sigma$ is the statistical error.
Since a number of model parameters are used in our CANs, our method has
additional error budgets when reconstructing the summary...
more |
pdf
| html
Figures
Tweets
arxiv_cs_LG:
Denoising Weak Lensing Mass Maps with Deep Learning. Masato Shirasaki, Naoki Yoshida, and Shiro Ikeda https://t.co/gBhXy9sCtA
Github
TensorFlow implementation of "Image-to-Image Translation Using Conditional Adversarial Networks".
Stargazers: 661
Subscribers: 33
Subscribers: 33
Forks: 219
Open Issues: 27
Open Issues: 27
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 3
Total Words: 9341
Unqiue Words: 2180
2.801 Mikeys
#3. Speech and Speaker Recognition from Raw Waveform with SincNet
Mirco Ravanelli, Yoshua Bengio
Deep neural networks can learn complex and abstract representations, that are
progressively obtained by combining simpler ones. A recent trend in speech and
speaker recognition consists in discovering these representations starting from
raw audio samples directly. Differently from standard hand-crafted features
such as MFCCs or FBANK, the raw waveform can potentially help neural networks
discover better and more customized representations. The high-dimensional raw
inputs, however, can make training significantly more challenging. This paper
summarizes our recent efforts to develop a neural architecture that efficiently
processes speech from audio waveforms. In particular, we propose SincNet, a
novel Convolutional Neural Network (CNN) that encourages the first layer to
discover meaningful filters by exploiting parametrized sinc functions. In
contrast to standard CNNs, which learn all the elements of each filter, only
low and high cutoff frequencies of band-pass filters are directly learned from
data. This inductive bias offers a...
more |
pdf
| html
Figures
Tweets
Github
SincNet is a neural architecture for efficiently processing raw audio samples.
Stargazers: 172
Subscribers: 15
Subscribers: 15
Forks: 43
Open Issues: 1
Open Issues: 1
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 2
Total Words: 4401
Unqiue Words: 1599
2.79 Mikeys
#4. HD219666b: a hot-Neptune from TESS Sector 1
M. Esposito, D. J. Armstrong, D. Gandolfi, V. Adibekyan, M. Fridlund, N. C. Santos, J. H. Livingston, E. Delgado Mena, L. Fossati, J. Lillo-Box, O. Barragán, D. Barrado, P. E. Cubillos, B. Cooke, A. B. Justesen, F. Meru, R. F. Díaz, F. Dai, L. D. Nielsen, C. M. Persson, P. J. Wheatley, A. P. Hatzes, V. Van Eylen, M. M. Musso, R. Alonso, P. Beck, S. C. C. Barros, D. Bayliss, A. S. Bonomo, F. Bouchy, D. J. A. Brown, E. Bryant, J. Cabrera, W. D. Cochran, S. Csizmadia, H. Deeg, O. Demangeon, M. Deleuil, X. Dumusque, P. Eigmüller, M. Endl, A. Erikson, F. Faedi, P. Figueira, A. Fukui, S. Grziwa, E. W. Guenther, D. Hidalgo, M. Hjorth, T. Hirano, S. Hojjatpanah, E. Knudstrup, J. Korth, K. W. F. Lam, J. de Leon, M. N. Lund, R. Luque, S. Mathur, P. Montañés Rodríguez, N. Narita, D. Nespral, P. Niraula, G. Nowak, H. P. Osborn, E. Pallé, M. Pätzold, D. Pollacco, J. Prieto-Arranz, H. Rauer, S. Redfield, I. Ribas, S. G. Sousa, A. M. S. Smith, M. Tala-Pinto, S. Udry, J. N. Winn
We report the discovery of a transiting planet orbiting the old and inactive
G7 dwarf star HD219666 (Mstar = 0.92 +/- 0.03 MSun, Rstar = 1.03 +/- 0.03 RSun,
tau_star = 10 +/- 2 Gyr). With a mass of Mb = 16.6 +/- 1.3 MEarth, a radius of
Rb = 4.71 +/- 0.17 REarth, and an orbital period of P ~ 6 days, HD219666b is a
new member of a rare class of exoplanets: the hot-Neptunes. The Transiting
Exoplanet Survey Satellite (TESS) observed HD219666 (also known as TOI-118) in
its Sector 1 and the light curve shows four transit-like events, equally spaced
in time. We confirmed the planetary nature of the candidate by gathering
precise radial velocity measurements with HARPS@ESO3.6m. We used the co-added
HARPS spectrum to derive the host star fundamental parameters (Teff = 5527 +/-
65 K, log g = 4.40 +/- 0.11 (cgs), [Fe/H]= 0.04 +/- 0.04 dex, log R'HK = -5.07
+/- 0.03), as well as the abundances of many volatile and refractory elements.
