Sparse Encoding Research Articles

GluA1-containing AMPA receptors are necessary for sparse memory engram formation.

Memory formation depends on the selective recruitment of neuronal ensembles into circuits known as engrams, which represent the physical substrate of memory. Sparse encoding of these ensembles is essential for memory specificity and efficiency. AMPA receptor (AMPAR) subunits, particularly GluA1, play a central role in synaptic plasticity, which underpins memory encoding. This study investigates how GluA1 expression influences the recruitment of neurons into memory engrams. Using global GluA1 knockout (GluA1KO) mice, localized knockout models, and contextual fear-conditioning paradigms, we evaluated the role of GluA1 in memory formation and engram sparsity. GluA1KO mice exhibited impaired short-term memory retention but preserved 24-hour contextual memory. Despite this, these mice displayed increased expression of the immediate early gene Arc in hippocampal neurons, indicative of a broader engram network. Electrophysiological analyses revealed reduced synaptic strength in GluA1-deficient neurons, regardless of their Arc expression. Localized GluA1 knockout confirmed that GluA1 deficiency increases neuronal recruitment into engrams, disrupting the sparse encoding typically observed in wild-type mice. These findings demonstrate that GluA1-containing AMPARs constrain engram size, ensuring selective recruitment of neurons for efficient memory encoding. By regulating synaptic plasticity, GluA1 facilitates both the encoding and size of memory circuits. This study highlights the critical role of GluA1 in maintaining sparse engram formation and provides insight into mechanisms underlying memory deficits in conditions where synaptic composition is altered.

Emotion Recognition in a Closed-Cabin Environment: An Exploratory Study Using Millimeter-Wave Radar and Respiration Signals

In the field of psychology and cognition within closed cabins, noncontact vital sign detection holds significant potential as it can enhance the user’s experience by utilizing objective measurements to assess emotions, making the process more sustainable and easier to deploy. To evaluate the capability of noncontact methods for emotion recognition in closed spaces, such as submarines, this study proposes an emotion recognition method that employs a millimeter-wave radar to capture respiration signals and uses a machine-learning framework for emotion classification. Respiration signals were collected while the participants watched videos designed to elicit different emotions. An automatic sparse encoder was used to extract features from respiration signals, and two support vector machines were employed for emotion classification. The proposed method was experimentally validated using the FaceReader software, which is based on audiovisual signals, and achieved an emotion classification accuracy of 68.21%, indicating the feasibility and effectiveness of using respiration signals to recognize and assess the emotional states of individuals in closed cabins.

Control of innate olfactory valence by segregated cortical amygdala circuits.

Animals exhibit innate behaviors that are stereotyped responses to specific evolutionarily relevant stimuli in the absence of prior learning or experience. These behaviors can be reduced to an axis of valence, whereby specific odors evoke approach or avoidance responses. The posterolateral cortical amygdala (plCoA) mediates innate attraction and aversion to odor. However, little is known about how this brain area gives rise to behaviors of opposing motivational valence. Here, we sought to define the circuit features of plCoA that give rise to innate attraction and aversion to odor. We characterized the physiology, gene expression, and projections of this structure, identifying a divergent, topographic organization that selectively controls innate attraction and avoidance to odor. First, we examined odor-evoked responses in these areas and found sparse encoding of odor identity, but not valence. We next considered a topographic organization and found that optogenetic stimulation of the anterior and posterior domains of plCoA elicits attraction and avoidance, respectively, suggesting a functional axis for valence. Using single cell and spatial RNA sequencing, we identified the molecular cell types in plCoA, revealing an anteroposterior gradient in cell types, whereby anterior glutamatergic neurons preferentially express VGluT2 and posterior neurons express VGluT1. Activation of these respective cell types recapitulates appetitive and aversive behaviors, and chemogenetic inhibition reveals partial necessity for responses to innate appetitive or aversive odors. Finally, we identified topographically organized circuits defined by projections, whereby anterior neurons preferentially project to medial amygdala, and posterior neurons preferentially project to nucleus accumbens, which are respectively sufficient and necessary for innate attraction and aversion. Together, these data advance our understanding of how the olfactory system generates stereotypic, hardwired attraction and avoidance, and supports a model whereby distinct, topographically distributed plCoA populations direct innate olfactory responses by signaling to divergent valence-specific targets, linking upstream olfactory identity to downstream valence behaviors, through a population code. This suggests a novel amygdala circuit motif in which valence encoding is represented not by the firing properties of individual neurons, but by population level identity encoding that is routed through divergent targets to mediate distinct behaviors of opposing appetitive and aversive responses.

