Posted Content•
Neural Architecture Search with Reinforcement Learning
TL;DR: This paper uses a recurrent network to generate the model descriptions of neural networks and trains this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set.
Abstract: Neural networks are powerful and flexible models that work well for many difficult learning tasks in image, speech and natural language understanding. Despite their success, neural networks are still hard to design. In this paper, we use a recurrent network to generate the model descriptions of neural networks and train this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set. On the CIFAR-10 dataset, our method, starting from scratch, can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy. Our CIFAR-10 model achieves a test error rate of 3.65, which is 0.09 percent better and 1.05x faster than the previous state-of-the-art model that used a similar architectural scheme. On the Penn Treebank dataset, our model can compose a novel recurrent cell that outperforms the widely-used LSTM cell, and other state-of-the-art baselines. Our cell achieves a test set perplexity of 62.4 on the Penn Treebank, which is 3.6 perplexity better than the previous state-of-the-art model. The cell can also be transferred to the character language modeling task on PTB and achieves a state-of-the-art perplexity of 1.214.
Citations
More filters
26 Aug 2021
TL;DR: This paper compares two AutoML feature engineering methods in the context of the National Assessment of Educational Progress (NAEP) data mining competition and develops a novel feature selection method that addresses problems applying AutoML features in this context, where there were many heterogeneous features and relatively few students.
Abstract: Automatic machine learning (AutoML) methods automate the time-consuming, feature-engineering process so that researchers produce accurate student models more quickly and easily. In this paper, we compare two AutoML feature engineering methods in the context of the National Assessment of Educational Progress (NAEP) data mining competition. The methods we compare, Featuretools and TSFRESH (Time Series FeatuRe Extraction on basis of Scalable Hypothesis tests), have rarely been applied in the context of student interaction log data. Thus, we address research questions regarding the accuracy of models built with AutoML features, how AutoML feature types compare to each other and to expert-engineered features, and how interpretable the features are. Additionally, we developed a novel feature selection method that addresses problems applying AutoML feature engineering in this context, where there were many heterogeneous features (over 4,000) and relatively few students. Our entry to the NAEP competition placed 3rd overall on the final held-out dataset and 1st on the public leaderboard, with a final Cohen's kappa = .212 and area under the receiver operating characteristic curve (AUC) = .665 when predicting whether students would manage their time effectively on a math assessment. We found that TSFRESH features were significantly more effective than either Featuretools features or expert-engineered features in this context; however, they were also among the most difficult features to interpret based on a survey of six experts' judgments. Finally, we discuss the tradeoffs between effort and interpretability that arise in AutoML-based student modeling.
1 citations
TL;DR: A novel Learned Depthwise Separable Convolution (LdsConv) operation that is smart but has a strong capacity for learning is proposed that integrates the pruning technique into the design of convolutional filters.
Abstract: Standard convolutional filters usually capture unnecessary overlap of features resulting in a waste of computational cost. In this paper, we aim to solve this problem by proposing a novel Learned Depthwise Separable Convolution (LdsConv) operation that is smart but has a strong capacity for learning. It integrates the pruning technique into the design of convolutional filters, formulated as a generic convolutional unit that can be used as a direct replacement of convolutions without any adjustments of the architecture. To show the effectiveness of the proposed method, experiments are carried out using the state-of-the-art convolutional neural networks (CNNs), including ResNet, DenseNet, SE-ResNet and MobileNet, respectively. The results show that by simply replacing the original convolution with LdsConv in these CNNs, it can achieve a significantly improved accuracy while reducing computational cost. For the case of ResNet50, the FLOPs can be reduced by 40.9%, meanwhile the accuracy on the associated ImageNet increases.
1 citations
01 Jan 2019
TL;DR: In this article, the authors proposed a framework to unify varied shadows into a limited number of patterns, with the convolutional-neural network (CNNs)-powered shape classification for efficient shadow-based building height estimation and integrating semantic analysis into a framework of various state-of-the-art CNNs to support human-level understanding of map content.
Abstract: Big data that contain geo-referenced attributes have
significantly reformed the way that I process and analyze
geospatial data. Compared with the expected benefits received in
the data-rich environment, more data have not always contributed to
more accurate analysis. “Big but valueless” has becoming a critical
concern to the community of GIScience and data-driven geography. As
a highly-utilized function of GeoAI technique, deep learning models
designed for processing geospatial data integrate powerful
computing hardware and deep neural networks into various dimensions
of geography to effectively discover the representation of data.
However, limitations of these deep learning models have also been
reported when People may have to spend much time on preparing
training data for implementing a deep learning model. The objective
of this dissertation research is to promote state-of-the-art deep
learning models in discovering the representation, value and hidden
knowledge of GIS and remote sensing data, through three research
approaches. The first methodological framework aims to unify varied
shadow into limited number of patterns, with the convolutional
neural network (CNNs)-powered shape classification, multifarious
shadow shapes with a limited number of representative shadow
patterns for efficient shadow-based building height estimation. The
second research focus integrates semantic analysis into a framework
of various state-of-the-art CNNs to support human-level
understanding of map content. The final research approach of this
dissertation focuses on normalizing geospatial domain knowledge to
promote the transferability of a CNN’s model to land-use/land-cover
classification. This research reports a method designed to discover
detailed land-use/land-cover types that might be challenging for a
state-of-the-art CNN’s model that previously performed well on
land-cover classification only.
