Search

Minimizers and convolutional neural networks (CNNs) are two quite distinct popular techniques that have both been employed to analyze categorical biological sequences. At face value, the methods seem entirely dissimilar. Minimizers use min-wise hashing on a rolling window to extract a single important k-mer feature per window. CNNs start with a wide array of randomly initialized convolutional filters, paired with a pooling operation, and then multiple additional neural layers to learn both the filters themselves and how they can be used to classify the sequence. In this study, our main result is a careful mathematical analysis of hash function properties showing that for sequences over a categorical alphabet, random Gaussian initialization of convolutional filters with max-pooling is equivalent to choosing a minimizer ordering such that selected k-mers are (in Hamming distance) far from the k-mers within the sequence but close to other minimizers. In empirical experiments, we find that this property manifests as decreased density in repetitive regions, both in simulation and on real human telomeres. We additionally train from scratch a CNN embedding of synthetic short-reads from the SARS-CoV-2 genome into 3D Euclidean space that locally recapitulates the linear sequence distance of the read origins, a modest step toward building a deep learning assembler, although it is at present too slow to be practical. In total, this article provides a partial explanation for the effectiveness of CNNs in categorical sequence analysis. [ABSTRACT FROM AUTHOR]

Purpose: BUS‐Set is a reproducible benchmark for breast ultrasound (BUS) lesion segmentation, comprising of publicly available images with the aim of improving future comparisons between machine learning models within the field of BUS. Method: Four publicly available datasets were compiled creating an overall set of 1154 BUS images, from five different scanner types. Full dataset details have been provided, which include clinical labels and detailed annotations. Furthermore, nine state‐of‐the‐art deep learning architectures were selected to form the initial benchmark segmentation result, tested using five‐fold cross‐validation and MANOVA/ANOVA with Tukey statistical significance test with a threshold of 0.01. Additional evaluation of these architectures was conducted, exploring possible training bias, and lesion size and type effects. Results: Of the nine state‐of‐the‐art benchmarked architectures, Mask R‐CNN obtained the highest overall results, with the following mean metric scores: Dice score of 0.851, intersection over union of 0.786 and pixel accuracy of 0.975. MANOVA/ANOVA and Tukey test results showed Mask R‐CNN to be statistically significant better compared to all other benchmarked models with a p‐value >0.01. Moreover, Mask R‐CNN achieved the highest mean Dice score of 0.839 on an additional 16 image dataset, that contained multiple lesions per image. Further analysis on regions of interest was conducted, assessing Hamming distance, depth‐to‐width ratio (DWR), circularity, and elongation, which showed that the Mask R‐CNN's segmentations maintained the most morphological features with correlation coefficients of 0.888, 0.532, 0.876 for DWR, circularity, and elongation, respectively. Based on the correlation coefficients, statistical test indicated that Mask R‐CNN was only significantly different to Sk‐U‐Net. Conclusions: BUS‐Set is a fully reproducible benchmark for BUS lesion segmentation obtained through the use of public datasets and GitHub. Of the state‐of‐the‐art convolution neural network (CNN)‐based architectures, Mask R‐CNN achieved the highest performance overall, further analysis indicated that a training bias may have occurred due to the lesion size variation in the dataset. All dataset and architecture details are available at GitHub: https://github.com/corcor27/BUS‐Set, which allows for a fully reproducible benchmark. [ABSTRACT FROM AUTHOR]

