The perceptual loss has been widely used as an effective loss term in image synthesis tasks including image super-resolution, and style transfer. It was believed that the success lies in the high-level perceptual feature representations extracted from CNNs pretrained with a large set of images. Here we reveal that, what matters is the network structure instead of the trained weights. Without any learning, the structure of a deep network is sufficient to capture the dependencies between multiple levels of variable statistics using multiple layers of CNNs. This insight removes the requirements of pre-training and a particular network structure (commonly, VGG) that are previously assumed for the perceptual loss, thus enabling a significantly wider range of applications. To this end, we demonstrate that a randomly-weighted deep CNN can be used to model the structured dependencies of outputs. On a few dense per-pixel prediction tasks such as semantic segmentation, depth estimation and instance segmentation, we show improved results of using the extended randomized perceptual loss, compared to the baselines using pixel-wise loss alone. We hope that this simple, extended perceptual loss may serve as a generic structured-output loss that is applicable to most structured output learning tasks.
In online applications with streaming data, awareness of how far the training or test set has shifted away from the original dataset can be crucial to the performance of the model. However, we may not have access to historical samples in the data stream. To cope with such situations, we propose a novel method, Continual Density Ratio Estimation (CDRE), for estimating density ratios between the initial and current distributions ($p/q_t$) of a data stream in an iterative fashion without the need of storing past samples, where $q_t$ is shifting away from $p$ over time $t$. We demonstrate that CDRE can be more accurate than standard DRE in terms of estimating divergences between distributions, despite not requiring samples from the original distribution. CDRE can be applied in scenarios of online learning, such as importance weighted covariate shift, tracing dataset changes for better decision making. In addition, (CDRE) enables the evaluation of generative models under the setting of continual learning. To the best of our knowledge, there is no existing method that can evaluate generative models in continual learning without storing samples from the original distribution.
As the successor of H.265/HEVC, the new versatile video coding standard (H.266/VVC) can provide up to 50% bitrate saving with the same subjective quality, at the cost of increased decoding complexity. To accelerate the application of the new coding standard, a real-time H.266/VVC software decoder that can support various platforms is implemented, where SIMD technologies, parallelism optimization, and the acceleration strategies based on the characteristics of each coding tool are applied. As the mobile devices have become an essential carrier for video services nowadays, the mentioned optimization efforts are not only implemented for the x86 platform, but more importantly utilized to highly optimize the decoding performance on the ARM platform in this work. The experimental results show that when running on the Apple A14 SoC (iPhone 12pro), the average single-thread decoding speed of the present implementation can achieve 53fps (RA and LB) for full HD (1080p) bitstreams generated by VTM-11.0 reference software using 8bit Common Test Conditions (CTC). When multi-threading is enabled, an average of 32 fps (RA) can be achieved when decoding the 4K bitstreams.
In an order-driven financial market, the price of a financial asset is discovered through the interaction of orders - requests to buy or sell at a particular price - that are posted to the public limit order book (LOB). Therefore, LOB data is extremely valuable for modelling market dynamics. However, LOB data is not freely accessible, which poses a challenge to market participants and researchers wishing to exploit this information. Fortunately, trades and quotes (TAQ) data - orders arriving at the top of the LOB, and trades executing in the market - are more readily available. In this paper, we present the LOB recreation model, a first attempt from a deep learning perspective to recreate the top five price levels of the LOB for small-tick stocks using only TAQ data. Volumes of orders sitting deep in the LOB are predicted by combining outputs from: (1) a history compiler that uses a Gated Recurrent Unit (GRU) module to selectively compile prediction relevant quote history; (2) a market events simulator, which uses an Ordinary Differential Equation Recurrent Neural Network (ODE-RNN) to simulate the accumulation of net order arrivals; and (3) a weighting scheme to adaptively combine the predictions generated by (1) and (2). By the paradigm of transfer learning, the source model trained on one stock can be fine-tuned to enable application to other financial assets of the same class with much lower demand on additional data. Comprehensive experiments conducted on two real world intraday LOB datasets demonstrate that the proposed model can efficiently recreate the LOB with high accuracy using only TAQ data as input.
The majority of Chinese characters are monophonic, i.e.their pronunciations are unique and thus can be induced easily using a check table. As for their counterparts, polyphonic characters have more than one pronunciation. To perform linguistic computation tasks related to spoken Mandarin Chinese, the correct pronunciation for each polyphone must be identified among several candidates according to its context. This process is called Polyphone Disambiguation, a key procedure in the Grapheme-to-phoneme (G2P) conversion step of a Chinese text-to-speech (TTS) system. The problem is well explored with both knowledge-based and learning-based approaches, yet it remains challenging due to the lack of publicly available datasets and complex language phenomenon concerned polyphone. In this paper, we propose a novel semi-supervised learning (SSL) framework for Mandarin Chinese polyphone disambiguation that can potentially leverage unlimited unlabeled text data. We explore the effect of various proxy labeling strategies including entropy-thresholding and lexicon-based labeling. As for the architecture, a pre-trained model of Electra is combined with Convolution BLSTM layers to fine-tune on our task. Qualitative and quantitative experiments demonstrate that our method achieves state-of-the-art performance in Mandarin Chinese polyphone disambiguation. In addition, we publish a novel dataset specifically for the polyphone disambiguation task to promote further researches.
