Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “NODES”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 235 records · Page 13

A Multi-Purpose Modular Electronics Integration Node for Exploration Extravehicular Activity

As NASA works to develop an effective integrated portable life support system design for exploration Extravehicular activity (EVA), alternatives to the current system s electrical power and control architecture are needed to support new requirements for flexibility, maintainability, reliability, and reduced mass and volume. Experience with the current Extravehicular Mobility Unit (EMU) has demonstrated that the current architecture, based in a central power supply, monitoring and control unit, with dedicated analog wiring harness connections to active components in the system has a significant impact on system packaging and seriously constrains design flexibility in adapting to component obsolescence and changing system needs over time. An alternative architecture based in the use of a digital data bus offers possible wiring harness and system power savings, but risks significant penalties in component complexity and cost. A hybrid architecture that relies on a set of electronic and power interface nodes serving functional models within the Portable Life Support System (PLSS) is proposed to minimize both packaging and component level penalties. A common interface node hardware design can further reduce penalties by reducing the nonrecurring development costs, making miniaturization more practical, maximizing opportunities for maturation and reliability growth, providing enhanced fault tolerance, and providing stable design interfaces for system components and a central control. Adaptation to varying specific module requirements can be achieved with modest changes in firmware code within the module. A preliminary design effort has developed a common set of hardware interface requirements and functional capabilities for such a node based on anticipated modules comprising an exploration PLSS, and a prototype node has been designed assembled, programmed, and tested. One instance of such a node has been adapted to support testing the swingbed carbon dioxide and humidity control element in NASA s advanced PLSS 2.0 test article. This paper will describe the common interface node design concept, results of the prototype development and test effort, and plans for use in NASA PLSS 2.0 integrated tests.

Hodgson, Edward↗

Estimates of SEU rates from heavy ions in devices exhibiting dual-node susceptibility

A single-event effect (SEE) occurs when a single ionizing particle passing through an electronic device produces a detectable event in the device. The type of SEE most frequently discussed in the literature is a "single-node upset" problem in which an upset at a given node depends only on conditions seen by that node. In contrast, some devices are designed so that an upset becomes a "dual-node" problem in which a device upset depends on conditions simultaneously seen by two nodes. This report derives a method for estimating heavy-ion induced upset rates for the dual-node problem.

Edmonds, Larry D.↗

Single and Multi-Node Modeling of Direct, Submerged, and Self-Pressurization of A Cryogenic Propellant Tank Using Nodal Tools

The pressurization of cryogenic propellant tanks will always be an important process so long as cryogenic liquids are being considered as fuel sources or used for other in-space applications. Pressure control of the tank ullage is necessary for managing propellant flowrates to an engine or a receiver tank, and modeling of the process is used to predict the pressurant requirements and the amount of propellant boiloff. Direct ullage pressurization is the more traditional approach to tank pressurization, as the physics are straight-forward, and ample test (flight) data have been collected and analyzed over the past several decades. Submerged injection pressurization is an alternate method for tank pressurization and has been shown to reduce pressurant requirements, subcool the propellant, and reduce the risk of ullage collapse. Additionally, the pressurant gas entering the ullage is usually much colder when using the submerged pressurization approach, resulting in reduced propellant boiloff. These benefits are at the expense of vaporizing a small percentage of the propellent. Both tank pressurization methods are viable options for current and future space missions, and it is important to have the capability of analyzing the tank ullage conditions for both approaches. Our previous work has demonstrated the development of a Generalized Fluid System Simulation Program (GFSSP) model, which contains a thermodynamic equilibrium heat and mass transfer subroutine capable of effectively analyzing both direct and submerged pressurization systems [1-2]. This subroutine has most recently been enhanced to include the non-equilibrium effect of pressurant dissolution into the propellant. To date the ullage has always been represented as a single node, and although the simulated single-node temperatures have good comparison with the volume-averaged ullage temperatures computed from test data, the physics of the thermal stratification in the ullage were never captured, and adjustment factors in the model were required. The purpose of this paper is to introduce the development of a multi-node ullage model using GFSSP and to discuss the improvements of the simulated ullage temperature distribution and its resulting effects on ullage heat transfer processes. Test data from the Cryogenic Propellant Storage and Transfer Engineering Developmental Unit (CPST EDU) was used for model validation. For additional comparison, a Thermal Desktop (TD) model was also developed to analyze the CPST EDU direct ullage pressurization tests using both a single node and multi-node approach. The model includes the direct pressurant line, vent line, fill/drain line, and a TD FloCAD Compartment. The TD FloCAD Compartment is employed to represent the liquid and ullage as single volumes inside the tank, to include a liquid/vapor interface, and to generate network level objects such as lumps (analogous to nodes in GFSSP), paths, and ties between the fluid and thermal elements. An established heat load on the model tank was leveraged from a pre-existing higher-fidelity model correlated to CPST EDU test data.

