No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Exploiting Class Learnability in Noisy Data
Abstract
In many domains, collecting sufficient labeled training data for supervised machine learning requires easily accessible but noisy sources, such as crowdsourcing services or tagged Web data. Noisy labels occur frequently in data sets harvested via these means, sometimes resulting in entire classes of data on which learned classifiers generalize poorly. For real world applications, we argue that it can be beneficial to avoid training on such classes entirely. In this work, we aim to explore the classes in a given data set, and guide supervised training to spend time on a class proportional to its learnability. By focusing the training process, we aim to improve model generalization on classes with a strong signal. To that end, we develop an online algorithm that works in conjunction with classifier and training algorithm, iteratively selecting training data for the classifier based on how well it appears to generalize on each class. Testing our approach on a variety of data sets, we show our algorithm learns to focus on classes for which the model has low generalization error relative to strong baselines, yielding a classifier with good performance on learnable classes.
... Finally, on the border between global and local post-hoc explanation approaches lie various methods that create simplified representations of the policy or the environment (or its observed version) and then use them to generate local explanations. For instance, in [9,10] authors use NN to generate a graph representation of scenes (images, for instance) that can be later used by a reasoning engine. Another way of doing it is to build a full Markov decision process and traverse it as a graph from the query state to the main reward state [12]. ...
The advancement on explainability techniques is quite relevant in the field of Reinforcement Learning (RL) and its applications can be beneficial for the development of intelligent agents that are understandable by humans and are able cooperate with them. When dealing with Deep RL some approaches already exist in the literature, but a common problem is that it can be tricky to define whether the explanations generated for an agent really reflect the behaviour of the trained agent. In this work we will apply an approach for explainability based on the creation of a Policy Graph (PG) that represents the agent’s behaviour. Our main contribution is a way to measure the similarity between the explanations and the agent’s behaviour, by building another agent that follows a policy based on the explainability method and comparing the behaviour of both agents.
Molten pool characteristics have a significant effect on printing quality in laser powder bed fusion (PBF), and quantitative predictions of printing parameters and molten pool dimensions are critical to the intelligent control of the complex processes in PBF. Thus far, bidirectional predictions of printing parameters and molten pool dimensions have been challenging due to the highly nonlinear correlations involved. To address this issue, we integrate an experiment on molten pool characteristics, a mechanistic model, and deep learning to achieve both forward and inverse predictions of key parameters and molten pool characteristics during laser PBF. The experiment provides fundamental data, the mechanistic model significantly augments the dataset, and the multilayer perceptron (MLP) deep learning model predicts the molten pool dimensions and process parameters based on the dataset built from the experiment and the mechanistic model. The results show that bidirectional predictions of the molten pool dimensions and process parameters can be realized, with the highest prediction accuracies approaching 99.9% and mean prediction accuracies of over 90.0%. Moreover, the prediction accuracy of the MLP model is closely related to the characteristics of the dataset—that is, the learnability of the dataset has a crucial impact on the prediction accuracy. The highest prediction accuracy is 97.3% with enhancement of the dataset via the mechanistic model, while the highest prediction accuracy is 68.3% when using only the experimental dataset. The prediction accuracy of the MLP model largely depends on the quality of the dataset as well. The research results demonstrate that bidirectional predictions of complex correlations using MLP are feasible for laser PBF, and offer a novel and useful framework for the determination of process conditions and outcomes for intelligent additive manufacturing.
In this article, we propose a novel method to simultaneously solve the data problem of dirty quality and poor quantity for person reidentification (ReID). Dirty quality refers to the wrong labels in image annotations. Poor quantity means that some identities have very few images (FewIDs). Training with these mislabeled data or FewIDs with triplet loss will lead to low generalization performance. To solve the label error problem, we propose a weighted label correction based on cross-entropy (wLCCE) strategy. Specifically, according to the influence range of the wrong labels, we first classify the mislabeled images into point label error and set label error. Then, we propose a weighted triplet loss (WTL) to correct the two label errors, respectively. To alleviate the poor quantity issue, we propose a feature simulation based on autoencoder (FSAE) method to generate some virtual samples for FewID. For the authenticity of the simulated features, we transfer the difference pattern of identities with multiple images (MultIDs) to FewIDs by training an autoencoder (AE)-based simulator. In this way, the FewIDs obtain richer expressions to distinguish from other identities. By dealing with a dirty and poor data problem, we can learn more robust ReID models using the triplet loss. We conduct extensive experiments on two public person ReID datasets: 1) Market-1501 and 2) DukeMTMC-reID, to verify the effectiveness of our approach.