The host star brightness (V = 9.9) makes it suitable for further
characterization by means of in-transit...
more |
pdf
| html
Figures
None.
Tweets
scimichael:
HD219666b: a hot-Neptune from TESS Sector 1
"report the discovery of a transiting planet orbiting the old and inactive G7 dwarf star HD219666.. mass of Mb = 16.6 +/- 1.3 MEarth, a radius of Rb = 4.71 +/- 0.17 REarth, & an orbital period of P ~ 6 days.."
https://t.co/uYLNwFbvGn
oscaribv:
Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https://t.co/FE3ge47qKc
SpaceToday1:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
Lunarheritage:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
DrJorgeMelendez:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
CosmicRami:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
NESTA_US:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
safreitas_c:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
_Gustavobc:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
SnowRaptor:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
resonances:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
mwgc1995:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
astro_who:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
Mabel_Valerdi:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
procrastiwalter:
RT @oscaribv: Check our new paper on the discovery and characterization of another TESS planet: HD 219666 b!
https://t.co/pQ2wP2MMEQ https…
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 76
Total Words: 0
Unqiue Words: 0
2.518 Mikeys
#5. Online gradient-based mixtures for transfer modulation in meta-learning
Ghassen Jerfel, Erin Grant, Thomas L. Griffiths, Katherine Heller
Learning-to-learn or meta-learning leverages data-driven inductive bias to
increase the efficiency of learning on a novel task. This approach encounters
difficulty when transfer is not mutually beneficial, for instance, when tasks
are sufficiently dissimilar or change over time. Here, we use the connection
between gradient-based meta-learning and hierarchical Bayes (Grant et al.,
2018) to propose a mixture of hierarchical Bayesian models over the parameters
of an arbitrary function approximator such as a neural network. Generalizing
the model-agnostic meta-learning (MAML) algorithm (Finn et al., 2017), we
present a stochastic expectation maximization procedure to jointly estimate
parameter initializations for gradient descent as well as a latent assignment
of tasks to initializations. This approach better captures the diversity of
training tasks as opposed to consolidating inductive biases into a single set
of hyperparameters. Our experiments demonstrate better generalization
performance on the standard miniImageNet benchmark for...
more |
pdf
| html
Figures
None.
Tweets
Miles_Brundage:
"Online gradient-based mixtures for transfer modulation in meta-learning," Jerfel and @ermgrant et al.: https://t.co/vBRtfYz16L
BrundageBot:
Online gradient-based mixtures for transfer modulation in meta-learning. Ghassen Jerfel, Erin Grant, Thomas L. Griffiths, and Katherine Heller https://t.co/gLYAJ5BYAM
M157q_News_RSS:
Online gradient-based mixtures for transfer modulation in meta-learning. (arXiv:1812.06080v1 [cs.LG])
https://t.co/VcNZTFkOVu
Learning-to-learn or meta-learning leverages data-driven inductive bias to increase the efficiency of learning on a novel task. This approach encounters d
arxivml:
"Online gradient-based mixtures for transfer modulation in meta-learning",
Ghassen Jerfel, Erin Grant, Thomas L． Gr…
https://t.co/CKmdWCZFh8
arxiv_cs_LG:
Online gradient-based mixtures for transfer modulation in meta-learning. Ghassen Jerfel, Erin Grant, Thomas L. Griffiths, and Katherine Heller https://t.co/9MX33PGLYQ
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 4
Total Words: 7399
Unqiue Words: 2199
2.498 Mikeys
#6. Rethinking Layer-wise Feature Amounts in Convolutional Neural Network Architectures
Martin Mundt, Sagnik Majumder, Tobias Weis, Visvanathan Ramesh
We characterize convolutional neural networks with respect to the relative
amount of features per layer. Using a skew normal distribution as a
parametrized framework, we investigate the common assumption of monotonously
increasing feature-counts with higher layers of architecture designs. Our
evaluation on models with VGG-type layers on the MNIST, Fashion-MNIST and
CIFAR-10 image classification benchmarks provides evidence that motivates
rethinking of our common assumption: architectures that favor larger early
layers seem to yield better accuracy.