CNN-based hashing methods have greatly boosted the performance of image retrieval, under the strong supervision of large amounts of manually annotated labels. In recent years, a large number of social media images with user tags have been generated on the Internet. These images can be regarded as weakly labeled training data, which can provide rich samples for training hash network, and greatly reduce the cost of obtaining training data. However, there are noise and visual irrelevant tags in user tags, and different tags may describe different objects in the image. In the previous CNN-based hashing method, a training image usually corresponds to a manual label and generates a hash code. These methods are difficult to use the image described by user tags with noise. For solving the above problem, we propose a CNN-based objectness hash learning method using user tags as a guide for training. First of all, the user tags are roughly filtered to remove noise tags that are not related to the visual content of images. Secondly, we quantify user tags into a unified semantic space and extract the highest-frequency words of the semantic space from similar objectness areas as their labels. Then, these objectness areas with their labels are grouped into a series of triple units as training data. So that the generated hash code can inherit the semantic similarity of the objectness areas well, that is, the Hamming distance between hash codes generated by similar objectness areas is closer, the reverse is the farther. Experimental results on NUS-WIDE and Flickr datasets show that our method can effectively extract object-level semantic information from weak user tags, and improve the accuracy of image retrieval. [ABSTRACT FROM AUTHOR]

Background: Protein–protein interaction (PPI) is vital for life processes, disease treatment, and drug discovery. The computational prediction of PPI is relatively inexpensive and efficient when compared to traditional wet-lab experiments. Given a new protein, one may wish to find whether the protein has any PPI relationship with other existing proteins. Current computational PPI prediction methods usually compare the new protein to existing proteins one by one in a pairwise manner. This is time consuming. Results: In this work, we propose a more efficient model, called deep hash learning protein-and-protein interaction (DHL-PPI), to predict all-against-all PPI relationships in a database of proteins. First, DHL-PPI encodes a protein sequence into a binary hash code based on deep features extracted from the protein sequences using deep learning techniques. This encoding scheme enables us to turn the PPI discrimination problem into a much simpler searching problem. The binary hash code for a protein sequence can be regarded as a number. Thus, in the pre-screening stage of DHL-PPI, the string matching problem of comparing a protein sequence against a database with M proteins can be transformed into a much more simpler problem: to find a number inside a sorted array of length M. This pre-screening process narrows down the search to a much smaller set of candidate proteins for further confirmation. As a final step, DHL-PPI uses the Hamming distance to verify the final PPI relationship. Conclusions: The experimental results confirmed that DHL-PPI is feasible and effective. Using a dataset with strictly negative PPI examples of four species, DHL-PPI is shown to be superior or competitive when compared to the other state-of-the-art methods in terms of precision, recall or F1 score. Furthermore, in the prediction stage, the proposed DHL-PPI reduced the time complexity from O (M 2) to O (M log M) for performing an all-against-all PPI prediction for a database with M proteins. With the proposed approach, a protein database can be preprocessed and stored for later search using the proposed encoding scheme. This can provide a more efficient way to cope with the rapidly increasing volume of protein datasets. [ABSTRACT FROM AUTHOR]

Recommendation efficiency and data sparsity problems have been regarded as two challenges of improving performance for online recommendation. Most of the previous related work focus on improving recommendation accuracy instead of efficiency. In this paper, we propose a Deep Pairwise Hashing (DPH) to map users and items to binary vectors in Hamming space, where a user's preference for an item can be efficiently calculated by Hamming distance, which significantly improves the efficiency of online recommendation. To alleviate data sparsity and cold-start problems, the user-item interactive information and item content information are unified to learn effective representations of items and users. Specifically, we first pre-train robust item representation from item content data by a Denoising Auto-encoder instead of other deterministic deep learning frameworks; then we finetune the entire framework by adding a pairwise loss objective with discrete constraints; moreover, DPH aims to minimize a pairwise ranking loss that is consistent with the ultimate goal of recommendation. Finally, we adopt the alternating optimization method to optimize the proposed model with discrete constraints. Extensive experiments on three different datasets show that DPH can significantly advance the state-of-the-art frameworks regarding data sparsity and item cold-start recommendation. [ABSTRACT FROM AUTHOR]