SIBS: A sparse encoder utilizing self-inspired bases for efficient image representation

Addressing the limitations of pre-defined dictionaries in image processing, this study introduces Self-Inspired Bases-based Sparse Encoder (SIBS), a novel approach that dynamically generates image-specific atoms by leveraging existing dictionaries and input data. Extensive numerical analysis and simulations demonstrate that SIBS enhances sparse coding performance in three key areas: improved quality resolution via precise image reconstruction, accelerated convergence speed, and competitive compression performance rivaling techniques like JPEG and JP2000. Experiments show that SIBS enhances quality levels and NMSE at the same number of non-zero coefficients and outperforms atoms learned directly from the image at low error levels. SIBS also effectively improves the performance of various learned and structured dictionaries by serving as auxiliary bases, enhancing quality resolution, convergence rate, and dictionary efficiency. Additionally, SIBS has potential as a competitive image compression method. The study’s discussion on SIBS limitations guides future research, including examining its performance with other greedy pursuit algorithms and reevaluating the encoder with other lossless and filter techniques.

Application of optimized sparse encoding algorithm in data compression

Data transmission is crucial in the process of equipment monitoring. The compression algorithms are adopted to reduce the amount of data transmission. When sparse encoding algorithms are used for this purpose, despite extensive attempts to improve their performance, some problems still remain. The main issues pertain to: (1) Dictionary construction − As practical dictionary better matches the input signal, such a dictionary must be efficiently established; 2) The setting of sparsity value − It is difficult to determine the size of sparsity values when prior knowledge is insufficient (when the parameter settings are too large, the cost of data transmission increases considerably, while excessively small values can result in poor reconstruction accuracy); and (3) How to implement engineering applications of data compression based on sparse encoding. To address these issues, in this work, an optimized sparse encoding algorithm is proposed by combining the non-negative Online Dictionary Learning (ODL) with the optimized Orthogonal Matching Pursuit (OMP) algorithm. First, a non-negative ODL algorithm is adopted for dictionary learning based on the training dataset to obtain an effective dictionary. Next, the optimized OMP algorithm is used to obtain the sparsity and the sparse coefficient matrix. Further analyses confirm that the reconstructed signal has a small reconstruction error. Finally, A lossy compression framework is proposed for the compressing equipment condition monitoring data using sparse encoding algorithms. Compared with the compression algorithms based on compressive sensing and DCT, the average compression ratio can reach 42.7 when the reconstruction accuracy is similar. In terms of comprehensive compression performance, the quality score is also higher compared to other algorithms.

Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification

Occluded person re-identification (ReID) is a challenging problem due to contamination from occluders. Existing approaches address the issue with prior knowledge cues, such as human body key points and semantic segmentations, which easily fail in the presence of heavy occlusion and other humans as occluders. In this paper, we propose a feature pruning and consolidation (FPC) framework to circumvent explicit human structure parsing. The framework mainly consists of a sparse encoder, a multi-view feature mathcing module, and a feature consolidation decoder. Specifically, the sparse encoder drops less important image tokens, mostly related to background noise and occluders, solely based on correlation within the class token attention. Subsequently, the matching stage relies on the preserved tokens produced by the sparse encoder to identify k-nearest neighbors in the gallery by measuring the image and patch-level combined similarity. Finally, we use the feature consolidation module to compensate pruned features using identified neighbors for recovering essential information while disregarding disturbance from noise and occlusion. Experimental results demonstrate the effectiveness of our proposed framework on occluded, partial, and holistic Re-ID datasets. In particular, our method outperforms state-of-the-art results by at least 8.6% mAP and 6.0% Rank-1 accuracy on the challenging Occluded-Duke dataset.

Semi-Symmetrical, Fully Convolutional Masked Autoencoder for TBM Muck Image Segmentation

Deep neural networks are effectively utilized for the instance segmentation of muck images from tunnel boring machines (TBMs), providing real-time insights into the surrounding rock condition. However, the high cost of obtaining quality labeled data limits the widespread application of this method. Addressing this challenge, this study presents a semi-symmetrical, fully convolutional masked autoencoder designed for self-supervised pre-training on extensive unlabeled muck image datasets. The model features a four-tier sparse encoder for down-sampling and a two-tier sparse decoder for up-sampling, connected via a conventional convolutional neck, forming a semi-symmetrical structure. This design enhances the model’s ability to capture essential low-level features, including geometric shapes and object boundaries. Additionally, to circumvent the trivial solutions in pixel regression that the original masked autoencoder faced, Histogram of Oriented Gradients (HOG) descriptors and Laplacian features have been integrated as novel self-supervision targets. Testing shows that the proposed model can effectively discern essential features of muck images in self-supervised training. When applied to subsequent end-to-end training tasks, it enhances the model’s performance, increasing the prediction accuracy of Intersection over Union (IoU) for muck boundaries and regions by 5.9% and 2.4%, respectively, outperforming the enhancements made by the original masked autoencoder.