more |
pdf
| html
Figures
Tweets
BrundageBot:
Rethinking Layer-wise Feature Amounts in Convolutional Neural Network Architectures. Martin Mundt, Sagnik Majumder, Tobias Weis, and Visvanathan Ramesh https://t.co/gKjrNsU9lq
arxivml:
"Rethinking Layer-wise Feature Amounts in Convolutional Neural Network Architectures",
Martin Mundt, Sagnik Majumde…
https://t.co/scX7V6Tzvw
arxiv_cs_LG:
Rethinking Layer-wise Feature Amounts in Convolutional Neural Network Architectures. Martin Mundt, Sagnik Majumder, Tobias Weis, and Visvanathan Ramesh https://t.co/xazYk40PdZ
Github
PyTorch implementation of our paper "Rethinking Layer-wise Feature Amounts in Convolutional Neural Network Architectures"
Stargazers: 0
Subscribers: 1
Subscribers: 1
Forks: 0
Open Issues: 0
Open Issues: 0
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 4
Total Words: 2455
Unqiue Words: 1030
2.427 Mikeys
#7. A Probe into Understanding GAN and VAE models
Jingzhao Zhang, Lu Mi, Macheng Shen
Both generative adversarial network models and variational autoencoders have
been widely used to approximate probability distributions of datasets. Although
they both use parametrized distributions to approximate the underlying data
distribution, whose exact inference is intractable, their behaviors are very
different. In this report, we summarize our experiment results that compare
these two categories of models in terms of fidelity and mode collapse. We
provide a hypothesis to explain their different behaviors and propose a new
model based on this hypothesis. We further tested our proposed model on MNIST
dataset and CelebA dataset.
more |
pdf
| html
Figures
Tweets
arxiv_org:
A Probe into Understanding GAN and VAE models. https://t.co/QwCm74Z4L0 https://t.co/nKQIiMOheW
BrundageBot:
A Probe into Understanding GAN and VAE models. Jingzhao Zhang, Lu Mi, and Macheng Shen https://t.co/hHO5Q2bHXC
arxivml:
"A Probe into Understanding GAN and VAE models",
Jingzhao Zhang, Lu Mi, Macheng Shen
https://t.co/nHaW4tWGEW
yapp1e:
A Probe into Understanding GAN and VAE models. (arXiv:1812.05676v1 [cs.LG]) https://t.co/nT6hE5yMEn
Both generative adversarial network models and variational autoencoders have been widely used to approximate probability distributions of datasets. Although they both use parame…
udmrzn:
RT @arxiv_org: A Probe into Understanding GAN and VAE models. https://t.co/QwCm74Z4L0 https://t.co/nKQIiMOheW
RexDouglass:
RT @arxiv_org: A Probe into Understanding GAN and VAE models. https://t.co/QwCm74Z4L0 https://t.co/nKQIiMOheW
morioka:
RT @arxiv_org: A Probe into Understanding GAN and VAE models. https://t.co/QwCm74Z4L0 https://t.co/nKQIiMOheW
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 3
Total Words: 4580
Unqiue Words: 1352
2.388 Mikeys
#8. Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem
Matthias Hein, Maksym Andriushchenko, Julian Bitterwolf
Classifiers used in the wild, in particular for safety-critical systems,
should not only have good generalization properties but also should know when
they don't know, in particular make low confidence predictions far away from
the training data. We show that ReLU type neural networks which yield a
piecewise linear classifier function fail in this regard as they produce almost
always high confidence predictions far away from the training data. For bounded
domains like images we propose a new robust optimization technique similar to
adversarial training which enforces low confidence predictions far away from
the training data. We show that this technique is surprisingly effective in
reducing the confidence of predictions far away from the training data while
maintaining high confidence predictions and similar test error on the original
classification task compared to standard training.
more |
pdf
| html
Figures
Tweets
arxiv_org:
Why ReLU networks yield high-confidence predictions far away from the training data and h... https://t.co/a8traWSLvc https://t.co/tuu1GWY8Pl
BrundageBot:
Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem. Matthias Hein, Maksym Andriushchenko, and Julian Bitterwolf https://t.co/FiprsV8D5k
arxivml:
"Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the proble…
https://t.co/IXhPHWavZY
yapp1e:
Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem. (arXiv:1812.05720v1 [cs.LG]) https://t.co/y8SlO18j87
Classifiers used in the wild, in particular for safety-critical systems, should not only have good generali…
disigandalf:
RT @arxiv_org: Why ReLU networks yield high-confidence predictions far away from the training data and h... https://t.co/a8traWSLvc https:/…
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 3
Total Words: 8384
Unqiue Words: 2141
2.382 Mikeys
#9. Scaling shared model governance via model splitting
Miljan Martic, Jan Leike, Andrew Trask, Matteo Hessel, Shane Legg, Pushmeet Kohli
Currently the only techniques for sharing governance of a deep learning model
are homomorphic encryption and secure multiparty computation. Unfortunately,
neither of these techniques is applicable to the training of large neural
networks due to their large computational and communication overheads. As a
scalable technique for shared model governance, we propose splitting deep
learning model between multiple parties. This paper empirically investigates
the security guarantee of this technique, which is introduced as the problem of
model completion: Given the entire training data set or an environment
simulator, and a subset of the parameters of a trained deep learning model, how
much training is required to recover the model's original performance? We
define a metric for evaluating the hardness of the model completion problem and
study it empirically in both supervised learning on ImageNet and reinforcement
learning on Atari and DeepMind~Lab. Our experiments show that (1) the model
completion problem is harder in reinforcement...
more |
pdf
| html
Figures
None.
