Richard M. Golden

The :20Brain-state-in-a-box Neural model is a gradient descent algorithm

Abstract The Brain-State-in-a-Box (BSB) neural model ( J. A. Anderson, J. W. Silverstein, S. A. Ritz, & R. S. Jones, 1977 , Psychological Review , 84 , 413–451) is a pattern categorization device inspired by neurophysiological considerations. This model has additionally been applied to a fairly diverse range of psychological phenomena. In this paper, the BSB model is demonstrated to be a deterministic constrained gradient descent algorithm that minimizes a quadratic cost function. A formal proof that all trajectories of the BSB algorithm in state vector space approach the set of system equilibrium points, under certain specific conditions, is presented. Some conditions regarding the existence of global energy minima are also briefly discussed.

A unified framework for connectionist systems

Pattern classification using connectionist (i.e., neural network) models is viewed within a statistical framework. A connectionist networks subjective beliefs about its statistical environment are derived. This belief structure is the networks “subjective” probability distribution. Stimulus classification is interpreted as computing the “most probable” response for a given stimulus with respect to the subjective probability distribution. Given the subjective probability distribution, learning algorithms can be analyzed and designed using maximum likelihood estimation techniques, and statistical tests can be developed to evaluate and compare network architectures. The framework is applicable to many connectionist networks including those of Hopfield (1982, 1984), Cohen and Grossberg (1983), Anderson et al. (1977), and Rumelhart et al. (1986b).

Effective Software Fault Localization Using an RBF Neural Network

We propose the application of a modified radial basis function neural network in the context of software fault localization, to assist programmers in locating bugs effectively. This neural network is trained to learn the relationship between the statement coverage information of a test case and its corresponding execution result, success or failure. The trained network is then given as input a set of virtual test cases, each covering a single statement. The output of the network, for each virtual test case, is considered to be the suspiciousness of the corresponding covered statement. A statement with a higher suspiciousness has a higher likelihood of containing a bug, and thus statements can be ranked in descending order of their suspiciousness. The ranking can then be examined one by one, starting from the top, until a bug is located. Case studies on 15 different programs were conducted, and the results clearly show that our proposed technique is more effective than several other popular, state of the art fault localization techniques. Further studies investigate the robustness of the proposed technique, and illustrate how it can easily be applied to programs with multiple bugs as well.

Predictors of general quality of life in patients with benign prostate hyperplasia or prostate cancer

PURPOSE Studies in disease specific populations have emphasized disease specific quality of life with little study of general quality of life. Furthermore, studies of general quality of life in disease specific populations have mostly examined the importance of disease specific variables, and have generally yielded poor correlations of such variables and general quality of life. We attempted to model the emotional component of general quality of life in patients with prostate disease. MATERIALS AND METHODS We integrated prospectively collected disease specific and nonspecific clinical and self-reported patient data. We also applied neural network and more conventional statistical tools to examine the relative use of various available analytical methodologies in modeling general quality of life. RESULTS Neural networks created reasonably good models of the emotional component of general quality of life. Logistic regression analysis also created reasonably good models and, given current computational schemes, allowed for identification of significant inputs in the models more readily than did the feed-forward, back propagation neural networks. All models of general quality of life relied primarily on disease nonspecific inputs, including social support, activities of daily living and coping. CONCLUSIONS Our observations suggested that efforts to optimize general quality of life in patients with prostate disease must integrate disease nonspecific variables.

Discrepancy Risk Model Selection Test theory for comparing possibly misspecified or nonnested models

A new model selection statistical test is proposed for testing the null hypothesis that two probability models equally effectively fit the underlying data generating process (DGP). The new model selection test, called the Discrepancy Risk Model Selection Test (DRMST), extends previous work (see Vuong, 1989) on this problem in four distinct ways. First, generalized goodness-of-fit measures (which include log-likelihood functions) can be used. Second, unlike the classical likelihood ratio test, the models are not required to be fully nested where the nesting concept is defined for generalized goodness-of-fit measures. The DRMST also differs from the likelihood ratio test by not requiring that either competing model provides a completely accurate representation of the DGP. And, fourth, the DRMST may be used to compare competing time-series models using correlated observations as well as data consisting of independent and identically distributed observations.

A parallel distributed processing model of story comprehension and recall

An optimal control theory of story comprehension and recall is proposed within the framework of a “situation”‐state space. A point in situation‐state space is specified by a collection of propositions, each of which can have the values of either “present” or “absent.” A trajectory in situation‐state space is a temporally ordered sequence of situations. A readers knowledge that the occurrence of one situation is likely to cause the occurrence of another situation is represented by a subjective conditional probability distribution. A multistate probabilistic (MSP) causal chain notation is also introduced for conveniently describing the knowledge structures implicitly represented by the subjective conditional probability distribution. A story is represented as a partially specified trajectory in situation‐state space, and thus, story comprehension is defined as the problem of inferring the most probable missing features of the partially specified story trajectory. The story‐recall process is also viewed as ...