We introduce the "exponential linear unit" (ELU) which speeds up learning in deep neural networks and leads to higher classification accuracies. Like rectified linear units (ReLUs), leaky ReLUs (LReLUs) and parametrized ReLUs (PReLUs), ELUs alleviate the vanishing gradient problem via the identity for positive values. However, ELUs have improved learning characteristics compared to the units with other activation functions. In contrast to ReLUs, ELUs have negative values which allows them to push mean unit activations closer to zero like batch normalization but with lower computational complexity. Mean shifts toward zero speed up learning by bringing the normal gradient closer to the unit natural gradient because of a reduced bias shift effect. While LReLUs and PReLUs have negative values, too, they do not ensure a noise-robust deactivation state. ELUs saturate to a negative value with smaller inputs and thereby decrease the forward propagated variation and information. Therefore, ELUs code the degree of presence of particular phenomena in the input, while they do not quantitatively model the degree of their absence. In experiments, ELUs lead not only to faster learning, but also to significantly better generalization performance than ReLUs and LReLUs on networks with more than 5 layers. On CIFAR-100 ELUs networks significantly outperform ReLU networks with batch normalization while batch normalization does not improve ELU networks. ELU networks are among the top 10 reported CIFAR-10 results and yield the best published result on CIFAR-100, without resorting to multi-view evaluation or model averaging. On ImageNet, ELU networks considerably speed up learning compared to a ReLU network with the same architecture, obtaining less than 10% classification error for a single crop, single model network.
Despite progress in perceptual tasks such as image classification, computers still perform poorly on cognitive tasks such as image description and question answering. Cognition is core to tasks that involve not just recognizing, but reasoning about our visual world. However, models used to tackle the rich content in images for cognitive tasks are still being trained using the same datasets designed for perceptual tasks. To achieve success at cognitive tasks, models need to understand the interactions and relationships between objects in an image. When asked "What vehicle is the person riding?", computers will need to identify the objects in an image as well as the relationships riding(man, carriage) and pulling(horse, carriage) in order to answer correctly that "the person is riding a horse-drawn carriage". In this paper, we present the Visual Genome dataset to enable the modeling of such relationships. We collect dense annotations of objects, attributes, and relationships within each image to learn these models. Specifically, our dataset contains over 100K images where each image has an average of 21 objects, 18 attributes, and 18 pairwise relationships between objects. We canonicalize the objects, attributes, relationships, and noun phrases in region descriptions and questions answer pairs to WordNet synsets. Together, these annotations represent the densest and largest dataset of image descriptions, objects, attributes, relationships, and question answers.
We consider the sequential Bayesian optimization problem with bandit feedback, adopting a formulation that allows for the reward function to vary with time. We model the reward function using a Gaussian process whose evolution obeys a simple Markov model. We introduce two natural extensions of the classical Gaussian process upper confidence bound (GP-UCB) algorithm. The first, R-GP-UCB, resets GP-UCB at regular intervals. The second, TV-GP-UCB, instead forgets about old data in a smooth fashion. Our main contribution comprises of novel regret bounds for these algorithms, providing an explicit characterization of the trade-off between the time horizon and the rate at which the function varies. We illustrate the performance of the algorithms on both synthetic and real data, and we find the gradual forgetting of TV-GP-UCB to perform favorably compared to the sharp resetting of R-GP-UCB. Moreover, both algorithms significantly outperform classical GP-UCB, since it treats stale and fresh data equally.
Obtaining effective mid-level representations has become an increasingly important task in computer vision. In this paper, we propose a fully automatic algorithm which harvests visual concepts from a large number of Internet images (more than a quarter of a million) using text-based queries. Existing approaches to visual concept learning from Internet images either rely on strong supervision with detailed manual annotations or learn image-level classifiers only. Here, we take the advantage of having massive well organized Google and Bing image data, visual concepts (around 14, 000) are automatically exploited from images using word-based queries. Using the learned visual concepts, we show state-of-the-art performances on a variety of benchmark datasets, which demonstrate the effectiveness of the learned mid-level representations: being able to generalize well to general natural images. Our method shows significant improvement over the competing systems in image classification, including those with strong supervision.
