The conventional approach to image recognition has been based on raster graphics, which can suffer from aliasing and information loss when scaled up or down. In this paper, we propose a novel approach that leverages the benefits of vector graphics for object localization and classification. Our method, called YOLaT (You Only Look at Text), takes the textual document of vector graphics as input, rather than rendering it into pixels. YOLaT builds multi-graphs to model the structural and spatial information in vector graphics and utilizes a dual-stream graph neural network (GNN) to detect objects from the graph. However, for real-world vector graphics, YOLaT only models in flat GNN with vertexes as nodes ignore higher-level information of vector data. Therefore, we propose YOLaT++ to learn Multi-level Abstraction Feature Learning from a new perspective: Primitive Shapes to Curves and Points. On the other hand, given few public datasets focus on vector graphics, data-driven learning cannot exert its full power on this format. We provide a large-scale and challenging dataset for Chart-based Vector Graphics Detection and Chart Understanding, termed VG-DCU, with vector graphics, raster graphics, annotations, and raw data drawn for creating these vector charts. Experiments show that the YOLaT series outperforms both vector graphics and raster graphics-based object detection methods on both subsets of VG-DCU in terms of both accuracy and efficiency, showcasing the potential of vector graphics for image recognition tasks.

We present a method to formulate algorithm discovery as program search, and apply it to discover optimization algorithms for deep neural network training. We leverage efficient search techniques to explore an infinite and sparse program space. To bridge the large generalization gap between proxy and target tasks, we also introduce program selection and simplification strategies. Our method discovers a simple and effective optimization algorithm, $\textbf{Lion}$ ($\textit{Evo$\textbf{L}$ved S$\textbf{i}$gn M$\textbf{o}$me$\textbf{n}$tum}$). It is more memory-efficient than Adam as it only keeps track of the momentum. Different from adaptive optimizers, its update has the same magnitude for each parameter calculated through the sign operation. We compare Lion with widely used optimizers, such as Adam and Adafactor, for training a variety of models on different tasks. On image classification, Lion boosts the accuracy of ViT by up to 2\% on ImageNet and saves up to 5x the pre-training compute on JFT. On vision-language contrastive learning, we achieve 88.3\% $\textit{zero-shot}$ and 91.1\% $\textit{fine-tuning}$ accuracy on ImageNet, surpassing the previous best results by 2\% and 0.1\%, respectively. On diffusion models, Lion outperforms Adam by achieving a better FID score and reducing the training compute by up to 2.3x. For autoregressive, masked language modeling, and fine-tuning, Lion exhibits a similar or better performance compared to Adam. Our analysis of Lion reveals that its performance gain grows with the training batch size. It also requires a smaller learning rate than Adam due to the larger norm of the update produced by the sign function. Additionally, we examine the limitations of Lion and identify scenarios where its improvements are small or not statistically significant.

The mixture proportions of pretraining data domains (e.g., Wikipedia, books, web text) greatly affect language model (LM) performance. In this paper, we propose Domain Reweighting with Minimax Optimization (DoReMi), which first trains a small proxy model using group distributionally robust optimization (Group DRO) over domains to produce domain weights (mixture proportions) without knowledge of downstream tasks. We then resample a dataset with these domain weights and train a larger, full-sized model. In our experiments, we use DoReMi on a 280M-parameter proxy model to set the domain weights for training an 8B-parameter model (30x larger) more efficiently. On The Pile, DoReMi improves perplexity across all domains, even when it downweights a domain. DoReMi improves average few-shot downstream accuracy by 6.5% points over a baseline model trained using The Pile's default domain weights and reaches the baseline accuracy with 2.6x fewer training steps. On the GLaM dataset, DoReMi, which has no knowledge of downstream tasks, even matches the performance of using domain weights tuned on downstream tasks.