Tweets
BrundageBot:
Scaling shared model governance via model splitting. Miljan Martic, Jan Leike, Andrew Trask, Matteo Hessel, Shane Legg, and Pushmeet Kohli https://t.co/0IkHMhAUd6
M157q_News_RSS:
Scaling shared model governance via model splitting. (arXiv:1812.05979v1 [cs.LG])
https://t.co/4RzdwL6nb4
Currently the only techniques for sharing governance of a deep learning model are homomorphic encryption and secure multiparty computation. Unfortunately, neither of these te
arxivml:
"Scaling shared model governance via model splitting",
Miljan Martic, Jan Leike, Andrew Trask, Matteo Hessel, Shane…
https://t.co/qiZHlPcQwW
arxiv_cs_LG:
Scaling shared model governance via model splitting. Miljan Martic, Jan Leike, Andrew Trask, Matteo Hessel, Shane Legg, and Pushmeet Kohli https://t.co/mc2hcYKvXa
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 6
Total Words: 9550
Unqiue Words: 2475
2.382 Mikeys
#10. Why ReLU Units Sometimes Die: Analysis of Single-Unit Error Backpropagation in Neural Networks
Scott C. Douglas, Jiutian Yu
Recently, neural networks in machine learning use rectified linear units
(ReLUs) in early processing layers for better performance. Training these
structures sometimes results in "dying ReLU units" with near-zero outputs. We
first explore this condition via simulation using the CIFAR-10 dataset and
variants of two popular convolutive neural network architectures. Our
explorations show that the output activation probability Pr[y>0] is generally
less than 0.5 at system convergence for layers that do not employ skip
connections, and this activation probability tends to decrease as one
progresses from input layer to output layer. Employing a simplified model of a
single ReLU unit trained by a variant of error backpropagation, we then perform
a statistical convergence analysis to explore the model's evolutionary
behavior. Our analysis describes the potentially-slower convergence speeds of
dying ReLU units, and this issue can occur regardless of how the weights are
initialized.
more |
pdf
| html
Figures
None.
Tweets
BrundageBot:
Why ReLU Units Sometimes Die: Analysis of Single-Unit Error Backpropagation in Neural Networks. Scott C. Douglas and Jiutian Yu https://t.co/HrlYDFZaWL
M157q_News_RSS:
Why ReLU Units Sometimes Die: Analysis of Single-Unit Error Backpropagation in Neural Networks. (arXiv:1812.05981v1 [cs.LG])
https://t.co/oKhmyW1WTn
Recently, neural networks in machine learning use rectified linear units (ReLUs) in early processing layers for better performance.
arxivml:
"Why ReLU Units Sometimes Die: Analysis of Single-Unit Error Backpropagation in Neural Networks",
Scott C． Douglas,…
https://t.co/Y3MReb35Oz
arxiv_cs_LG:
Why ReLU Units Sometimes Die: Analysis of Single-Unit Error Backpropagation in Neural Networks. Scott C. Douglas and Jiutian Yu https://t.co/JO4Me4TVaG
Github
None.
Youtube
None.
Other stats
Sample Sizes : None.
Authors: 2
Total Words: 0
Unqiue Words: 0
About
Assert is a website where the best academic papers on arXiv (computer science, math, physics), bioRxiv (biology), BITSS (reproducibility), EarthArXiv (earth science), engrXiv (engineering), LawArXiv (law), PsyArXiv (psychology), SocArXiv (social science), and SportRxiv (sport research) bubble to the top each day.
Papers are scored (in real-time) based on how verifiable they are (as determined by their Github repos) and how interesting they are (based on Twitter).
To see top papers, follow us on twitter @assertpub_ (arXiv), @assert_pub (bioRxiv), and @assertpub_dev (everything else).
To see beautiful figures extracted from papers, follow us on Instagram.
Tracking 68,758 papers.
Search
Sort results based on if they are interesting or reproducible.
Interesting
Reproducible