pressurization↗

Determining when a set of compute nodes participating in a barrier operation on a parallel computer are ready to exit the barrier operation

Methods, apparatus, and products are disclosed for determining when a set of compute nodes participating in a barrier operation on a parallel computer are ready to exit the barrier operation that includes, for each compute node in the set: initializing a barrier counter with no counter underflow interrupt; configuring, upon entering the barrier operation, the barrier counter with a value in dependence upon a number of compute nodes in the set; broadcasting, by a DMA engine on the compute node to each of the other compute nodes upon entering the barrier operation, a barrier control packet; receiving, by the DMA engine from each of the other compute nodes, a barrier control packet; modifying, by the DMA engine, the value for the barrier counter in dependence upon each of the received barrier control packets; exiting the barrier operation if the value for the barrier counter matches the exit value.

97 MATHEMATICS AND COMPUTING↗

Asynchronous broadcast for ordered delivery between compute nodes in a parallel computing system where packet header space is limited

Disclosed is a mechanism on receiving processors in a parallel computing system for providing order to data packets received from a broadcast call and to distinguish data packets received at nodes from several incoming asynchronous broadcast messages where header space is limited. In the present invention, processors at lower leafs of a tree do not need to obtain a broadcast message by directly accessing the data in a root processor's buffer. Instead, each subsequent intermediate node's rank id information is squeezed into the software header of packet headers. In turn, the entire broadcast message is not transferred from the root processor to each processor in a communicator but instead is replicated on several intermediate nodes which then replicated the message to nodes in lower leafs. Hence, the intermediate compute nodes become "virtual root compute nodes" for the purpose of replicating the broadcast message to lower levels of a tree.

97 MATHEMATICS AND COMPUTING↗

Packet sniffer node and system including the same to assess wireless communication performance

A wireless system includes a plurality of wireless nodes structured to participate in a system test of the wireless nodes. The system test includes a plurality of test packets. A packet sniffer node includes a wireless receiver, a memory storing a schedule defining transmission and reception of the test packets by the wireless nodes, and a processor cooperating with the wireless receiver and the memory to receive at least some of the test packets responsive to the schedule, and to output data corresponding to the received test packets of the system test. A monitoring computer receives the output data from the processor of the packet sniffer node. Operation of the packet sniffer node, which does not transmit, does interfere with or alter execution of the system test.

Armstrong, Brian S. R.↗

Distributed gather/scatter operations across a network of memory nodes

Devices, methods, and systems for distributed gather and scatter operations in a network of memory nodes. A responding memory node includes a memory; a communications interface having circuitry configured to communicate with at least one other memory node; and a controller. The controller includes circuitry configured to receive a request message from a requesting node via the communications interface. The request message indicates a gather or scatter operation, and instructs the responding node to retrieve data elements from a source memory data structure and store the data elements to a destination memory data structure. The controller further includes circuitry configured to transmit a response message to the requesting node via the communications interface. The response message indicates that the data elements have been stored into the destination memory data structure.

97 MATHEMATICS AND COMPUTING↗

Node Augmentation Technique in Bayesian Network Evidence Analysis and Marshaling

Given a Bayesian network, sensitivity analysis is an important activity. This paper begins by describing a network augmentation technique which can simplifY the analysis. Next, we present two techniques which allow the user to determination the probability distribution of a hypothesis node under conditions of uncertain evidence; i.e. the state of an evidence node or nodes is described by a user specified probability distribution. Finally, we conclude with a discussion of three criteria for ranking evidence nodes based on their influence on a hypothesis node. All of these techniques have been used in conjunction with a commercial software package. A Bayesian network based on a directed acyclic graph (DAG) G is a graphical representation of a system of random variables that satisfies the following Markov property: any node (random variable) is independent of its non-descendants given the state of all its parents (Neapolitan, 2004). For simplicities sake, we consider only discrete variables with a finite number of states, though most of the conclusions may be generalized.