A variety of real-world applications rely on far future information to make decisions, thus calling for efficient and accurate long sequence multivariate time series forecasting. While recent attention-based forecasting models show strong abilities in capturing long-term dependencies, they still suffer from two key limitations. First, canonical self attention has a quadratic complexity w.r.t. the input time series length, thus falling short in efficiency. Second, different variables’ time series often have distinct temporal dynamics, which existing studies fail to capture, as they use the same model parameter space, e.g., projection matrices, for all variables’ time series, thus falling short in accuracy. To ensure high efficiency and accuracy, we propose Triformer, a triangular, variable-specific attention. (i) Linear complexity: we introduce a novel patch attention with linear complexity. When stacking multiple layers of the patch attentions, a triangular structure is proposed such that the layer sizes shrink exponentially, thus maintaining linear complexity. (ii) Variable-specific parameters: we propose a light-weight method to enable distinct sets of model parameters for different variables’ time series to enhance accuracy without compromising efficiency and memory usage. Strong empirical evidence on four datasets from multiple domains justifies our design choices, and it demonstrates that Triformer outperforms state-of-the-art methods w.r.t. both accuracy and efficiency. Source code is publicly available at https://github.com/razvanc92/triformer.

Neural architecture search (NAS) has attracted a lot of attention and has been illustrated to bring tangible benefits in a large number of applications in the past few years. Architecture topology and architecture size have been regarded as two of the most important aspects for the performance of deep learning models and the community has spawned lots of searching algorithms for both of those aspects of the neural architectures. However, the performance gain from these searching algorithms is achieved under different search spaces and training setups. This makes the overall performance of the algorithms incomparable and the improvement from a sub-module of the searching model unclear. In this paper, we propose NATS-Bench, a unified benchmark on searching for both topology and size, for (almost) any up-to-date NAS algorithm. NATS-Bench includes the search space of 15,625 neural cell candidates for architecture topology and 32,768 for architecture size on three datasets. We analyze the validity of our benchmark in terms of various criteria and performance comparison of all candidates in the search space. We also show the versatility of NATS-Bench by benchmarking 13 recent state-of-the-art NAS algorithms on it. All logs and diagnostic information trained using the same setup for each candidate are provided. This facilitates a much larger community of researchers to focus on developing better NAS algorithms in a more comparable and computationally effective environment. All codes are publicly available at: https://xuanyidong.com/assets/projects/NATS-Bench.

In this paper, we consider a different data format for images: vector graphics. In contrast to raster graphics which are widely used in image recognition, vector graphics can be scaled up or down into any resolution without aliasing or information loss, due to the analytic representation of the primitives in the document. Furthermore, vector graphics are able to give extra structural information on how low-level elements group together to form high level shapes or structures. These merits of graphic vectors have not been fully leveraged in existing methods. To explore this data format, we target on the fundamental recognition tasks: object localization and classification. We propose an efficient CNN-free pipeline that does not render the graphic into pixels (i.e. rasterization), and takes textual document of the vector graphics as input, called YOLaT (You Only Look at Text). YOLaT builds multi-graphs to model the structural and spatial information in vector graphics, and a dual-stream graph neural network is proposed to detect objects from the graph. Our experiments show that by directly operating on vector graphics, YOLaT outperforms raster-graphic based object detection baselines in terms of both average precision and efficiency. Code is available at https://github.com/microsoft/YOLaT-VectorGraphicsRecognition.

Existing visual navigation approaches leverage classification neural networks to extract global features from visual data for navigation. However, these networks are not originally designed for navigation tasks. Thus, the neural architectures might not be suitable to capture scene con-tents. Fortunately, neural architecture search (NAS) brings a hope to solve this problem. In this paper, we propose an Auto-Navigator to customize a specialized network for visual navigation. However, as navigation tasks mainly rely on reinforcement learning (RL) rewards in training, such weak supervision is insufficiently indicative for NAS to optimize visual perception network. Thus, we introduce imitation learning (IL) with optimal paths to optimize navigation policies while selecting an optimal architecture. As Auto-Navigator can obtain a direct supervision in every step, such guidance greatly facilitates architecture search. In particular, we initialize our Auto-Navigator with a learnable distribution over the search space of visual perception architecture, and then optimize the distribution with IL supervision. Afterwards, we employ an RL reward function to fine-tune our Auto-Navigator to improve the generalization ability of our model. Extensive experiments demonstrate that our Auto-Navigator outperforms baseline methods on Gibson and Matterport3D without significantly increasing network parameters.

