Early exiting allows instances to exit at different layers according to the estimation of difficulty. Previous works usually adopt heuristic metrics such as the entropy of internal outputs to measure instance difficulty, which suffers from generalization and threshold-tuning. In contrast, learning to exit, or learning to predict instance difficulty is a more appealing way. Though some effort has been devoted to employing such "learn-to-exit" modules, it is still unknown whether and how well the instance difficulty can be learned. As a response, we first conduct experiments on the learnability of instance difficulty, which demonstrates that modern neural models perform poorly on predicting instance difficulty. Based on this observation, we propose a simple-yet-effective Hash-based Early Exiting approach (HashEE) that replaces the learn-to-exit modules with hash functions to assign each token to a fixed exiting layer. Different from previous methods, HashEE requires no internal classifiers nor extra parameters, and therefore is more efficient. Experimental results on classification, regression, and generation tasks demonstrate that HashEE can achieve higher performance with fewer FLOPs and inference time compared with previous state-of-the-art early exiting methods.
Robustness and discrimination power are two fundamental requirements in visual object tracking. In most tracking paradigms, we find that the features extracted by the popular Siamese-like networks cannot fully discriminatively model the tracked targets and distractor objects, hindering them from simultaneously meeting these two requirements. While most methods focus on designing robust correlation operations, we propose a novel target-dependent feature network inspired by the self-/cross-attention scheme. In contrast to the Siamese-like feature extraction, our network deeply embeds cross-image feature correlation in multiple layers of the feature network. By extensively matching the features of the two images through multiple layers, it is able to suppress non-target features, resulting in instance-varying feature extraction. The output features of the search image can be directly used for predicting target locations without extra correlation step. Moreover, our model can be flexibly pre-trained on abundant unpaired images, leading to notably faster convergence than the existing methods. Extensive experiments show our method achieves the state-of-the-art results while running at real-time. Our feature networks also can be applied to existing tracking pipelines seamlessly to raise the tracking performance. Code will be available.
Semantic segmentation is a difficult task even when trained in a supervised manner on photographs. In this paper, we tackle the problem of semantic segmentation of artistic paintings, an even more challenging task because of a much larger diversity in colors, textures, and shapes and because there are no ground truth annotations available for segmentation. We propose an unsupervised method for semantic segmentation of paintings using domain adaptation. Our approach creates a training set of pseudo-paintings in specific artistic styles by using style-transfer on the PASCAL VOC 2012 dataset, and then applies domain confusion between PASCAL VOC 2012 and real paintings. These two steps build on a new dataset we gathered called DRAM (Diverse Realism in Art Movements) composed of figurative art paintings from four movements, which are highly diverse in pattern, color, and geometry. To segment new paintings, we present a composite multi-domain adaptation method that trains on each sub-domain separately and composes their solutions during inference time. Our method provides better segmentation results not only on the specific artistic movements of DRAM, but also on other, unseen ones. We compare our approach to alternative methods and show applications of semantic segmentation in art paintings. The code and models for our approach are publicly available at: https://github.com/Nadavc220/SemanticSegmentationInArtPaintings.
We introduce UDAAN, an open-source post-editing tool that can reduce manual editing efforts to quickly produce publishable-standard documents in different languages. UDAAN has an end-to-end Machine Translation (MT) plus post-editing pipeline wherein users can upload a document to obtain raw MT output. Further, users can edit the raw translations using our tool. UDAAN offers several advantages: a) Domain-aware, vocabulary-based lexical constrained MT. b) source-target and target-target lexicon suggestions for users. Replacements are based on the source and target texts lexicon alignment. c) Suggestions for translations are based on logs created during user interaction. d) Source-target sentence alignment visualisation that reduces the cognitive load of users during editing. e) Translated outputs from our tool are available in multiple formats: docs, latex, and PDF. Although we limit our experiments to English-to-Hindi translation for the current study, our tool is independent of the source and target languages. Experimental results based on the usage of the tools and users feedback show that our tool speeds up the translation time approximately by a factor of three compared to the baseline method of translating documents from scratch.
In scalable machine learning systems, model training is often parallelized over multiple nodes that run without tight synchronization. Most analysis results for the related asynchronous algorithms use an upper bound on the information delays in the system to determine learning rates. Not only are such bounds hard to obtain in advance, but they also result in unnecessarily slow convergence. In this paper, we show that it is possible to use learning rates that depend on the actual time-varying delays in the system. We develop general convergence results for delay-adaptive asynchronous iterations and specialize these to proximal incremental gradient descent and block-coordinate descent algorithms. For each of these methods, we demonstrate how delays can be measured on-line, present delay-adaptive step-size policies, and illustrate their theoretical and practical advantages over the state-of-the-art.