97 MATHEMATICS AND COMPUTING↗

Morphological predictors for lymph node metastases on computed tomography in colon cancer

Introduction/Background: The aim of the study was to assess morphological predictors for lymph node metastases (Stage III disease) in colon cancer on computed tomography. Methods and materials: Ninety-four patients with histology-proven colon cancer (adenocarcinoma) who underwent elective primary curative resection between the years 2012 and 2014 were included. Contrast-enhanced CT examinations were independently reviewed by two blinded observers regarding tumor location, depth of tumor invasion, and presence of lymph node metastases. Ocular presence of internal heterogeneity and presence of irregular outer border were used as morphological criteria for lymph node involvement. Protocol-based histopathology after curative surgery served as reference standard. Sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and accuracy for each morphological criterion for prediction of stage III disease were calculated. Inter-observer agreement was compared using Kappa statistics. Results: According to histopathology, 59 patients were staged as I–II disease and 35 patients were staged as stage III disease. The presence of internal heterogeneity in a lymph node on CT resulted in moderate sensitivity (66–77%) but high specificity (95–95%) for prediction of Stage III disease by both observers. The presence of irregular outer border also resulted in poor sensitivity (49–54%) but high specificity (97–97%). The combination of either internal heterogeneity and/or irregular outer border per patient resulted in a moderate sensitivity (67–77%) and high specificity (95–95%), PPV (89–96%), and NPV (84–88%). Inter-observer agreement (Cohens Kappa) was 0.72. Consensus reading for the combined criteria resulted in sensitivity and specificity of 69% and 100%, respectively. Conclusion: Using morphological criteria for lymph node metastases on CT examination in patients with colon cancer results in high specificity but moderate sensitivity in predicting stage III disease.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Performance of a new system using a one-step nucleic acid amplification assay for detecting lymph node metastases in breast cancer

One-step nucleic acid amplification (OSNA) for CK19 mRNA is an intraoperative diagnostic procedure for detection of lymph node metastasis. Automated Gene Amplification Detector RD-200 and the LYNOAMP CK19 gene amplification reagent as components of a new OSNA system have been developed. As an improvement over a conventional system, the new system can analyze 14 samples per run, evaluate two lymph nodes in ~ 17 min, and reduce inhibition of reactions. This study was aimed at evaluating clinical performance of the new system by comparing it with performance of histopathological analysis and a conventional OSNA system. A total of 150 lymph nodes in 63 breast cancer patients (T1–3) who underwent sentinel lymph node biopsy or axillary lymph node dissection were examined intraoperatively with the new OSNA system, the conventional system, and histopathological analysis. In comparison with histopathological analysis, sensitivity, specificity, and concordance rate of the new system were 93.9, 98.8, and 96.7%, respectively. In comparison with the conventional system, similar corresponding values were obtained: 96.9, 98.8, and 98.0%, respectively. The results show that clinical performance of the new OSNA system is equivalent to that of histopathological diagnosis and the conventional OSNA system. The new system is superior to the conventional one because of processing of a greater number of samples, shorter testing time, and the absence of inhibited reactions.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Breast Cancer Regional Radiation Fields for Supraclavicular and Axillary Lymph Node Treatment: Is a Posterior Axillary Boost Field Technique Optimal?