We present supervision by registration and triangulation (SRT), an unsupervised approach that utilizes unlabeled multi-view video to improve the accuracy and precision of landmark detectors. Being able to utilize unlabeled data enables our detectors to learn from massive amounts of unlabeled data freely available and not be limited by the quality and quantity of manual human annotations. To utilize unlabeled data, there are two key observations: (I) The detections of the same landmark in adjacent frames should be coherent with registration, i.e., optical flow. (II) The detections of the same landmark in multiple synchronized and geometrically calibrated views should correspond to a single 3D point, i.e., multi-view consistency. Registration and multi-view consistency are sources of supervision that do not require manual labeling, thus it can be leveraged to augment existing training data during detector training. End-to-end training is made possible by differentiable registration and 3D triangulation modules. Experiments with 11 datasets and a newly proposed metric to measure precision demonstrate accuracy and precision improvements in landmark detection on both images and video.

Neural networks are sensitive to hyper-parameter and architecture choices. Automated Machine Learning (AutoML) is a promising paradigm for automating these choices. Current ML software libraries, however, are quite limited in handling the dynamic interactions among the components of AutoML. For example, efficient NAS algorithms, such as ENAS and DARTS, typically require an implementation coupling between the search space and search algorithm, the two key components in AutoML. Furthermore, implementing a complex search flow, such as searching architectures within a loop of searching hardware configurations, is difficult. To summarize, changing the search space, search algorithm, or search flow in current ML libraries usually requires a significant change in the program logic. In this paper, we introduce a new way of programming AutoML based on symbolic programming. Under this paradigm, ML programs are mutable, thus can be manipulated easily by another program. As a result, AutoML can be reformulated as an automated process of symbolic manipulation. With this formulation, we decouple the triangle of the search algorithm, the search space and the child program. This decoupling makes it easy to change the search space and search algorithm (without and with weight sharing), as well as to add search capabilities to existing code and implement complex search flows. We then introduce PyGlove, a new Python library that implements this paradigm. Through case studies on ImageNet and NAS-Bench-101, we show that with PyGlove users can easily convert a static program into a search space, quickly iterate on the search spaces and search algorithms, and craft complex search flows to achieve better results.

Co-training is a well-known semi-supervised learning approach which trains classifiers on two or more different views and exchanges pseudo labels of unlabeled instances in an iterative way. During the co-training process, pseudo labels of unlabeled instances are very likely to be false especially in the initial training, while the standard co-training algorithm adopts a 'draw without replacement' strategy and does not remove these wrongly labeled instances from training stages. Besides, most of the traditional co-training approaches are implemented for two-view cases, and their extensions in multi-view scenarios are not intuitive. These issues not only degenerate their performance as well as available application range but also hamper their fundamental theory. Moreover, there is no optimization model to explain the objective a co-training process manages to optimize. To address these issues, in this study we design a unified self-paced multi-view co-training (SPamCo) framework which draws unlabeled instances with replacement. Two specified co-regularization terms are formulated to develop different strategies for selecting pseudo-labeled instances during training. Both forms share the same optimization strategy which is consistent with the iteration process in co-training and can be naturally extended to multi-view scenarios. A distributed optimization strategy is also introduced to train the classifier of each view in parallel to further improve the efficiency of the algorithm. Furthermore, the SPamCo algorithm is proved to be PAC learnable, supporting its theoretical soundness. Experiments conducted on synthetic, text categorization, person re-identification, image recognition and object detection data sets substantiate the superiority of the proposed method.

Deeper and wider convolutional neural networks (CNNs) achieve superior performance but bring expensive computation cost. Accelerating such overparameterized neural network has received increased attention. A typical pruning algorithm is a three-stage pipeline, i.e., training, pruning, and retraining. Prevailing approaches fix the pruned filters to zero during retraining and, thus, significantly reduce the optimization space. Besides, they directly prune a large number of filters at first, which would cause unrecoverable information loss. To solve these problems, we propose an asymptotic soft filter pruning (ASFP) method to accelerate the inference procedure of the deep neural networks. First, we update the pruned filters during the retraining stage. As a result, the optimization space of the pruned model would not be reduced but be the same as that of the original model. In this way, the model has enough capacity to learn from the training data. Second, we prune the network asymptotically. We prune few filters at first and asymptotically prune more filters during the training procedure. With asymptotic pruning, the information of the training set would be gradually concentrated in the remaining filters, so the subsequent training and pruning process would be stable. The experiments show the effectiveness of our ASFP on image classification benchmarks. Notably, on ILSVRC-2012, our ASFP reduces more than 40% FLOPs on ResNet-50 with only 0.14% top-5 accuracy degradation, which is higher than the soft filter pruning by 8%.