A study of self-reflection and intelligent adjustment mechanism in artificial intelligence-assisted vocal training

Abstract In the current domestic vocal music industry, there are many misunderstandings in many people’s singing, such as singing loudly, overpowering heavy skills, and non-uniformity of vocal areas. Based on the neural network structure model, this paper proposes a relevant method for extracting music features from vocal signals. After downsampling, normalization, frame splitting, and other processing, the time-frequency transformation of the audio signal is achieved, so as to improve the analysis of the information it carries. The audio segmentation algorithm DIS based on metric distance is used to segment the notes, while the spectral variance method is used to determine the turbid and non-turbid segments in the sound clips. Using the amplitude compression-based fundamental estimation filter. The fundamental frequency is extracted to complete the tracking of the dominant fundamental frequency trajectory of vocal features. On the basis of the neural network model, a sparse encoder and softmax classifier are added to build the training fundamental frequency discrimination model. In an empirical investigation of vocal training, the resonance peaks of the vocal singing clip of “Song of the Mother Deer”, the 3rd resonance peak appeared in the interval of 0s-0.52s, 3.81s-4s, 7.15s-9s, 10.85s-11s, and 14.87s-16.24s, with the change amplitude ranging from 2574Hz to 4436Hz, and there is a certain rule of change of the other resonance peaks. The performance of the students trained with the vocal training model was 78.61, and the performance of the students trained with the traditional vocal training was 75.34, and the vocal training model has a good practical effect.

Camera-Lidar fusion algorithm formed by fast sparse encoder and transformer decoder

Camera-Lidar fusion algorithm formed by fast sparse encoder and transformer decoder

A mechanistic account of visual discomfort.

Much of the neural machinery of the early visual cortex, from the extraction of local orientations to contextual modulations through lateral interactions, is thought to have developed to provide a sparse encoding of contour in natural scenes, allowing the brain to process efficiently most of the visual scenes we are exposed to. Certain visual stimuli, however, cause visual stress, a set of adverse effects ranging from simple discomfort to migraine attacks, and epileptic seizures in the extreme, all phenomena linked with an excessive metabolic demand. The theory of efficient coding suggests a link between excessive metabolic demand and images that deviate from natural statistics. Yet, the mechanisms linking energy demand and image spatial content in discomfort remain elusive. Here, we used theories of visual coding that link image spatial structure and brain activation to characterize the response to images observers reported as uncomfortable in a biologically based neurodynamic model of the early visual cortex that included excitatory and inhibitory layers to implement contextual influences. We found three clear markers of aversive images: a larger overall activation in the model, a less sparse response, and a more unbalanced distribution of activity across spatial orientations. When the ratio of excitation over inhibition was increased in the model, a phenomenon hypothesised to underlie interindividual differences in susceptibility to visual discomfort, the three markers of discomfort progressively shifted toward values typical of the response to uncomfortable stimuli. Overall, these findings propose a unifying mechanistic explanation for why there are differences between images and between observers, suggesting how visual input and idiosyncratic hyperexcitability give rise to abnormal brain responses that result in visual stress.

Improved Smart Healthcare System of Cloud-Based IoT Framework for the Prediction of Heart Disease

Smart healthcare systems in the cloud-based IoT framework for the prediction of heart disease improve the patient's health status and minimizes the death rate. The prediction of heart disease is a challenging one. Early prediction of heart disease may reduce the risk of patient illness and monitoring in real-time to avoid the risk. The view of existing algorithms is inaccurate in early prediction which took a lot of time for the prediction and inaccurate early prediction of heart disease. To overcome these issues, this paper proposed a sparse autoencoder with Galactic Swarm Optimization (SAE-GSO) algorithm. A sparse encoder predicts heart disease and enhances the accurate prediction, tuning the parameters of sparsity regularity in the sparse autoencoder, Galactic Swarm optimization algorithm is implemented. The proposed work enhances the prediction rate of heart diseases, minimizing the error rate, and maximizing the accuracy. The accuracy rate of the proposed work of SAE-GSO in the Cleveland Dataset produces got 92.23 %, GBT got 65.12 %, SAE got 87.34%, and NB got 83.16 %. The accuracy rate of the proposed work of SAE-GSO in the Framingham Dataset produced 92.59 %, GBT got 69.16 %, SAE got 86.25%, and NB got 82.37%.