Purpose: To assess whether using an anterior oblique supraclavicular (SCV) field with a posterior axillary boost (PAB) field is an optimal technique for targeting axillary (AX) lymph nodes compared with two computed tomography (CT)-based techniques: (1) an SCV field with an anterior boost field and (2) intensity-modulated radiotherapy (IMRT). Methods and Materials: Ten patients with CT simulation data treated with postmastectomy radiation that included an SCV field were selected for the study. Supraclavicular nodes and AX Level I-III nodes within the SCV field were contoured and defined as the treatment target. Plans using the three techniques were generated and evaluated for each patient. Results: The anterior axillary boost field and IMRT resulted in superior dose coverage compared with PAB. Namely, treatment volumes that received 105%, 80%, and 30% of prescribed dose for IMRT plans were significantly less than those for the anterior axillary boost plans, which were significantly less than PAB. For PAB and anterior axillary boost plans, there was a linear correlation between treatment volume receiving 105% of prescribed dose and maximum target depth. Furthermore, the IMRT technique resulted in better lung sparing and dose conformity to the target than anterior axillary boost, which again was significantly better than PAB. The maximum cord dose for IMRT was small, but higher than for the other two techniques. More monitor units were required to deliver the IMRT plan than the PAB plan, which was more than the anterior axillary boost plan. Conclusions: The PAB technique is not optimal for treatment of AX lymph nodes in an SCV field. We conclude that CT treatment planning with dose optimization around delineated target volumes should become standard for radiation treatments of supraclavicular and AX lymph nodes.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Lymph Node Ratio as a Prognostic Factor in Patients With Stage III Rectal Cancer Treated With Total Mesorectal Excision Followed by Chemoradiotherapy

Purpose: To investigate the prognostic impact of lymph node ratio (LNR) on survival in the patients with Stage III rectal cancer. Methods and Materials: We retrospectively reviewed the data of 421 consecutive patients who underwent total mesorectal excision followed by chemoradiotherapy for rectal cancer from 1996 to 2006. The 232 patients with positive lymph nodes (LNs) were divided into four groups according to LNR quartiles: LNR {<=}0.1 (n = 69), {<=}0.2 (n = 49), {<=}0.4 (n = 54), and >0.4 (n = 60). The association between LNR and survival was evaluated by the Kaplan-Meier method and multivariate analysis with covariates of prognostic significance in univariate analysis. Results: The median numbers of examined and positive LNs were 17 and 3, respectively, and the median LNR was 0.20 (range, 0.03-1). There was a strong correlation between the number of positive LNs and LNR (r = 0.724, p < 0.001). After a median follow-up of 53 months (range, 9-138 months), the actuarial overall survival and disease-free survival rates at 5 years were 69% and 56%, respectively. The 5-year survival rate decreased as LNR increased ({<=}0.1, 89%; {<=}0.2, 67%; {<=}0.4, 64%; >0.4, 50%; p < 0.001). Lymph node ratio was also a significant prognostic factor on Cox regression analysis ({<=}0.1, hazard ratio [HR] = 1; {<=}0.2, HR = 1.3, p = 0.623; {<=}0.4, HR = 2.4, p = 0.047; >0.4, HR = 3.7, p = 0.005). Lymph node ratio had a prognostic effect on overall survival in subgroups of patients with N1 (p = 0.032) and N2 (p = 0.034) tumors. Conclusion: Lymph node ratio was the most significant predictor of survival in the patients with Stage III rectal cancer who had undergone postoperative chemoradiation.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Influence of Pathological Nodal Status and Maximal Standardized Uptake Value of the Primary Tumor and Regional Lymph Nodes on Treatment Plans in Patients With Advanced Oral Cavity Squamous Cell Carcinoma

Purpose: A better understanding of the prognostic factors in oral cavity squamous cell carcinoma (OSCC) may optimize the therapeutic approach. In this study, we sought to investigate whether the combination of clinical information, pathologic results, and preoperative maximal standardized uptake value (SUVmax) at the primary tumor and regional lymph nodes might improve the prognostic stratification in this patient group. Methods and Materials: A total of 347 consecutive OSCC patients were investigated. All participants underwent fluorodeoxyglucose-positron emission tomography within 2 weeks before surgery and neck dissection. The duration of follow-up was at least 24 months in all surviving patients. The optimal cutoff values for SUVmax at the primary tumor (SUVtumor-max) and regional lymph nodes (SUVnodal-max) were selected according to the 5-year disease-free survival (DFS) rate. Independent prognosticators were identified by Cox regression analysis. Results: In multivariate analysis, a cutoff SUVtumor-max of 8.6, a cutoff SUVnodal-max of 5.7, and the presence of pathologic lymph node metastases were found to be significant prognosticators for the 5-year DFS. A scoring system using these three prognostic factors was formulated to define distinct prognostic groups. The 5-year rates for patients with a score between 0 and 3 were as follows: neck control, 94%, 86%, 77%, 59% (p < 0.0001); distant metastases, 1%, 7%, 22%, 47% (p < 0.0001); disease-specific survival, 93%, 85%, 61%, 36%, respectively (p < 0.0001). Conclusion: Based on the study findings, the combined evaluation of pathologic node status and SUVmax at the primary tumor and regional lymph nodes may improve prognostic stratification in OSCC patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Impact of Lymph Node Status on Clinical Outcomes After Accelerated Partial Breast Irradiation