Most person re-identification (re-ID) approaches are based on supervised learning, which requires intensive manual annotation for training data. However, it is not only resourceintensive to acquire identity annotation but also impractical to label the large-scale real-world data. To relieve this problem, we propose a bottom-up clustering (BUC) approach to jointly optimize a convolutional neural network (CNN) and the relationship among the individual samples. Our algorithm considers two fundamental facts in the re-ID task, i.e., diversity across different identities and similarity within the same identity. Specifically, our algorithm starts with regarding individual sample as a different identity, which maximizes the diversity over each identity. Then it gradually groups similar samples into one identity, which increases the similarity within each identity. We utilizes a diversity regularization term in the bottom-up clustering procedure to balance the data volume of each cluster. Finally, the model achieves an effective trade-off between the diversity and similarity. We conduct extensive experiments on the large-scale image and video re-ID datasets, including Market-1501, DukeMTMCreID, MARS and DukeMTMC-VideoReID. The experimental results demonstrate that our algorithm is not only superior to state-of-the-art unsupervised re-ID approaches, but also performs favorably than competing transfer learning and semi-supervised learning methods.

Conventional neural architecture search (NAS) approaches are usually based on reinforcement learning or evolutionary strategy, which take more than 1000 GPU hours to find a good model on CIFAR-10. We propose an efficient NAS approach, which learns the searching approach by gradient descent. Our approach represents the search space as a directed acyclic graph (DAG). This DAG contains thousands of sub-graphs, each of which indicates a kind of neural architecture. To avoid traversing all the possibilities of the sub-graphs, we develop a differentiable sampler over the DAG. This sampler is learnable and optimized by the validation loss after training the sampled architecture. In this way, our approach can be trained in an end-to-end fashion by gradient descent, named Gradient-based search using Differentiable Architecture Sampler (GDAS). In experiments, we can finish one searching procedure in four GPU hours on CIFAR-10, and the discovered model obtains a test error of 2.82% with only 2.5M parameters, which is on par with the state-of-the-art.

Facial landmark detection aims to localize the anatomically defined points of human faces. In this paper, we study facial landmark detection from partially labeled facial images. A typical approach is to (1) train a detector on the labeled images; (2) generate new training samples using this detector's prediction as pseudo labels of unlabeled images; (3) retrain the detector on the labeled samples and partial pseudo labeled samples. In this way, the detector can learn from both labeled and unlabeled data and become robust. In this paper, we propose an interaction mechanism between a teacher and two students to generate more reliable pseudo labels for unlabeled data, which are beneficial to semi-supervised facial landmark detection. Specifically, the two students are instantiated as dual detectors. The teacher learns to judge the quality of the pseudo labels generated by the students and filter out unqualified samples before the retraining stage. In this way, the student detectors get feedback from their teacher and are retrained by premium data generated by itself. Since the two students are trained by different samples, a combination of their predictions will be more robust as the final prediction compared to either prediction. Extensive experiments on 300-W and AFLW benchmarks show that the interactions between teacher and students contribute to better utilization of the unlabeled data and achieves state-of-the-art performance.

Neural architecture search (NAS) aims to automate the search procedure of architecture instead of manual design. Even if recent NAS approaches finish the search within days, lengthy training is still required for a specific architecture candidate to get the parameters for its accurate evaluation. Recently one-shot NAS methods are proposed to largely squeeze the tedious training process by sharing parameters across candidates. In this way, the parameters for each candidate can be directly extracted from the shared parameters instead of training them from scratch. However, they have no sense of which candidate will perform better until evaluation so that the candidates to evaluate are randomly sampled and the top-1 candidate is considered the best. In this paper, we propose a Self-Evaluated Template Network (SETN) to improve the quality of the architecture candidates for evaluation so that it is more likely to cover competitive candidates. SETN consists of two components: (1) an evaluator, which learns to indicate the probability of each individual architecture being likely to have a lower validation loss. The candidates for evaluation can thus be selectively sampled according to this evaluator. (2) a template network, which shares parameters among all candidates to amortize the training cost of generated candidates. In experiments, the architecture found by SETN achieves the state-of-the-art performance on CIFAR and ImageNet benchmarks within comparable computation costs.