The recently introduced continuous Skip-gram model is an efficient method for
learning high-quality distributed vector representations that capture a large
number of precise syntactic and semantic word relationships. In this paper we
present several extensions that improve both the quality of the vectors and the
training speed. By subsampling of the frequent words we obtain significant
speedup and also learn more regular word representations. We also describe a
simple alternative to the hierarchical softmax called negative sampling. An
inherent limitation of word representations is their indifference to word order
and their inability to represent idiomatic phrases. For example, the meanings
of "Canada" and "Air" cannot be easily combined to obtain "Air Canada".
Motivated by this example, we present a simple method for finding phrases in
text, and show that learning good vector representations for millions of
phrases is possible.
The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called ldquoImageNetrdquo, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.
Many applications require optimizing an unknown, noisy function that is expensive to evaluate. We formalize this task as a multi-armed bandit problem, where the payoff function is either sampled from a Gaussian process (GP) or has low RKHS norm. We resolve the important open problem of deriving regret bounds for this setting, which imply novel convergence rates for GP optimization. We analyze GP-UCB, an intuitive upper-confidence based algorithm, and bound its cumulative regret in terms of maximal information gain, establishing a novel connection between GP optimization and experimental design. Moreover, by bounding the latter in terms of operator spectra, we obtain explicit sublinear regret bounds for many commonly used covariance functions. In some important cases, our bounds have surprisingly weak dependence on the dimensionality. In our experiments on real sensor data, GP-UCB compares favorably with other heuristical GP optimization approaches.
We introduce a new descriptor for images which allows the construction of efficient and compact classifiers with good accuracy
on object category recognition. The descriptor is the output of a large number of weakly trained object category classifiers
on the image. The trained categories are selected from an ontology of visual concepts, but the intention is not to encode
an explicit decomposition of the scene. Rather, we accept that existing object category classifiers often encode not the category
per se but ancillary image characteristics; and that these ancillary characteristics can combine to represent visual classes unrelated
to the constituent categories’ semantic meanings.
The advantage of this descriptor is that it allows object-category queries to be made against image databases using efficient
classifiers (efficient at test time) such as linear support vector machines, and allows these queries to be for novel categories.
Even when the representation is reduced to 200 bytes per image, classification accuracy on object category recognition is
comparable with the state of the art (36% versus 42%), but at orders of magnitude lower computational cost.
Caltech-UCSD Birds 200 (CUB-200) is a challenging image dataset annotated with 200 bird species. It was created to enable the study of subordinate categorization, which is not possible with other popular datasets that focus on basic level categories (such as PASCAL VOC, Caltech-101, etc). The images were downloaded from the website Flickr and filtered by workers on Amazon Mechanical Turk. Each image is annotated with a bounding box, a rough bird segmentation, and a set of attribute labels.
Exploratory activities seem to be intrinsically rewarding for children and crucial for their cognitive development. Can a machine be endowed with such an intrinsic motivation system? This is the question we study in this paper, presenting a number of computational systems that try to capture this drive towards novel or curious situations. After discussing related research coming from developmental psychology, neuroscience, developmental robotics, and active learning, this paper presents the mechanism of Intelligent Adaptive Curiosity, an intrinsic motivation system which pushes a robot towards situations in which it maximizes its learning progress. This drive makes the robot focus on situations which are neither too predictable nor too unpredictable, thus permitting autonomous mental development. The complexity of the robot's activities autonomously increases and complex developmental sequences self-organize without being constructed in a supervised manner. Two experiments are presented illustrating the stage-like organization emerging with this mechanism. In one of them, a physical robot is placed on a baby play mat with objects that it can learn to manipulate. Experimental results show that the robot first spends time in situations which are easy to learn, then shifts its attention progressively to situations of increasing difficulty, avoiding situations in which nothing can be learned. Finally, these various results are discussed in relation to more complex forms of behavioral organization and data coming from developmental psychology
In the multi-armed bandit problem, a gambler must decide which arm of K non-identical slot machines to play in a sequence of trials so as to maximize his reward. This classical problem has received much attention because of the simple model it provides of the trade-off between exploration (trying out each arm to find the best one) and exploitation (playing the arm believed to give the best payoff). Past solutions for the bandit problem have almost always relied on assumptions about the statistics of the slot machines. In this work, we make no statistical assumptions whatsoever about the nature of the process generating the payoffs of the slot machines. We give a solution to the bandit problem in which an adversary, rather than a well-behaved stochastic process, has complete control over the payoffs. In a sequence of T plays, we prove that the expected per-round payoff of our algorithm approaches that of the best arm at the rate O(T Gamma1=2 ), and we give an improved rate of converg...
