Training large neural networks is possible by training a smaller hypernetwork that predicts parameters for the large ones. A recently released Graph HyperNetwork (GHN) trained this way on one million smaller ImageNet architectures is able to predict parameters for large unseen networks such as ResNet-50. While networks with predicted parameters lose performance on the source task, the predicted parameters have been found useful for fine-tuning on other tasks. We study if fine-tuning based on the same GHN is still useful on novel strong architectures that were published after the GHN had been trained. We found that for recent architectures such as ConvNeXt, GHN initialization becomes less useful than for ResNet-50. One potential reason is the increased distribution shift of novel architectures from those used to train the GHN. We also found that the predicted parameters lack the diversity necessary to successfully fine-tune parameters with gradient descent. We alleviate this limitation by applying simple post-processing techniques to predicted parameters before fine-tuning them on a target task and improve fine-tuning of ResNet-50 and ConvNeXt.
相關內容
The impressive performances of deep learning architectures is associated to massive increase of models complexity. Millions of parameters need be tuned, with training and inference time scaling accordingly. But is massive fine-tuning necessary? In this paper, focusing on image classification, we consider a simple transfer learning approach exploiting pretrained convolutional features as input for a fast kernel method. We refer to this approach as top-tuning, since only the kernel classifier is trained. By performing more than 2500 training processes we show that this top-tuning approach provides comparable accuracy w.r.t. fine-tuning, with a training time that is between one and two orders of magnitude smaller. These results suggest that top-tuning provides a useful alternative to fine-tuning in small/medium datasets, especially when training efficiency is crucial.
Neural Architecture Search (NAS) has significantly improved productivity in the design and deployment of neural networks (NN). As NAS typically evaluates multiple models by training them partially or completely, the improved productivity comes at the cost of significant carbon footprint. To alleviate this expensive training routine, zero-shot/cost proxies analyze an NN at initialization to generate a score, which correlates highly with its true accuracy. Zero-cost proxies are currently designed by experts conducting multiple cycles of empirical testing on possible algorithms, data-sets, and neural architecture design spaces. This lowers productivity and is an unsustainable approach towards zero-cost proxy design as deep learning use-cases diversify in nature. Additionally, existing zero-cost proxies fail to generalize across neural architecture design spaces. In this paper, we propose a genetic programming framework to automate the discovery of zero-cost proxies for neural architecture scoring. Our methodology efficiently discovers an interpretable and generalizable zero-cost proxy that gives state of the art score-accuracy correlation on all data-sets and search spaces of NASBench-201 and Network Design Spaces (NDS). We believe that this research indicates a promising direction towards automatically discovering zero-cost proxies that can work across network architecture design spaces, data-sets, and tasks.
Neural architecture search (NAS) has become a common approach to developing and discovering new neural architectures for different target platforms and purposes. However, scanning the search space is comprised of long training processes of many candidate architectures, which is costly in terms of computational resources and time. Regression algorithms are a common tool to predicting a candidate architecture's accuracy, which can dramatically accelerate the search procedure. We aim at proposing a new baseline that will support the development of regression algorithms that can predict an architecture's accuracy just from its scheme, or by only training it for a minimal number of epochs. Therefore, we introduce the NAAP-440 dataset of 440 neural architectures, which were trained on CIFAR10 using a fixed recipe. Our experiments indicate that by using off-the-shelf regression algorithms and running up to 10% of the training process, not only is it possible to predict an architecture's accuracy rather precisely, but that the values predicted for the architectures also maintain their accuracy order with a minimal number of monotonicity violations. This approach may serve as a powerful tool for accelerating NAS-based studies and thus dramatically increase their efficiency. The dataset and code used in the study have been made public.
Network architecture search (NAS) has become a common approach to developing and discovering new neural architectures for different target platforms and purposes. However, scanning the search space is comprised of long training processes of many candidate architectures, which is costly in terms of computational resources and time. Regression algorithms are a common tool to predicting a candidate architecture's accuracy, which can dramatically accelerate the search procedure. We aim at proposing a new baseline that will support the development of regression algorithms that can predict an architecture's accuracy just from its scheme, or by only training it for a minimal number of epochs. Therefore, we introduce the NAAP-440 dataset of 440 neural architectures, which were trained on CIFAR10 using a fixed recipe. Our experiments indicate that by using off-the-shelf regression algorithms and running up to 10% of the training process, not only is it possible to predict an architecture's accuracy rather precisely, but that the values predicted for the architectures also maintain their accuracy order with a minimal number of monotonicity violations. This approach may serve as a powerful tool for accelerating NAS-based studies and thus dramatically increase their efficiency. The dataset and code used in the study have been made public.