Prevailing deep convolutional neural networks (CNNs) for person re-IDentification (reID) are usually built upon ResNet or VGG backbones, which were originally designed for classification. Because reID is different from classification, the architecture should be modified accordingly. We propose to automatically search for a CNN architecture that is specifically suitable for the reID task. There are three aspects to be tackled. First, body structural information plays an important role in reID but it is not encoded in backbones. Second, Neural Architecture Search (NAS) automates the process of architecture design without human effort, but no existing NAS methods incorporate the structure information of input images. Third, reID is essentially a retrieval task but current NAS algorithms are merely designed for classification. To solve these problems, we propose a retrieval-based search algorithm over a specifically designed reID search space, named Auto-ReID. Our Auto-ReID enables the automated approach to find an efficient and effective CNN architecture for reID. Extensive experiments demonstrate that the searched architecture achieves state-of-the-art performance while reducing 50% parameters and 53% FLOPs compared to others.

Network pruning reduces the computation costs of an over-parameterized network without performance damage. Prevailing pruning algorithms pre-define the width and depth of the pruned networks, and then transfer parameters from the unpruned network to pruned networks. To break the structure limitation of the pruned networks, we propose to apply neural architecture search to search directly for a network with flexible channel and layer sizes. The number of the channels/layers is learned by minimizing the loss of the pruned networks. The feature map of the pruned network is an aggregation of K feature map fragments (generated by K networks of different sizes), which are sampled based on the probability distribution. The loss can be back-propagated not only to the network weights, but also to the parameterized distribution to explicitly tune the size of the channels/layers. Specifically, we apply channel-wise interpolation to keep the feature map with different channel sizes aligned in the aggregation procedure. The maximum probability for the size in each distribution serves as the width and depth of the pruned network, whose parameters are learned by knowledge transfer, e.g., knowledge distillation, from the original networks. Experiments on CIFAR-10, CIFAR-100 and ImageNet demonstrate the effectiveness of our new perspective of network pruning compared to traditional network pruning algorithms. Various searching and knowledge transfer approaches are conducted to show the effectiveness of the two components. Code is at: https://github.com/D-X-Y/NAS-Projects

In this paper, we study object detection using a large pool of unlabeled images and only a few labeled images per category, named "few-example object detection". The key challenge consists in generating trustworthy training samples as many as possible from the pool. Using few training examples as seeds, our method iterates between model training and high-confidence sample selection. In training, easy samples are generated first and, then the poorly initialized model undergoes improvement. As the model becomes more discriminative, challenging but reliable samples are selected. After that, another round of model improvement takes place. To further improve the precision and recall of the generated training samples, we embed multiple detection models in our framework, which has proven to outperform the single model baseline and the model ensemble method. Experiments on PASCAL VOC'07, MS COCO'14, and ILSVRC'13 indicate that by using as few as three or four samples selected for each category, our method produces very competitive results when compared to the state-of-the-art weakly-supervised approaches using a large number of image-level labels.

In many real-world applications, data can be represented by multiple ways or multi-view features to describe various characteristics of data. In this sense, the prediction performance can be significantly improved by taking advantages of these features together. Late fusion, which combines the predictions of multiple features, is a commonly used approach to make the final decision for a test instance. However, it is ubiquitous that different features dispute the prediction on the same data with each other, leading to performance degeneration. In this paper, we propose an efficient and effective matrix factorization-based approach to fuse predictions from multiple sources. This approach leverages a hard constraint on the matrix rank to preserve the consistency of predictions by various features, and we thus named it as Hard-rank Constraint Matrix Factorization-based fusion (HCMF). HCMF can avoid the performance degeneration caused by the controversy of multiple features. Extensive experiments demonstrate the efficacy of HCMF for outlier detection and the performance improvement, which outperforms the state-of-the-art late fusion algorithms on many data sets.