In this paper, we introduce Recipe1M+, a new large-scale, structured corpus of over one million cooking recipes and 13 million food images. As the largest publicly available collection of recipe data, Recipe1M+ affords the ability to train high-capacity models on aligned, multimodal data. Using these data, we train a neural network to learn a joint embedding of recipes and images that yields impressive results on an image-recipe retrieval task. Moreover, we demonstrate that regularization via the addition of a high-level classification objective both improves retrieval performance to rival that of humans and enables semantic vector arithmetic. We postulate that these embeddings will provide a basis for further exploration of the Recipe1M+ dataset and food and cooking in general. Code, data and models are publicly available.
<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>
1.http://im2recipe.csail.mit.edu.
Automated photo tagging has established itself as one of the most compelling applications of deep learning. While deep convolutional neural networks have repeatedly demonstrated top performance on standard datasets for classification, there are a number of often overlooked but important considerations when deploying this technology in a real-world scenario. In this paper, we present our efforts in developing a large-scale photo tagging system for Flickr photo search. We discuss topics including how to 1) select the tags that matter most to our users; 2) develop lightweight, high-performance models for tag prediction; and 3) leverage the power of large amounts of noisy data for training. Our results demonstrate that, for real-world datasets, training exclusively with this noisy data yields performance on par with the standard paradigm of first pre-training on clean data and then fine-tuning. In addition, we observe that the models trained with user-generated data can yield better fine-tuning results when a small amount of clean data is available. As such, we advocate for the approach of harnessing user-generated data in large-scale systems.
Domain similarity measures can be used to gauge adaptability and select suitable data for transfer learning, but existing approaches define ad hoc measures that are deemed suitable for respective tasks. Inspired by work on curriculum learning, we propose to \emph{learn} data selection measures using Bayesian Optimization and evaluate them across models, domains and tasks. Our learned measures outperform existing domain similarity measures significantly on three tasks: sentiment analysis, part-of-speech tagging, and parsing. We show the importance of complementing similarity with diversity, and that learned measures are -- to some degree -- transferable across models, domains, and even tasks.
We introduce a method for automatically selecting the path, or syllabus, that a neural network follows through a curriculum so as to maximise learning efficiency. A measure of the amount that the network learns from each data sample is provided as a reward signal to a nonstationary multi-armed bandit algorithm, which then determines a stochastic syllabus. We consider a range of signals derived from two distinct indicators of learning progress: rate of increase in prediction accuracy, and rate of increase in network complexity. Experimental results for LSTM networks on three curricula demonstrate that our approach can significantly accelerate learning, in some cases halving the time required to attain a satisfactory performance level.
We created the Yahoo Flickr Creative Commons 100 Million Dataseta (YFCC100M) in 2014 as part of the Yahoo Webscope program, which is a reference library of interesting and scientifically useful datasets. The YFCC100M is the largest public multimedia collection ever released, with a total of 100 million media objects, of which approximately 99.2 million are photos and 0.8 million are videos, all uploaded to Flickr between 2004 and 2014 and published under a CC commercial or noncommercial license. The dataset is distributed through Amazon Web Services as a 12.5GB compressed archive containing only metadata. However, as with many datasets, the YFCC100M is constantly evolving; over time, we have released and will continue to release various expansion packs containing data not yet in the collection; for instance, the actual photos and videos, as well as several visual and aural features extracted from the data, have already been uploaded to the cloud, ensuring the dataset remains accessible and intact for years to come. The YFCC100M dataset overcomes many of the issues affecting existing multimedia datasets in terms of modalities, metadata, licensing, and, principally, volume.
