12 July - 25 July 2019
1814 new papers
Etymo Newsletter provides the latest development in machine learning research,
including the most popular datasets and the most trending papers in the past two weeks.
If you like this newsletter, you can subscribe to our fortnightly newsletters here.
Fortnight Summary
Popular datasets include MNIST, ImageNet, CIFAR-10, KITTI, COCO, and Cityscapes. Our trending phrases are Answer Set, for which we have five articles, Perception Bias, and Sound Events. In our trending papers section, we include a paper on the analogy between neural networks and the learning and development in human brains, a new SGD optimizer called Lookahead, and a new image dataset to significantly improve the accuracy and robustness of classifiers.Popular Datasets
Here are most mentioned datasets over the last two weeks.
| Name | Type | Number of Papers |
|---|---|---|
| MNIST | Handwritten Digits | 58 |
| ImageNet | Image Dataset | 46 |
| CIFAR-10 | Tiny Image Dataset in 10 Classes | 28 |
| KITTI | Autonomous Driving | 26 |
| COCO | Common Objects in Context | 20 |
| Cityscapes | Images from 50 different cities | 17 |
Trending Phrases
In this section, we present a list of phrases that appeared significantly more in this newsletter than the previous newsletters.
- Answer Set:
- Beyond NP: Quantifying over Answer Sets
- Better Paracoherent Answer Sets with Less Resources
- Abstract Solvers for Computing Cautious Consequences of ASP programs
- Partial Compilation of ASP Programs
- Incremental Answer Set Programming with Overgrounding
- Homophily and minority size explain perception biases in social networks
- Language Modelling for Sound Event Detection with Teacher Forcing and Scheduled Sampling
Perception Bias:
Sound Events:
Etymo Trending
Presented below is a list of the most trending papers added in the last two weeks.
-
What does it mean to understand a neural network?:
The authors give an analogy comparing neural networks to the development and learning in brains. -
Lookahead Optimizer: k steps forward, 1 step back:
Lookahead optimizer is a new optimization algorithm of stochastic gradient descent (SGD). Lookahead is orthogonal to a) adaptive learning rate schemes, such as AdaGrad and Adam, and b) accelerated schemes, such as heavy-ball and Nesterov momentum. The new Lookahead algorithm iteratively updates two sets of weights. -
Natural Adversarial Examples:
This is a new dataset to measure classifier robustness. This dataset contains 7,500 curated natural adversarial examples, which are real-world, unmodified, and naturally occurring examples.