We address rotation averaging (RA) and its application to real-world 3D reconstruction. Local optimisation based approaches are the defacto choice, though they only guarantee a local optimum. Global optimizers ensure global optimality in low noise conditions, but they are inefficient and may easily deviate under the influence of outliers or elevated noise levels. We push the envelope of rotation averaging by leveraging the advantages of global RA method and local RA method. Combined with a fast view graph filtering as preprocessing, the proposed hybrid approach is robust to outliers. We apply the proposed hybrid rotation averaging approach to incremental Structure from Motion (SfM) by adding the resulting global rotations as regularizers to bundle adjustment. Overall, we demonstrate high practicality of the proposed method as bad camera poses are effectively corrected and drift is reduced.
We address rotation averaging and its application to real-world 3D reconstruction. Local optimisation based approaches are the defacto choice, though they only guarantee a local optimum. Global optimizers ensure global optimality in low noise conditions, but they are inefficient and may easily deviate under the influence of outliers or elevated noise levels. We push the envelope of global rotation averaging by formulating it as a semi-definite program that can be solved efficiently by applying the Burer-Monteiro method. Both memory and time requirements are thereby largely reduced through a low-rank factorisation. Combined with a fast view graph filtering as preprocessing, and a local optimiser as post-processing, the proposed hybrid approach is robust to outliers. Compared against state-of-the-art globally optimal methods, our approach is 1 ~ 2 orders of magnitude faster while maintaining the same or better accuracy. We apply the proposed hybrid rotation averaging approach to incremental Structure from Motion (SfM) by adding the resulting global rotations as regularizers to bundle adjustment. Overall, we demonstrate high practicality of the proposed method as bad camera poses are effectively corrected and drift is reduced.
Food recommendation has become an important means to help guide users to adopt healthy dietary habits. Previous works on food recommendation either i) fail to consider users' explicit requirements, ii) ignore crucial health factors (e.g., allergies and nutrition needs), or iii) do not utilize the rich food knowledge for recommending healthy recipes. To address these limitations, we propose a novel problem formulation for food recommendation, modeling this task as constrained question answering over a large-scale food knowledge base/graph (KBQA). Besides the requirements from the user query, personalized requirements from the user's dietary preferences and health guidelines are handled in a unified way as additional constraints to the QA system. To validate this idea, we create a QA style dataset for personalized food recommendation based on a large-scale food knowledge graph and health guidelines. Furthermore, we propose a KBQA-based personalized food recommendation framework which is equipped with novel techniques for handling negations and numerical comparisons in the queries. Experimental results on the benchmark show that our approach significantly outperforms non-personalized counterparts (average 59.7% absolute improvement across various evaluation metrics), and is able to recommend more relevant and healthier recipes.
Despite the recent successes of deep learning in natural language processing (NLP), there remains widespread usage of and demand for techniques that do not rely on machine learning. The advantage of these techniques is their interpretability and low cost when compared to frequently opaque and expensive machine learning models. Although they may not be be as performant in all cases, they are often sufficient for common and relatively simple problems. In this paper, we aim to modernize these older methods while retaining their advantages by extending approaches from categorical or bag-of-words representations to word embeddings representations in the latent space. First, we show that entropy and Kullback-Leibler divergence can be efficiently estimated using word embeddings and use this estimation to compare text across several categories. Next, we recast the heavy-tailed distribution known as Zipf's law that is frequently observed in the categorical space to the latent space. Finally, we look to improve the Jaccard similarity measure for sentence suggestion by introducing a new method of identifying similar sentences based on the set cover problem. We compare the performance of this algorithm against several baselines including Word Mover's Distance and the Levenshtein distance.
Due to the high variability of the traffic in the radio access network (RAN), fixed network configurations are not flexible to achieve the optimal performance. Our vendors provide several settings of the eNodeB to optimize the RAN performance, such as media access control scheduler, loading balance, etc. But the detailed mechanisms of the eNodeB configurations are usually very complicated and not disclosed, not to mention the large KPIs space needed to be considered. These make constructing simulator, offline tuning, or rule-based solutions difficult. We aim to build an intelligent controller without strong assumption or domain knowledge about the RAN and can run for 24/7 without supervision. To achieve this goal, we first build a closed-loop control testbed RAN in a lab environment with one eNodeB provided by one of the largest wireless vendors and four smartphones. Next, we build a double Q network agent that is trained with the live feedbacks of the key performance indicators from the RAN. Our work proved the effectiveness of applying deep reinforcement learning to improve network performance in a real RAN network environment.