In this paper, we focus on the one-example person re-identification (re-ID) task, where each identity has only one labeled example along with many unlabeled examples. We propose a progressive framework which gradually exploits the unlabeled data for person re-ID. In this framework, we iteratively (1) update the Convolutional Neural Network (CNN) model and (2) estimate pseudo labels for the unlabeled data. We split the training data into three parts, i.e., labeled data, pseudo-labeled data, and indexlabeled data. Initially, the re-ID model is trained using the labeled data. For the subsequent model training, we update the CNN model by the joint training on the three data parts. The proposed joint training method can optimize the model by both the data with labels (or pseudo labels) and the data without any reliable labels. For the label estimation step, instead of using a static sampling strategy, we propose a progressive sampling strategy to increase the number of the selected pseudo-labeled candidates step by step. We select a few candidates with most reliable pseudo labels from unlabeled examples as the pseudo-labeled data, and keep the rest as index-labeled data by assigning them with the data indexes. During iterations, the index-labeled data are dynamically transferred to pseudo-labeled data. Notably, the rank-1 accuracy of our method outperforms the state-of-the-art method by 21.6 points (absolute, i.e., 62.8% vs. 41.2%) on MARS, and 16.6 points on DukeMTMC-VideoReID. Extended to the few-example setting, our approach with only 20% labeled data surprisingly achieves comparable performance to the supervised state-of-the-art method with 100% labeled data.

Recent advances in facial landmark detection achieve success by learning discriminative features from rich deformation of face shapes and poses. Besides the variance of faces themselves, the intrinsic variance of image styles, e.g., grayscale vs. color images, light vs. dark, intense vs. dull, and so on, has constantly been overlooked. This issue becomes inevitable as increasing web images are collected from various sources for training neural networks. In this work, we propose a style-aggregated approach to deal with the large intrinsic variance of image styles for facial landmark detection. Our method transforms original face images to style-aggregated images by a generative adversarial module. The proposed scheme uses the style-aggregated image to maintain face images that are more robust to environmental changes. Then the original face images accompanying with style-aggregated ones play a duet to train a landmark detector which is complementary to each other. In this way, for each face, our method takes two images as input, i.e., one in its original style and the other in the aggregated style. In experiments, we observe that the large variance of image styles would degenerate the performance of facial landmark detectors. Moreover, we show the robustness of our method to the large variance of image styles by comparing to a variant of our approach, in which the generative adversarial module is removed, and no style-aggregated images are used. Our approach is demonstrated to perform well when compared with state-of-the-art algorithms on benchmark datasets AFLW and 300-W. Code is publicly available on GitHub: https://github.com/D-X-Y/SAN.

In this paper, we present supervision-by-registration, an unsupervised approach to improve the precision of facial landmark detectors on both images and video. Our key observation is that the detections of the same landmark in adjacent frames should be coherent with registration, i.e., optical flow. Interestingly, coherency of optical flow is a source of supervision that does not require manual labeling, and can be leveraged during detector training. For example, we can enforce in the training loss function that a detected landmark at framet-1followed by optical flow tracking from framet-1to frametshould coincide with the location of the detection at framet. Essentially, supervision-by-registration augments the training loss function with a registration loss, thus training the detector to have output that is not only close to the annotations in labeled images, but also consistent with registration on large amounts of unlabeled videos. End-to-end training with the registration loss is made possible by a differentiable Lucas-Kanade operation, which computes optical flow registration in the forward pass, and back-propagates gradients that encourage temporal coherency in the detector. The output of our method is a more precise image-based facial landmark detector, which can be applied to single images or video. With supervision-by-registration, we demonstrate (1) improvements in facial landmark detection on both images (300W, ALFW) and video (300VW, Youtube-Celebrities), and (2) significant reduction of jittering in video detections.

We focus on the one-shot learning for video-based person re-Identification (re-ID). Unlabeled tracklets for the person re-ID tasks can be easily obtained by preprocessing, such as pedestrian detection and tracking. In this paper, we propose an approach to exploiting unlabeled tracklets by gradually but steadily improving the discriminative capability of the Convolutional Neural Network (CNN) feature representation via stepwise learning. We first initialize a CNN model using one labeled tracklet for each identity. Then we update the CNN model by the following two steps iteratively: 1. sample a few candidates with most reliable pseudo labels from unlabeled tracklets; 2. update the CNN model according to the selected data. Instead of the static sampling strategy applied in existing works, we propose a progressive sampling method to increase the number of the selected pseudo-labeled candidates step by step. We systematically investigate the way how we should select pseudo-labeled tracklets into the training set to make the best use of them. Notably, the rank-1 accuracy of our method outperforms the state-of-the-art method by 21.46 points (absolute, i.e., 62.67% vs. 41.21%) on the MARS dataset, and 16.53 points on the DukeMTMC-VideoReID dataset.