Face swapping is an exciting new technology that allows the identification of a source face to be transferred to a target face while maintaining the target face's attributes. As a result, this study uses deep learning and subjective assessment for face swapping detection. In this study, face-swapping generates image-realistic and time-coherent sequences. Human recognition can aid in the prevention of face swapping attacks. Human volunteers determine accuracy in reconciling faces. They may believe a photograph is a forgery. Rankings of GIF emotions: A pair of Hamming-LUCBs is employed when you click an image. Selecting a switched type has a 50% failure rate. The CNN-LSTM and RNN-LSTM classifiers are compared. The RNN-LSTM algorithm outperformed the CNN-LSTM algorithm. Rather than that, we learned by exchanging faces. Validation made use of the system's default hyper parameters and epochs. We train and publish separately on both strategies. As a result, we compared them. 25-50-85 LSTM (CNN & RNN). We quantify the model's output. The dimensions of the faces are 0.2865 and 0.0415. R/F equals 0.1106. Real and fictitious faces have values of 0.3701 and 0.4229. Between 0.3741 and 0.1175. In Ae-GAN generated fictitious images, human connectedness is imprecise. The true identities of Nirkin and AEGAN (around 0.09 difference). This study makes extensive use of transfer learning. The Best Face Swap Detection Photos in the World Each model has over 1000 images taken in real life (the largest known). The model was confused because we put it to the test on humans. It makes a comparison of two photos. An image can identify an imposter; the context would then make it very evident that our model does not exist. As a result, the model can perceive a human. [ABSTRACT FROM AUTHOR]

In this work, we advance the security promise of split manufacturing through judicious handling of interconnects. First, we study the cost-security tradeoffs underlying for split manufacturing, which are limiting its adoption. Next, aiming to resolve these concerns, we propose three effective and efficient strategies to dedicatedly lift nets to higher metal layers. Toward this end, we design custom “elevating cells” and devise procedures for routing blockages. All our techniques are employed in a commercial-grade computer-aided design (CAD) framework. For our security analysis, we leverage various state-of-the-art attacks (network flow-based attack, routing-congestion-aware attack, and deep learning-based attack), established metrics (correct connection rate, output error rate, and Hamming distance), and advanced metrics (percentage of netlist recovery and mutual information). Our extensive experiments show that our scheme provides superior protection. Simultaneously, we induce reasonably low and controllable overheads on power and performance, without any silicon area costs. Besides, we support higher split layers, which helps to alleviate concerns on the practicality of split manufacturing. [ABSTRACT FROM AUTHOR]

Head pose estimation (HPE) represents a topic central to many relevant research fields and characterized by a wide application range. In particular, HPE performed using a singular RGB frame is particular suitable to be applied at best-frame-selection problems. This explains a growing interest witnessed by a large number of contributions, most of which exploit deep learning architectures and require extensive training sessions to achieve accuracy and robustness in estimating head rotations on three axes. However, methods alternative to machine learning approaches could be capable of similar if not better performance. To this regard, we present FASHE, an approach based on partitioned iterated function systems (PIFS) to represent auto-similarities within face image through a contractive affine function transforming the domain blocks extracted only once by a single frontal reference image, in a good approximation of the range blocks which the target image has been partitioned into. Pose estimation is achieved by finding the closest match between fractal code of target image and a reference array by means of Hamming distance. The results of experiments conducted exceed the state of the art on both Biwi and Ponting’04 datasets as well as approaching those of the best performing methods on the challenging AFLW2000 database. In addition, the applications to GOTCHA Video Dataset demonstrate that FASHE successfully operates in-the-wild. [ABSTRACT FROM AUTHOR]

In this paper, an improved deep learning-based end-to-end autoencoder is proposed for unmanned aerial vehicle (UAV) to ground free space optical communication to mitigate atmospheric turbulence. Deep neural network (DNN) is applied to the intensity modulation/direct detection (IM/DD) autoencoder, including transmitter, receiver as well as channel model. The performance is improved by two-stage deep learning training because the minimum Hamming distance between the codewords is increased through pre-training. Simulation results show that the bit error rate of our proposed scheme can reach the 7% hard-decision forward error correction (HD-FEC) threshold at signal-to-noise ratio of approximately 22 dB and in strong atmospheric turbulence where the maximum Rytov variance is 3.5. Our proposed scheme can outperform the state-of-the-art IM/DD system with PPM transmitter and maximum likelihood receiver by achieving approximately 12 dB improvement and reducing ∼51.3% of the decoder's running time without the need for accurate channel state information. • A data-driven deep learning-based autoencoder is proposed for UAV-to-Ground free space optical communication which can mitigate atmospheric turbulence. • The proposed scheme can reach the HD-FEC threshold at 22 dB with maximum Rytov variance 3.5. • The proposed scheme shows approximately 12 dB improvement compared with PPM transmitter and ∼51.3% reduction in decoder's running time compared with ML receiver without the need for perfect channel state information. [ABSTRACT FROM AUTHOR]