1 citations
29 Oct 2021
TL;DR: In this paper, a non-local network routing (NNR) approach is proposed for perceptual image super-resolution, which aims to globally optimize both procedures of learning an optimal perceptual loss and routing a spatial-adaptive network architecture in a unified reinforcement learning framework.
Abstract: In this paper, we propose a non-local network routing (NNR) approach for perceptual image super-resolution. Unlike conventional methods which generate visually-faked textures due to exiting hand-designed losses, our approach aims to globally optimize both procedures of learning an optimal perceptual loss and routing a spatial-adaptive network architecture in a unified reinforcement learning framework. To this end, we introduce a reward function to teach our objective to pay more attention on the visual quality of the super-resolved image. Moreover, we carefully design an offset operation inside the neural architecture search space, which typically deforms the receptive field on boundary refinement in a non-local manner. Experimentally, our proposed method surpasses the perceptual performance over state-of-the-art methods on several widely-evaluated benchmark datasets.
1 citations
TL;DR: A simpler SDC scheme with a higher classification accuracy, named NAS-SDC, is developed based on the neural architecture search (NAS) technique, where a max-pooling cell based on NASNet is introduced to reduce the search space and the number of network parameters, thus simplifying the candidate operators for the search.
Abstract: Surface defect classification (SDC) is the visual inspection of the surface of an object to identify appearance defects. Efficient and accurate SDC is i mportant for improving the quality of industrial products. A manually designed convolutional neural network (CNN) is traditionally used for SDC. In this study, a simpler SDC scheme with a higher classification accuracy, named NAS-SDC, is developed based on the neural architecture search (NAS) technique. A max-pooling cell based on NASNet is introduced to reduce the search space and the number of network parameters, thus simplifying the candidate operators for the search. Two network architectures are proposed to stack the search candidates or the best cells. The proposed method can be used to automatically design an efficient CNN model for SDC on a specific dataset. Experimental results show that the proposed method can find the best cells in ~11 h using a single graphics processing unit (GPU) and achieves higher classification accuracies (99.98%, 99.8% and 99.26%) than state-of-the-art methods on the Northeastern University (NEU-CLS), DAGM, and bridge defect datasets. The number of network parameters used in the proposed method is only 0.35 M, and the average test time per sample is approximately 61 ms, thus achieving a balance between performance and speed.
1 citations
References
More filters
27 Jun 2016
TL;DR: In this article, the authors proposed a residual learning framework to ease the training of networks that are substantially deeper than those used previously, which won the 1st place on the ILSVRC 2015 classification task.
Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.
123,388 citations
Proceedings Article•
01 Jan 2015TL;DR: This work introduces Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments, and provides a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework.
Abstract: We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.
111,197 citations
Proceedings Article•
04 Sep 2014TL;DR: This work investigates the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting using an architecture with very small convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers.
Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.
55,235 citations
"Neural Architecture Search with Rei..." refers methods in this paper
...Along with this success is a paradigm shift from feature designing to architecture designing, i.e., from SIFT (Lowe, 1999), and HOG (Dalal & Triggs, 2005), to AlexNet (Krizhevsky et al., 2012), VGGNet (Simonyan & Zisserman, 2014), GoogleNet (Szegedy et al., 2015), and ResNet (He et al., 2016a)....
[...]
01 Jan 1998
TL;DR: In this article, a graph transformer network (GTN) is proposed for handwritten character recognition, which can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters.
Abstract: Multilayer neural networks trained with the back-propagation algorithm constitute the best example of a successful gradient based learning technique. Given an appropriate network architecture, gradient-based learning algorithms can be used to synthesize a complex decision surface that can classify high-dimensional patterns, such as handwritten characters, with minimal preprocessing. This paper reviews various methods applied to handwritten character recognition and compares them on a standard handwritten digit recognition task. Convolutional neural networks, which are specifically designed to deal with the variability of 2D shapes, are shown to outperform all other techniques. Real-life document recognition systems are composed of multiple modules including field extraction, segmentation recognition, and language modeling. A new learning paradigm, called graph transformer networks (GTN), allows such multimodule systems to be trained globally using gradient-based methods so as to minimize an overall performance measure. Two systems for online handwriting recognition are described. Experiments demonstrate the advantage of global training, and the flexibility of graph transformer networks. A graph transformer network for reading a bank cheque is also described. It uses convolutional neural network character recognizers combined with global training techniques to provide record accuracy on business and personal cheques. It is deployed commercially and reads several million cheques per day.
42,067 citations
20 Jun 2005
TL;DR: It is shown experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection, and the influence of each stage of the computation on performance is studied.
Abstract: We study the question of feature sets for robust visual object recognition; adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of histograms of oriented gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.
31,952 citations
"Neural Architecture Search with Rei..." refers methods in this paper
...Along with this success is a paradigm shift from feature designing to architecture designing, i.e., from SIFT (Lowe, 1999), and HOG (Dalal & Triggs, 2005), to AlexNet (Krizhevsky et al., 2012), VGGNet (Simonyan & Zisserman, 2014), GoogleNet (Szegedy et al., 2015), and ResNet (He et al., 2016a)....
[...]