In this paper, we propose a patch-based architecture for multi-label classification problems where only a single positive label is observed in images of the dataset. Our contributions are twofold. First, we introduce a light patch architecture based on the attention mechanism. Next, leveraging on patch embedding self-similarities, we provide a novel strategy for estimating negative examples and deal with positive and unlabeled learning problems. Experiments demonstrate that our architecture can be trained from scratch, whereas pre-training on similar databases is required for related methods from the literature.
We propose a lightweight and accurate method for detecting anomalies in videos. Existing methods used multiple-instance learning (MIL) to determine the normal/abnormal status of each segment of the video. Recent successful researches argue that it is important to learn the temporal relationships among segments to achieve high accuracy, instead of focusing on only a single segment. Therefore we analyzed the existing methods that have been successful in recent years, and found that while it is indeed important to learn all segments together, the temporal orders among them are irrelevant to achieving high accuracy. Based on this finding, we do not use the MIL framework, but instead propose a lightweight model with a self-attention mechanism to automatically extract features that are important for determining normal/abnormal from all input segments. As a result, our neural network model has 1.3\% of the number of parameters of the existing method. We evaluated the frame-level detection accuracy of our method on three benchmark datasets (UCF-Crime, ShanghaiTech, and XD-Violence) and demonstrate that our method can achieve the comparable or better accuracy than state-of-the-art methods.
The dominating NLP paradigm of training a strong neural predictor to perform one task on a specific dataset has led to state-of-the-art performance in a variety of applications (eg. sentiment classification, span-prediction based question answering or machine translation). However, it builds upon the assumption that the data distribution is stationary, ie. that the data is sampled from a fixed distribution both at training and test time. This way of training is inconsistent with how we as humans are able to learn from and operate within a constantly changing stream of information. Moreover, it is ill-adapted to real-world use cases where the data distribution is expected to shift over the course of a model's lifetime. The first goal of this thesis is to characterize the different forms this shift can take in the context of natural language processing, and propose benchmarks and evaluation metrics to measure its effect on current deep learning architectures. We then proceed to take steps to mitigate the effect of distributional shift on NLP models. To this end, we develop methods based on parametric reformulations of the distributionally robust optimization framework. Empirically, we demonstrate that these approaches yield more robust models as demonstrated on a selection of realistic problems. In the third and final part of this thesis, we explore ways of efficiently adapting existing models to new domains or tasks. Our contribution to this topic takes inspiration from information geometry to derive a new gradient update rule which alleviate catastrophic forgetting issues during adaptation.
The time and effort involved in hand-designing deep neural networks is immense. This has prompted the development of Neural Architecture Search (NAS) techniques to automate this design. However, NAS algorithms tend to be slow and expensive; they need to train vast numbers of candidate networks to inform the search process. This could be alleviated if we could partially predict a network's trained accuracy from its initial state. In this work, we examine the overlap of activations between datapoints in untrained networks and motivate how this can give a measure which is usefully indicative of a network's trained performance. We incorporate this measure into a simple algorithm that allows us to search for powerful networks without any training in a matter of seconds on a single GPU, and verify its effectiveness on NAS-Bench-101, NAS-Bench-201, NATS-Bench, and Network Design Spaces. Our approach can be readily combined with more expensive search methods; we examine a simple adaptation of regularised evolutionary search. Code for reproducing our experiments is available at //github.com/BayesWatch/nas-without-training.
Deep Convolutional Neural Networks (CNNs) are a special type of Neural Networks, which have shown state-of-the-art results on various competitive benchmarks. The powerful learning ability of deep CNN is largely achieved with the use of multiple non-linear feature extraction stages that can automatically learn hierarchical representation from the data. Availability of a large amount of data and improvements in the hardware processing units have accelerated the research in CNNs and recently very interesting deep CNN architectures are reported. The recent race in deep CNN architectures for achieving high performance on the challenging benchmarks has shown that the innovative architectural ideas, as well as parameter optimization, can improve the CNN performance on various vision-related tasks. In this regard, different ideas in the CNN design have been explored such as use of different activation and loss functions, parameter optimization, regularization, and restructuring of processing units. However, the major improvement in representational capacity is achieved by the restructuring of the processing units. Especially, the idea of using a block as a structural unit instead of a layer is gaining substantial appreciation. This survey thus focuses on the intrinsic taxonomy present in the recently reported CNN architectures and consequently, classifies the recent innovations in CNN architectures into seven different categories. These seven categories are based on spatial exploitation, depth, multi-path, width, feature map exploitation, channel boosting and attention. Additionally, it covers the elementary understanding of the CNN components and sheds light on the current challenges and applications of CNNs.
Deep Learning has enabled remarkable progress over the last years on a variety of tasks, such as image recognition, speech recognition, and machine translation. One crucial aspect for this progress are novel neural architectures. Currently employed architectures have mostly been developed manually by human experts, which is a time-consuming and error-prone process. Because of this, there is growing interest in automated neural architecture search methods. We provide an overview of existing work in this field of research and categorize them according to three dimensions: search space, search strategy, and performance estimation strategy.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191