English

Huge diagrams have unique properties for organizations and research, such as client linkages in informal organizations and customer evaluation lattices in social channels. They necessitate a lot of financial assets to maintain because they are large and frequently continue to expand. Owners of large diagrams may need to use cloud resources due to the extensive arrangement of open cloud resources to increase capacity and computation flexibility. However, the cloud's accountability and protection of schematics have become a significant issue. In this study, we consider calculations for security savings for essential graph examination practices: schematic extraterrestrial examination for outsourcing graphs in the cloud server. We create the security-protecting variants of the two proposed Eigen decay computations. They are using two cryptographic algorithms: additional substance homomorphic encryption (ASHE) strategies and some degree homomorphic encryption (SDHE) methods. Inadequate networks also feature a distinctively confidential info adaptation convention to allow the trade-off between secrecy and data sparseness. Both dense and sparse structures are investigated. According to test results, calculations with sparse encoding can drastically reduce information. SDHE-based strategies have reduced computing time, while ASHE-based methods have reduced stockpiling expenses.

Token Sparsification for Faster Medical Image Segmentation.

Can we use sparse tokens for dense prediction, e.g., segmentation? Although token sparsification has been applied to Vision Transformers (ViT) to accelerate classification, it is still unknown how to perform segmentation from sparse tokens. To this end, we reformulate segmentation as a sparse encoding → token completion → dense decoding (SCD) pipeline. We first empirically show that naïvely applying existing approaches from classification token pruning and masked image modeling (MIM) leads to failure and inefficient training caused by inappropriate sampling algorithms and the low quality of the restored dense features. In this paper, we propose Soft-topK Token Pruning (STP) and Multi-layer Token Assembly (MTA) to address these problems. In sparse encoding, STP predicts token importance scores with a lightweight sub-network and samples the topK tokens. The intractable topK gradients are approximated through a continuous perturbed score distribution. In token completion, MTA restores a full token sequence by assembling both sparse output tokens and pruned multi-layer intermediate ones. The last dense decoding stage is compatible with existing segmentation decoders, e.g., UNETR. Experiments show SCD pipelines equipped with STP and MTA are much faster than baselines without token pruning in both training (up to 120% higher throughput) and inference (up to 60.6% higher throughput) while maintaining segmentation quality. Code is available here: https://github.com/cvlab-stonybrook/TokenSparse-for-MedSeg.

Visual Sentiment Classification via Low-Rank Regularization and Label Relaxation

In the human cognitive system, the emotional feeling is a complicated process. Visual sentiment classification aims to predict the human emotions evoked by different images. In this article, we proposed a novel visual sentiment classification algorithm by modeling this task as a low-rank subspace learning problem. To reduce the discrepancy between global and local features, image features of relevant regions are selected from the whole image by sparse encoding. The label relaxation item is employed for alleviating the label ambiguity caused by subjective evaluation. We develop an alternative iterative method to optimize the proposed objective function. This model can be naturally extended for online learning, which improves efficiency. We conduct extensive experiments on three publicly available data sets. Compared with several state-of-the-art methods, we achieve better performance.

Computational Intelligence Based Electronic Healthcare Data Analytics Using Feature Selection with Classification by Deep Learning Architecture

EHRs (Electronic health records) are a source of big data that offer a wealth of clinical patient health data. However, because these notes are free-form texts, writing formats and styles range greatly amongst various records, text data from eHRs, such as discharge rapid notes, provide analysis challenges. This research proposed novel technique in electronic healthcare data analysis based on feature selection and classification utilizingDL methods. here the input is collected as input EH data, is processed for dimensionality reduction, noise removal. A public data pre-processing method for dealing with HD-EHR data is dimensionality reduction, which tries to minimize amount of EHR representational features while enhancing effectiveness of following data analysis, such as classification. The processed data features has been selected utilizingweighted curvature based feature selection with support vector machine. Then this selected deep features has been classified using sparse encoder transfer learning. the experimental analysis has been carried out for various EH datasets in terms of accuracy of 96%, precision of 92%, recall of 77%, F-1 score of 72%, MAP of 65%

Reasoning discriminative dictionary-embedded network for fully automatic vertebrae tumor diagnosis.