This paper proposed a Soft Filter Pruning (SFP) method to accelerate the inference procedure of deep Convolutional Neural Networks (CNNs). Specifically, the proposed SFP enables the pruned filters to be updated when training the model after pruning. SFP has two advantages over previous works: (1) Larger model capacity. Updating previously pruned filters provides our approach with larger optimization space than fixing the filters to zero. Therefore, the network trained by our method has a larger model capacity to learn from the training data. (2) Less dependence on the pretrained model. Large capacity enables SFP to train from scratch and prune the model simultaneously. In contrast, previous filter pruning methods should be conducted on the basis of the pre-trained model to guarantee their performance. Empirically, SFP from scratch outperforms the previous filter pruning methods. Moreover, our approach has been demonstrated effective for many advanced CNN architectures. Notably, on ILSCRC-2012, SFP reduces more than 42% FLOPs on ResNet-101 with even 0.2% top-5 accuracy improvement, which has advanced the state-of-the-art. Code is publicly available on GitHub: https://github.com/he-y/softfilter-pruning

Given only a few image-text pairs, humans can learn to detect semantic concepts and describe the content. For machine learning algorithms, they usually require a lot of data to train a deep neural network to solve the problem. However, it is challenging for the existing systems to generalize well to the few-shot multi-modal scenario, because the learner should understand not only images and texts but also their relationships from only a few examples. In this paper, we tackle two multi-modal problems, i.e., image captioning and visual question answering (VQA), in the few-shot setting.

In this paper, we present a novel and general network structure towards accelerating the inference process of convolutional neural networks, which is more complicated in network structure yet with less inference complexity. The core idea is to equip each original convolutional layer with another low-cost collaborative layer (LCCL), and the element-wise multiplication of the ReLU outputs of these two parallel layers produces the layer-wise output. The combined layer is potentially more discriminative than the original convolutional layer, and its inference is faster for two reasons: 1) the zero cells of the LCCL feature maps will remain zero after element-wise multiplication, and thus it is safe to skip the calculation of the corresponding high-cost convolution in the original convolutional layer; 2) LCCL is very fast if it is implemented as a 1 × 1 convolution or only a single filter shared by all channels. Extensive experiments on the CIFAR-10, CIFAR-100 and ILSCRC-2012 benchmarks show that our proposed network structure can accelerate the inference process by 32% on average with negligible performance drop.

Co-training is a well-known semi-supervised learning approach which trains classifiers on two different views and exchanges labels of unlabeled instances in an iterative way. During co-training process, labels of unlabeled instances in the training pool are very likely to be false especially in the initial training rounds, while the standard co-training algorithm utilizes a “draw without replacement” manner and does not remove these false labeled instances from training. This issue not only tends to degenerate its performance but also hampers its fundamental theory. Besides, there is no optimization model to explain what objective a cotraining process optimizes. To these issues, in this study we design a new co-training algorithm named self-paced cotraining (SPaCo) with a “draw with replacement” learning mode. The rationality of SPaCo can be proved under theoretical assumptions utilized in traditional co-training research, and furthermore, the algorithm exactly complies with the alternative optimization process for an optimization model of self-paced curriculum learning, which can be finely explained in robust learning manner. Experimental results substantiate the superiority of the proposed method as compared with current state-of-the-art co-training methods.

A major challenge that arises in Weakly Supervised Object Detection (WSOD) is that only image-level labels are available, whereas WSOD trains instance-level object detectors. A typical approach to WSOD is to 1) generate a series of region proposals for each image and assign the image-level label to all the proposals in that image; 2) train a classifier using all the proposals; and 3) use the classifier to select proposals with high confidence scores as the positive instances for another round of training. In this way, the image-level labels are iteratively transferred to instance-level labels.