Current computational approaches to learning visual object categories require thousands of training images, are slow, cannot learn in an incremental manner and cannot incorporate prior information into the learning process. In addition, no algorithm presented in the literature has been tested on more than a handful of object categories. We present an method for learning object categories from just a few training images. It is quick and it uses prior information in a principled way. We test it on a dataset composed of images of objects belonging to 101 widely varied categories. Our proposed method is based on making use of prior information, assembled from (unrelated) object categories which were previously learnt. A generative probabilistic model is used, which represents the shape and appearance of a constellation of features belonging to the object. The parameters of the model are learnt incrementally in a Bayesian manner. Our incremental algorithm is compared experimentally to an earlier batch Bayesian algorithm, as well as to one based on maximum likelihood. The incremental and batch versions have comparable classification performance on small training sets, but incremental learning is significantly faster, making real-time learning feasible. Both Bayesian methods outperform maximum likelihood on small training sets.
In this paper, we present a large-scale object retrieval system. The
user supplies a query object by selecting a region of a query image,
and the system returns a ranked list of images that contain the same
object, retrieved from a large corpus. We demonstrate the scalability
and performance of our system on a dataset of over 1 million images
crawled from the photo-sharing site, Flickr [3], using Oxford landmarks
as queries. Building an image-feature vocabulary is a major time
and performance bottleneck, due to the size of our dataset. To address
this problem we compare different scalable methods for building a
vocabulary and introduce a novel quantization method based on randomized
trees which we show outperforms the current state-of-the-art on an
extensive ground-truth. Our experiments show that the quantization
has a major effect on retrieval quality. To further improve query
performance, we add an efficient spatial verification stage to re-rank
the results returned from our bag-of-words model and show that this
consistently improves search quality, though by less of a margin
when the visual vocabulary is large. We view this work as a promising
step towards much larger, "web-scale" image corpora.
We show how a standard tool from statistics | namely condence bounds | can be used to elegantly deal with situations which exhibit an exploitation-exploration trade-o. Our technique for designing and analyzing algorithms for such situations is general and can be applied when an algorithm has to make exploitation-versus-exploration decisions based on uncertain information provided by a random process. We apply our technique to two models with such an exploitation-exploration trade-o. For the adversarial bandit problem with shifting our new algorithm suers only ~ O (ST)1=2 regret with high probability over T trials with S shifts. Such a regret bound was previously known only in expectation. The second model we consider is associative reinforcement learning with linear value functions. For this model our technique improves the regret from ~ O T 3=4 to ~ O T 1=2 .
Until recently, statistical theory has been restricted to the design and analysis of sampling experiments in which the size and composition of the samples are completely determined before the experimentation begins. The reasons for this are partly historical, dating back to the time when the statistician was consulted, if at all, only after the experiment was over, and partly intrinsic in the mathematical difficulty of working with anything but a fixed number of independent random variables. A major advance now appears to be in the making with the creation of a theory of the sequential design of experiments, in which the size and composition of the samples are not fixed in advance but are functions of the observations themselves.
Human linguistic annotation is crucial for many natural language processing tasks but can be expensive and time-consuming. We ex- plore the use of Amazon's Mechanical Turk system, a significantly cheaper and faster method for collecting annotations from a broad base of paid non-expert contributors over the Web. We investigate five tasks: af- fect recognition, word similarity, recognizing textual entailment, event temporal ordering, and word sense disambiguation. For all five, we show high agreement between Mechani- cal Turk non-expert annotations and existing gold standard labels provided by expert label- ers. For the task of affect recognition, we also show that using non-expert labels for training machine learning algorithms can be as effec- tive as using gold standard annotations from experts. We propose a technique for bias correction that significantly improves annota- tion quality on two tasks. We conclude that many large labeling tasks can be effectively designed and carried out in this method at a fraction of the usual expense.
Learning what data to learn
- Y Fan
- F Tian
- T Qin
- J Bian
- T.-Y Liu
Fan, Y.; Tian, F.; Qin, T.; Bian, J.; and Liu, T.-Y. 2017.
Learning what data to learn. In International Conference on
Learning Representations (Workshop).
Training convolutional networks with noisy labels
- S Sukhbaatar
- J Bruna
- M Paluri
- L Bourdev
Sukhbaatar, S.; Bruna, J.; Paluri, M.; Bourdev, L.; and Fergus,
R. 2014. Training convolutional networks with noisy labels.
In International Conference on Learning Representations
(Workshop).