Fully automatic vertebrae tumor diagnosis (FAVTD) means using an end-to-end network to directly perform vertebrae recognition and tumor diagnosis from MRI images. FAVTD is clinically crucial for tumor screening and treatment, which helps prevent further metastasis and save the patients' lives. However, FAVTD has not yet been fully attempted due to the challenges raised by tumor appearance variability as well as MRI image field of view (FOV) and/or characteristics diversity. We propose a REasoning DiscriminativE diCtIonary-embeDded nEtwork (RE-DECIDE) to tackle the challenges in FAVTD. RE-DECIDE contains an elaborated enhanced-supervision recognition network (ERN) and a self-adaptive reasoning diagnosis network (SRDN). ERN is implemented in a feed-forward dictionary learning manner, which encodes each vertebra by the sparse codes and uses the sparse projections of the vertebrae coordinates onto multiple observation axes for supervision. ERN thus provides multiple sparse encodings of all vertebrae (and their ground truths) to enhance supervision, which reinforces the discrimination of different vertebrae and thus improves recognition performance. SRDN first highlights the most informative feature in the recognized vertebrae based on an attention mechanism. It then performs feature interaction, i.e., exchanges features of different vertebrae based on the graph reasoning mechanism. A reasoning controlling strategy is designed to prompt feature interaction in vertebrae with the same diagnosis labels and meanwhile reduces that in vertebrae with different labels, which avoids over-smoothing and improves diagnosis performance. RE-DECIDE is trained and evaluated using a challenging dataset consisting of 600 MRI images; the evaluation results show that RE-DECIDE achieves high performance in both recognition (accuracy: 0.940) and diagnosis (AUC: 0.947) tasks.

AirCode: A Robust Object Encoding Method

Object encoding and identification are crucial for many robotic tasks such as autonomous exploration and semantic relocalization. Existing works heavily rely on the tracking of detected objects but have difficulty recalling revisited objects precisely. In this paper, we propose a novel object encoding method, which is named as AirCode, based on a graph of key-points. To be robust to the number of key-points detected, we propose a feature sparse encoding and object dense encoding method to ensure that each key-point can only affect a small part of the object descriptors, leading it to be robust to viewpoint changes, scaling, occlusion, and even object deformation. In the experiments, we show that it achieves superior performance for object identification than the state-of-the-art algorithms and is able to provide reliable semantic relocalization. It is a plug-and-play module and we expect that it will play an important role in various applications.

The number of partial Steiner systems and d-partitions

We prove asymptotic upper bounds on the number of $d$-partitions (paving matroids of fixed rank) and partial Steiner systems (sparse paving matroids of fixed rank), using a mixture of entropy counting, sparse encoding, and the probabilistic method.

Sketched Stochastic Dictionary Learning for large‐scale data and application to high‐throughput mass spectrometry

AbstractFactorization of large data corpora has emerged as an essential technique to extract dictionaries (sets of patterns that are meaningful for sparse encoding). Following this line, we present a novel algorithm based on compressive learning theory. In this framework, the (arbitrarily large) dataset of interest is replaced by a fixed‐size sketch resulting from a random sampling of the data distribution characteristic function. We apply our algorithm to the extraction of chromatographic elution profiles in mass spectrometry data, where it demonstrates its efficiency and interest compared to other related algorithms.

Energy efficient Q learning based Kullback sparse encoder for traffic and congestion control data delivery in WSN

In dense traffic Wireless Sensor Network (WSN), traffic congestion can increase routing overhead and packet loss, which restricts the entire network performance, therefore a traffic-aware and congestion-control data delivery is required to control the traffic. The proposed Energy-efficient Q-learning and Kullback Sparse Encoder (EQ-KSE) method is used for traffic-aware routing and congestion-control data delivery method for WSN. The method enforces a traffic balancing strategy using the energy-efficient reward function and estimates the wireless link quality by the Energy-efficient and Q-Learning Routing algorithm. On the basis of the estimation of each wireless link, the Energy-efficient Q-Learning based Traffic-aware Routing model makes routing decisions through energy and queue length to reduce routing overhead and time significantly. With the obtained optimal routes data aggregation are performed at the sink node, causing a proportionate amount of congestion. To handle this issue, a Kullback Leibler Sparse Auto Encoder (KL-SAE) Congestion-control Data Delivery method is proposed. This KL-SAE model with the aid of the reconstruction loss function, through divergence reduces the congestion, therefore contributing to packet loss, and packet delivery ratio. Simulation results show that EQ-KSE method performs traffic-aware routing by minimizing both the route selection time and overhead. In high node density scenarios, it also betters the state-of-the-art methods in packet delivery ratio and packet loss rate.