Purpose: To compare outcomes after accelerated partial breast irradiation (APBI) between node-negative and node-positive patients. Methods and Materials: A total of 534 patients with early-stage breast cancer received APBI including 39 node-positive (N+) cases. Clinical, pathologic, and treatment-related factors were compared between node-negative (N-) and N+ cohorts. Local recurrence (LR), regional recurrence (RR), axillary failure (AF), distant metastases (DM), disease-free survival (DFS), cause-specific survival (CSS), and overall survival (OS) were analyzed. Results: N+ patients were younger (p = 0.04), had larger tumors (p < 0.001), and were more likely to receive chemotherapy (p < 0.001). Mean follow-up was 7.8 years for N+ patients and 6.3 years for N- patients (p = 0.06). No differences were seen in 5-year actuarial rates of LR (2.2% vs. 2.6%, p = 0.86), AF (0% vs. 0%, p = 0.69), DFS (90.0% vs. 88.0%, p = 0.79), or OS (91.0 vs. 84.0%, p = 0.65) between the two groups, whereas higher rates of RR (0% vs. 6.1%, p < 0.001) and DM (2.2% vs. 8.9%, p = 0.005) were noted in N+ patients. A trend for improved CSS (p = 0.06), was seen in N- patients. Age, tumor size, receptor status, T-stage, chemotherapy, APBI technique, and nodal status (p = 0.86) were not associated with LR, while a trend for an association with LR was noted with close/positive margins, (p = 0.07), and failure to receive adjuvant hormonal therapy (p = 0.06). Conclusions: No differences were seen in the rates of LR or AF between N- and N+ patients after APBI. These results support the continued enrollment of node-positive patients in Phase III trials evaluating the efficacy of APBI including the National Surgical Adjuvant Breast and Bowel Project-B39/Radiation Therapy Oncology Group 0413.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Long-Term Results of Radiochemotherapy for Solitary Lymph Node Metastasis After Curative Resection of Esophageal Cancer

Purpose: To evaluate the long-term efficacy and toxicity of definitive radiochemotherapy for solitary lymph node metastasis after curative surgery of esophageal cancer. Methods and Materials: We performed a retrospective review of 35 patients who underwent definitive radiochemotherapy at Tohoku University Hospital between 2000 and 2009 for solitary lymph node metastasis after curative esophagectomy with lymph node dissection for esophageal cancer. Radiotherapy doses ranged from 60 to 66 Gy (median, 60 Gy). Concurrent chemotherapy was platinum based in all patients. The endpoints of the present study were overall survival, cause-specific survival, progression-free survival, irradiated-field control, overall tumor response, and prognostic factors. Results: The median observation period for survivors was 70.0 months. The 5-year overall survival was 39.2% (median survival, 39.0 months). The 5-year cause-specific survival, progression-free survival, and irradiated-field control were 43.3%, 31.0% and 59.9%, respectively. Metastatic lesion, size of the metastatic lymph node, and performance status before radiochemotherapy were significantly correlated with prognosis. Complete response and partial response were observed in 22.9% and 57.1% of the patients, respectively. There was no Grade 3 or higher adverse effect based on theCommon Terminology Criteria for Adverse Events (CTCAE v3.0) in the late phase. Conclusions: Based on our study findings, approximately 40% of patients with solitary lymph node metastasis after curative resection for esophageal cancer have a chance of long-term survival with definitive radiochemotherapy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Postoperative Radiation Therapy With or Without Concurrent Chemotherapy for Node-Positive Thoracic Esophageal Squamous Cell Carcinoma

