recentpopularlog in


« earlier   
subject category scheme for a global book trade
12 days ago by wilbsimpson
Flag Stories
Classification of flags based on colors, aspect rations, elements, etc, etc etc.
flags  classification 
15 days ago by drmeme
[1607.01759] Bag of Tricks for Efficient Text Classification
This paper explores a simple and efficient baseline for text classification. Our experiments show that our fast text classifier fastText is often on par with deep learning classifiers in terms of accuracy, and many orders of magnitude faster for training and evaluation. We can train fastText on more than one billion words in less than ten minutes using a standard multicore~CPU, and classify half a million sentences among~312K classes in less than a minute.
natural-language-processing  text-mining  classification  machine-learning  heuristics  representation  rather-interesting  neural-networks  to-simulate  consider:feature-discovery 
15 days ago by Vaguery
[1809.02104] Are adversarial examples inevitable?
A wide range of defenses have been proposed to harden neural networks against adversarial attacks. However, a pattern has emerged in which the majority of adversarial defenses are quickly broken by new attacks. Given the lack of success at generating robust defenses, we are led to ask a fundamental question: Are adversarial attacks inevitable? This paper analyzes adversarial examples from a theoretical perspective, and identifies fundamental bounds on the susceptibility of a classifier to adversarial attacks. We show that, for certain classes of problems, adversarial examples are inescapable. Using experiments, we explore the implications of theoretical guarantees for real-world problems and discuss how factors such as dimensionality and image complexity limit a classifier's robustness against adversarial examples.
classification  adversarial-examples  robustness  machine-learning  neural-networks  problems-with-continuous-embeddings-for-discrete-problems  to-write-about  to-simulate 
16 days ago by Vaguery
View of “Mapping-with”: The Politics of (Counter-)classification in OpenStreetMap | Cartographic Perspectives
By unpacking the theoretical work of Donna Haraway, I also argue for a return to the critical potential of feminist science and technology studies within cartography, signposted by the ongoing work of feminist and queer geographers such as Pavlovskaya (2018), Giesking (2018), Leszczynski and Elwood (2015), and Kwan (2007)—not simply as a tool for a feminist critique, but a way of remaking worlds, rather than just remaking maps. That mapping has troubles is not a new argument: significant empirical research has been undertaken documenting and advancing our understanding of the technopositional (Wilson 2017), tacit (McHaffie 2002), institutionalised (Gekker 2016), and politicised (Thatcher and Imaoka 2018) practices undertaken by cartographers, educators, and geographic information scientists. Furthermore, that the politics of mappings are based in situated knowledges (Wilmott 2016), embodied (Lin 2006), vernacular (Gerlach 2015), and taken up in the everyday (Del Casino and Hanna 2005) is also well documented within cartographic research.
mapping  cartography  ontology  haraway  classification 
17 days ago by shannon_mattern
Excavating Training Sets
An *excellent* paper from Kate Crawford and Trevor Paglen, exploring some horrific examples of bias in the popular ImageNet set of AI training data, and the history of inherent bias in classification schemes.

One thing that we found really hard in SpamAssassin was carefully curating training data. It was very easy to wind up with unreliable, garbage results by training the SpamAssassin scoreset from poorly-labelled data. It's jaw-dropping to consider that the past 10 years of AI research and real-world results may have been built atop a heap of bad training data, incorporating some unpleasant examples of bias....

You open up a database of pictures used to train artificial intelligence systems. At first, things seem straightforward. You’re met with thousands of images: apples and oranges, birds, dogs, horses, mountains, clouds, houses, and street signs. But as you probe further into the dataset, people begin to appear: cheerleaders, scuba divers, welders, Boy Scouts, fire walkers, and flower girls. Things get strange: A photograph of a woman smiling in a bikini is labeled a “slattern, slut, slovenly woman, trollop.” A young man drinking beer is categorized as an “alcoholic, alky, dipsomaniac, boozer, lush, soaker, souse.” A child wearing sunglasses is classified as a “failure, loser, non-starter, unsuccessful person.” You’re looking at the “person” category in a dataset called ImageNet, one of the most widely used training sets for machine learning.

Something is wrong with this picture. [....]

For the last two years, we have been studying the underlying logic of how images are used to train AI systems to “see” the world. We have looked at hundreds of collections of images used in artificial intelligence, from the first experiments with facial recognition in the early 1960s to contemporary training sets containing millions of images. Methodologically, we could call this project an archeology of datasets: we have been digging through the material layers, cataloguing the principles and values by which something was constructed, and analyzing what normative patterns of life were assumed, supported, and reproduced. By excavating the construction of these training sets and their underlying structures, many unquestioned assumptions are revealed. These assumptions inform the way AI systems work — and fail — to this day.
ai  bias  training  ml  algorithms  imagenet  classification  kate-crawford  trevor-paglen 
25 days ago by jm
Rare Sound Event Detection Using Deep Learning and Data Augmentation
There is an increasing interest in smart environment and a growing adoption of smart devices. Smart assistants such as Google Home and Amazon Alexa, although focus on speech, could be extended to identify domestic events in real-time to provide more and better smart functions. Sound event detection aims to detect multiple target sound events that may happen simultaneously. The task is challenging due to the overlapping of sound events, the highly imbalanced nature of target and non-target data, and the complicated real-world background noise. In this paper, we proposed a unified approach that takes advantages of both the deep learning and data augmentation. A convolutional neural network (CNN) was combined with a feed-forward neural network (FNN) to improve the detection performance, and a dynamic time warping based data augmentation (DA) method was proposed to address the data imbalance problem. Experiments on several datasets showed a more than 7% increase in accuracy compared to the state-of-the-art approaches.
audio  classification  data-augmentation 
26 days ago by arsyed

Copy this bookmark:

to read