RNA secondary structural alignment serves as a foundational procedure in identifying conserved structural motifs among RNA sequences, crucially advancing our understanding of novel RNAs via comparative genomic analysis. While various computational strategies for RNA structural alignment exist, they often come with high computational complexity. Specifically, when addressing a set of RNAs with unknown structures, the task of simultaneously predicting their consensus secondary structure and determining the optimal sequence alignment requires an overwhelming computational effort of $$O(L^6)$$ for each RNA pair. Such an extremely high computational complexity makes these methods impractical for large-scale analysis despite their accurate alignment capabilities. In this paper, we introduce REDalign, an innovative approach based on deep learning for RNA secondary structural alignment. By utilizing a residual encoder-decoder network, REDalign can efficiently capture consensus structures and optimize structural alignments. In this learning model, the encoder network leverages a hierarchical pyramid to assimilate high-level structural features. Concurrently, the decoder network, enhanced with residual skip connections, integrates multi-level encoded features to learn detailed feature hierarchies with fewer parameter sets. REDalign significantly reduces computational complexity compared to Sankoff-style algorithms and effectively handles non-nested structures, including pseudoknots, which are challenging for traditional alignment methods. Extensive evaluations demonstrate that REDalign provides superior accuracy and substantial computational efficiency. REDalign presents a significant advancement in RNA secondary structural alignment, balancing high alignment accuracy with lower computational demands. Its ability to handle complex RNA structures, including pseudoknots, makes it an effective tool for large-scale RNA analysis, with potential implications for accelerating discoveries in RNA research and comparative genomics.

BACKGROUND: As the RNA secondary structure is highly related to its stability and functions, the structure prediction is of great value to biological research. The traditional computational prediction for RNA secondary prediction is mainly based on the thermodynamic model with dynamic programming to find the optimal structure. However, the prediction performance based on the traditional approach is unsatisfactory for further research. Besides, the computational complexity of the structure prediction using dynamic programming is [Formula: see text]; it becomes [Formula: see text] for RNA structure with pseudoknots, which is computationally impractical for large-scale analysis. RESULTS: In this paper, we propose REDfold, a novel deep learning-based method for RNA secondary prediction. REDfold utilizes an encoder-decoder network based on CNN to learn the short and long range dependencies among the RNA sequence, and the network is further integrated with symmetric skip connections to efficiently propagate activation information across layers. Moreover, the network output is post-processed with constrained optimization to yield favorable predictions even for RNAs with pseudoknots. Experimental results based on the ncRNA database demonstrate that REDfold achieves better performance in terms of efficiency and accuracy, outperforming the contemporary state-of-the-art methods.

Deep convolutional neural networks are good at accuracy while bad at efficiency. To improve the inference speed, two directions have been explored in the past, lightweight model designing and network weight pruning. Lightweight models have been proposed to improve the speed with good enough accuracy. It is, however, not trivial if we can further speed up these "compact" models by weight pruning. In this paper, we present a technique to gradually prune the depthwise separable convolution networks, such as MobileNet, for improving the speed of this kind of "dense" network. When pruning depthwise separable convolutions, we need to consider more structural constraints to ensure the speedup of inference. Instead of pruning the model with the desired ratio in one stage, the proposed multi-stage gradual pruning approach can stably prune the filters with a finer pruning ratio. Our method achieves satisfiable speedup with little accuracy drop for MobileNets. Code is available at https://github.com/ivclab/Multistage_Pruning.

Simultaneously running multiple modules is a key requirement for a smart multimedia system for facial applications including face recognition, facial expression understanding, and gender identification. To effectively integrate them, a continual learning approach to learn new tasks without forgetting is introduced. Unlike previous methods growing monotonically in size, our approach maintains the compactness in continual learning. The proposed packing-and-expanding method is effective and easy to implement, which can iteratively shrink and enlarge the model to integrate new functions. Our integrated multitask model can achieve similar accuracy with only 39.9% of the original size.

It is hard to detect on-road objects under various lighting conditions. To improve the quality of the classifier, three techniques are used. We define subclasses to separate daytime and nighttime samples. Then we skip similar samples in the training set to prevent overfitting. With the help of the outside training samples, the detection accuracy is also improved. To detect objects in an edge device, Nvidia Jetson TX2 platform, we exert the lightweight model ResNet-18 FPN as the backbone feature extractor. The FPN (Feature Pyramid Network) generates good features for detecting objects over various scales. With Cascade R-CNN technique, the bounding boxes are iteratively refined for better results.

Continual lifelong learning is essential to many applications. In this paper, we propose a simple but effective approach to continual deep learning. Our approach leverages the principles of deep model compression, critical weights selection, and progressive networks expansion. By enforcing their integration in an iterative manner, we introduce an incremental learning method that is scalable to the number of sequential tasks in a continual learning process. Our approach is easy to implement and owns several favorable characteristics. First, it can avoid forgetting (i.e., learn new tasks while remembering all previous tasks). Second, it allows model expansion but can maintain the model compactness when handling sequential tasks. Besides, through our compaction and selection/expansion mechanism, we show that the knowledge accumulated through learning previous tasks is helpful to build a better model for the new tasks compared to training the models independently with tasks. Experimental results show that our approach can incrementally learn a deep model tackling multiple tasks without forgetting, while the model compactness is maintained with the performance more satisfiable than individual task training.

We propose a novel method to merge convolutional neural-nets for the inference stage. Given two well-trained networks that may have different architectures that handle different tasks, our method aligns the layers of the original networks and merges them into a unified model by sharing the representative codes of weights. The shared weights are further re-trained to fine-tune the performance of the merged model. The proposed method effectively produces a compact model that may run original tasks simultaneously on resource-limited devices. As it preserves the general architectures and leverages the co-used weights of well-trained networks, a substantial training overhead can be reduced to shorten the system development time. Experimental results demonstrate a satisfactory performance and validate the effectiveness of the method.

Pedestrian detection is an important issue in the field of intelligent transportation systems. As a pedestrian is not an apparent object at nighttime, it brings about critical difficulties in effectively detecting a pedestrian for a driving assistant vision system. While using an infrared projector to enhance the illumination contrast, objects in a nighttime environment might reflect the infrared projected by the emitted spotlight. In some cases, however, the clothes on a pedestrian might absorb most of the infrared, thus causing the pedestrian to be partially invisible. To deal with this problem, a nighttime part-based pedestrian detection method is proposed. It divides a pedestrian into parts for a moving vehicle with a camera and a near-infrared lighting projector. Due to a high computation load, selecting effective parts becomes imperative. By analyzing the spatial relationship between every pair of parts, the confidence of the detected parts can be enhanced even when some parts are occluded. At the last stage of this system, the pedestrian detection result is refined by a block-based segmentation method. The system is verified by experiments, and the appealing results are demonstrated.

To robustly detect people in a video sequence is hard due to various challenges. One of the most successful discriminative features for finding people goes to the Histograms of Oriented Gradients (HOG). Although the major contour information is encoded in the HOG feature well, the background clutter disturbs the gradient information. Thus, an extension of HOG, called histograms of oriented gradient of granules (HOGG), is proposed. Instead of collecting gradient information at each pixel, the histograms of gradients in small regions are computed. HOGG with different granularity can describe the contour while ignoring the noisy edges. Moreover, the clutter background problem can be solved by encoding extra region information. With the help of the integral image technique, the evaluation of HOGG can be efficient. The final HOG+HOGG classifier obtains 92% detection rate at 10-4false positive per window in the experiments.