Purpose: To retrospectively compare the efficacy of radiation therapy (RT) and chemotherapy plus RT (CRT) for the postoperative treatment of node-positive thoracic esophageal squamous cell carcinoma (TESCC) and to determine the incidence and severity of toxic reactions. Methods and Materials: We retrospectively reviewed data from 304 patients who had undergone esophagectomy with 3-field lymph node dissection for TESCC and were determined by postoperative pathology to have lymph node metastasis without distant hematogenous metastasis. Of these patients, 164 underwent postoperative chemotherapy (cisplatin 80 mg/m{sup 2}, average days 1-3, plus paclitaxel 135 mg/m{sup 2}, day 1; 21-day cycle) plus RT (50 Gy), and 140 underwent postoperative RT alone. Results: The 5-year overall survival rates for the CRT and RT groups were 47.4% and 38.6%, respectively (P=.030). The distant metastasis rate, the mixed (regional lymph node and distant) metastasis rate, and the overall recurrence rate were significantly lower in the CRT group than in the RT group (P<.05). However, mild and severe early toxic reactions, including neutropenia, radiation esophagitis, and gastrointestinal reaction, were significantly more common in the CRT group than in the RT group (P<.05). No significant differences in incidence of late toxic reactions were found between the 2 groups. Conclusions: Our results show that in node-positive TESCC patients, postoperative CRT is significantly more effective than RT alone at increasing the overall survival and decreasing the rates of distant metastasis, mixed metastasis, and overall recurrence. Severe early toxic reactions were more common with CRT than with RT alone, but patients could tolerate CRT.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Distribution of Cervical Lymph Node Metastases From Squamous Cell Carcinoma of the Oropharynx in the Era of Risk Stratification Using Human Papillomavirus and Smoking Status

Purpose/Objective(s): To investigate the factors contributing to the clinical presentation of oropharyngeal squamous cell carcinoma (OPSCC) in the era of risk stratification using human papilloma virus (HPV) and smoking status. Methods and Materials: All patients with OPSCC presenting to our institutional multidisciplinary clinic from January 2009 to June 2015 were reviewed from a prospective database. The patients were grouped as being at low risk, intermediate risk, and high risk in the manner described by Ang et al. Variance in clinical presentation was examined using χ{sup 2}, Kruskal-Wallis, Mann-Whitney, and logistic regression analyses. Results: The rates of HPV/p16 positivity (P<.001), never-smoking (P=.016), and cervical lymph node metastases (P=.023) were significantly higher for patients with OPSCC of the tonsil, base of tongue (BOT), or vallecula subsites when compared with pharyngeal wall or palate subsites. Low-risk patients with tonsil, base of tongue, or vallecula primary tumors presented with nodal stage N2a at a much higher than expected frequency (P=.007), and high-risk patients presented with tumor stage T4 at a much higher than expected frequency (P=.003). Patients with BOT primary tumors who were never-smokers were less likely to have clinically involved ipsilateral neck disease than were former smokers (odds ratio 1.8; P=.038). The distribution of cervical lymph node metastases was not associated with HPV/p16 positivity, risk group, or subsite. When these data were compared with those in historical series, no significant differences were seen in the patterns of cervical lymph node metastases for patients with OPSCC. Conclusions: For patients with OPSCC differences in HPV status, smoking history and anatomic subsite were associated with differences in clinical presentation but not with distribution of cervical lymph node metastases. Historical series describing the patterns of cervical lymph node metastases in patients with OPSCC remain clinically relevant.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Autoencoder node saliency: Selecting relevant latent representations

The autoencoder is an artificial neural network that performs nonlinear dimension reduction and learns hidden representations of unlabeled data. With a linear transfer function it is similar to the principal component analysis (PCA). While both methods use weight vectors for linear transformations, the autoencoder does not come with any indication similar to the eigenvalues in PCA that are paired with eigenvectors. In this work, we propose a novel autoencoder node saliency method that examines whether the features constructed by autoencoders exhibit properties related to known class labels. The supervised node saliency ranks the nodes based on their capability of performing a learning task. It is coupled with the normalized entropy difference (NED). We establish a property for NED values to verify classifying behaviors among the top ranked nodes. Lastly, by applying our methods to real datasets, we demonstrate their ability to provide indications on the performing nodes and explain the learned tasks in autoencoders.

97 MATHEMATICS AND COMPUTING↗