• We achieve effective deep hash retrieval by mapping and reversely mapping feature in multiple-GANs framework to simultaneously reduce storage cost truly and to obtain satisfactory user acceptance on the basis of acceptable retrieval precision. • We propose supervised manifold similarity to obtain better retrieval performance including retrieval precision and user acceptance followed by detailed demonstration. • We prove that Poisson distribution induced by tremendous hash codes can be initialized as generative distribution to fit real distribution. As an extension, any additive distribution can be utilized to initialize generative distribution to fit real distribution. • Experiments show that BH2I-GAN yields competitive retrieval performance comparing with state-of-the-art hashing methods, and obtains significant storage reduction as well as high-quality reconstruction from hash code. Besides, all retrieved images locate in the neighborhood of queries, which makes satisfactory user acceptance. Given the benefits of high retrieval efficiency and low storage cost, hashing method has received an increasing attention. In particular, deep learning-based hashing has been widely used in data mining and information retrieval. However, almost all the existing methods only achieve the goal of high retrieval precision, and limit the evaluation of hashing methods to objective aspect. In this paper, we propose a novel bidirectional hash_code-to-image translation model by using multi-generative multi-adversarial nets to reduce storage cost truly and obtain satisfactory user acceptance on the basis of high retrieval precision. Firstly, we propose supervised manifold metric to reduce Hamming distance between similar instances while increasing the Hamming distance between dissimilar instances, which have been proved to be helpful for high retrieval precision and good user acceptance. Then, we utilize multi-generative and multi-adversarial networks to construct hash mapping and inverse hash generation. During inverse generation, theoretical analysis is conducted to show that inverse hash network can avoid unstable training and mode collapse. Besides, we prove that Poisson distribution induced by hash codes can be initialized as generative distribution to fit real distribution. Experimental results show that our method outperforms several state-of-the-art approaches on three popular datasets. Specifically, ours yields average about 9.3% increment in Mean Average Precision(MAP) on three datasets, and achieves over 90% user satisfaction. Besides, it successfully reduces storage cost by 1,634 times in COCO 2017 large-scale dataset. [ABSTRACT FROM AUTHOR]

In many spatial trajectory-based applications, it is necessary to map raw trajectory data points onto road networks in digital maps, which is commonly referred to as a map-matching process. While most previous map-matching methods have focused on using rule-based algorithms to deal with the map-matching problems, in this paper, we consider the map-matching task from the data-driven perspective, proposing a deep learning-based map-matching model. We build a Transformer-based map-matching model with a transfer learning approach. We generate trajectory data to pre-train the Transformer model and then fine-tune the model with a limited number of labeled data to minimize the model development cost and reduce the real-to-virtual gaps. Three metrics (Average Hamming Distance, F-score, and BLEU) at two levels (point and segment level) are used to evaluate the model performance. The model is tested with real-world datasets, and the results show that the proposed map-matching model outperforms other existing map-matching models. We also analyze the matching mechanisms of the Transformer in the map-matching process, which helps to interpret the input data's internal correlation and the external relation between input data and matching results. In addition, the proposed model shows the possibility of using generated trajectories to solve the map-matching problems in the limited labeled data environment. [Display omitted] • Design a transfer learning approach to solve labeled data sparsity problems. • Develop a Transformer-based map-matching model with high performance. • Evaluate the model performance using three metrics at two levels. • Analyze the results to improve the model's explainability and interpretability. [ABSTRACT FROM AUTHOR]