Efficient data collection methods play a major role in helping us better understand the Earth and its ecosystems. In many applications, the usage of unmanned aerial vehicles (UAVs) for monitoring and remote sensing is rapidly gaining momentum due to their high mobility, low cost, and flexible deployment. A key challenge is planning missions to maximize the value of acquired data in large environments given flight time limitations. This is, for example, relevant for monitoring agricultural fields. This paper addresses the problem of adaptive path planning for accurate semantic segmentation of using UAVs. We propose an online planning algorithm which adapts the UAV paths to obtain high-resolution semantic segmentations necessary in areas with fine details as they are detected in incoming images. This enables us to perform close inspections at low altitudes only where required, without wasting energy on exhaustive mapping at maximum image resolution. A key feature of our approach is a new accuracy model for deep learning-based architectures that captures the relationship between UAV altitude and semantic segmentation accuracy. We evaluate our approach on different domains using real-world data, proving the efficacy and generability of our solution.
Motivated by their broad applications in reinforcement learning, we study the linear two-time-scale stochastic approximation, an iterative method using two different step sizes for finding the solutions of a system of two equations. Our main focus is to characterize the finite-time complexity of this method under time-varying step sizes and Markovian noise. In particular, we show that the mean square errors of the variables generated by the method converge to zero at a sublinear rate $\Ocal(k^{2/3})$, where $k$ is the number of iterations. We then improve the performance of this method by considering the restarting scheme, where we restart the algorithm after every predetermined number of iterations. We show that using this restarting method the complexity of the algorithm under time-varying step sizes is as good as the one using constant step sizes, but still achieving an exact converge to the desired solution. Moreover, the restarting scheme also helps to prevent the step sizes from getting too small, which is useful for the practical implementation of the linear two-time-scale stochastic approximation.
Background: Computational models of speech recognition often assume that the set of target words is already given. This implies that these models do not learn to recognise speech from scratch without prior knowledge and explicit supervision. Visually grounded speech models learn to recognise speech without prior knowledge by exploiting statistical dependencies between spoken and visual input. While it has previously been shown that visually grounded speech models learn to recognise the presence of words in the input, we explicitly investigate such a model as a model of human speech recognition. Methods: We investigate the time-course of word recognition as simulated by the model using a gating paradigm to test whether its recognition is affected by well-known word-competition effects in human speech processing. We furthermore investigate whether vector quantisation, a technique for discrete representation learning, aids the model in the discovery and recognition of words. Results/Conclusion: Our experiments show that the model is able to recognise nouns in isolation and even learns to properly differentiate between plural and singular nouns. We also find that recognition is influenced by word competition from the word-initial cohort and neighbourhood density, mirroring word competition effects in human speech comprehension. Lastly, we find no evidence that vector quantisation is helpful in discovering and recognising words. Our gating experiments even show that the vector quantised model requires more of the input sequence for correct recognition.
As a relatively new field, network neuroscience has tended to focus on aggregate behaviours of the brain averaged over many successive experiments or over long recordings in order to construct robust brain models. These models are limited in their ability to explain dynamic state changes in the brain which occurs spontaneously as a result of normal brain function. Hidden Markov Models (HMMs) trained on neuroimaging time series data have since arisen as a method to produce dynamical models that are easy to train but can be difficult to fully parametrise or analyse. We propose an interpretation of these neural HMMs as multiplex brain state graph models we term Hidden Markov Graph Models (HMGMs). This interpretation allows for dynamic brain activity to be analysed using the full repertoire of network analysis techniques. Furthermore, we propose a general method for selecting HMM hyperparameters in the absence of external data, based on the principle of maximum entropy, and use this to select the number of layers in the multiplex model. We produce a new tool for determining important communities of brain regions using a spatiotemporal random walk-based procedure that takes advantage of the underlying Markov structure of the model. Our analysis of real multi-subject fMRI data provides new results that corroborate the modular processing hypothesis of the brain at rest as well as contributing new evidence of functional overlap between and within dynamic brain state communities. Our analysis pipeline provides a way to characterise dynamic network activity of the brain under novel behaviours or conditions.
The multi-dithering method has been well verified in phase locking of polarization coherent combination experiment. However, it is hard to apply to low repetition frequency pulsed lasers, since there exists an overlap frequency domain between pulse laser and the amplitude phase noise and traditional filters cannot effectively separate phase noise. Aiming to solve the problem in this paper, we propose a novel method of pulse noise detection, identification, and filtering based on the autocorrelation characteristics between noise signals. In the proposed algorithm, a self-designed window algorithm is used to identify the pulse, and then the pulse signal group in the window is replaced by interpolation, which effectively filter the pulse signal doped in the phase noise within 0.1 ms. After filtering the pulses in the phase noise, the phase difference of two pulsed beams (10 kHz) is successfully compensated to zero in 1 ms, and the coherent combination of closed-loop phase lock is realized. At the same time, the phase correction times are few, the phase lock effect is stable, and the final light intensity increases to the ideal value (0.9 Imax).