Early blood product and crystalloid volume resuscitation: Risk association with multiple organ dysfunction after severe blunt traumatic injury

BACKGROUND Elements of volume resuscitation from hemorrhagic shock, such as amount of blood product and crystalloid administration, have been shown to be associated with multiple organ dysfunction (MOD). However, it is unknown whether these are causative factors or merely markers of an underlying requirement for large-volume resuscitation. We sought to further delineate the relevance of the major individual components of early volume resuscitation to onset of MOD after severe blunt traumatic injury. METHODS We performed a secondary analysis of a large, multicenter prospective observational cohort of severely injured blunt trauma patients, the NIGMS Trauma Glue Grant, to assess the relevance of individual components of resuscitation administered in the first 12 hours of resuscitation including packed red blood cells (PRBC), fresh frozen plasma (FFP), and isotonic crystalloid, to the onset of MOD within the first 28 days after injury. Deaths within 48 hours of injury were excluded. We used a two tiered, exhaustive logistic regression model search technique to adjust for potential confounders from clinically relevant MOD covariates, including indicators of shock severity, injury severity, comorbidities, age, and gender. RESULTS The study cohort consisted of 1,366 severely injured blunt trauma patients (median new Injury Severity Score = 34). Incidence of 28-day Marshall MOD was 19.6%. Transfusion of ≥10 Units of PRBC in the first 12 hours (odds ratio, 2.06; 95% confidence interval 1.44-2.94), but not FFP (≥8 U) or large volume crystalloid administration (≥12 L), was independently associated with onset of 28-day Marshall MOD. PRBC:FFP ratio in the first 12 hours was not significantly associated with MOD. CONCLUSIONS When controlling for all major components of acute volume resuscitation, massive-transfusion volumes of PRBCs within the first 12 hours of resuscitation are modestly associated with MOD, whereas FFP and large volume crystalloid administration are not independently associated with MOD. Previous reported associations of blood products and large-volume crystalloid with MOD may be reflecting overall resuscitation requirements and burden of injury rather than independent causation.

Comparing clinical predictors of deep venous thrombosis versus pulmonary embolus after severe injury: A new paradigm for posttraumatic venous thromboembolism?

BACKGROUND: The traditional paradigm is that deep venous thrombosis (DVT) and pulmonary embolus (PE) are different temporal phases of a single disease process, most often labeled as the composite end point venous thromboembolism (VTE). However, we theorize that after severe blunt injury, DVT and PE may represent independent thrombotic entities rather than different stages of a single pathophysiologic process and therefore exhibit different clinical risk factor profiles. METHODS: We examined a large, multicenter prospective cohort of severely injured blunt trauma patients to compare clinical risk factors for DVT and PE, including indicators of injury severity, shock, resuscitation parameters, comorbidities, and VTE prophylaxis. Independent risk factors for each outcome were determined by cross‐validated logistic regression modeling using advanced exhaustive model search procedures. RESULTS: The study cohort consisted of 1,822 severely injured blunt trauma patients (median Injury Severity Score [ISS], 33; median base deficit, ‐9.5). Incidence of DVT and PE were 5.1% and 3.9%, respectively. Only 9 (5.7%) of 73 patients with a PE were also diagnosed with DVT. Independent risk factors associated with DVT include prophylaxis initiation within 48 hours (odds ratio [OR], 0.57; 95% confidence interval [CI], 0.36–0.90) and thoracic Abbreviated Injury Scale (AIS) score of 3 or greater (OR, 1.82; 95% CI, 1.12–2.95), while independent risk factors for PE were serum lactate of greater than 5 (OR, 2.33; 95% CI, 1.43–3.79) and male sex (OR, 2.12; 95% CI, 1.17–3.84). Both DVT and PE exhibited differing risk factor profiles from the classic composite end point of VTE. CONCLUSION: DVT and PE exhibit differing risk factor profiles following severe injury. Clinical risk factors for diagnosis of DVT after severe blunt trauma include the inability to initiate prompt pharmacologic prophylaxis and severe thoracic injury, which may represent overall injury burden. In contrast, risk factors for PE are male sex and physiologic evidence of severe shock. We hypothesize that postinjury DVT and PE may represent a broad spectrum of pathologic thrombotic processes as opposed to the current conventional wisdom of peripheral thrombosis and subsequent embolus. LEVEL OF EVIDENCE: Prognostic study, level III.

Multiuser interference suppression using block Shanno constant modulus algorithm

We extend the results from Wang and Dowling (see. Proc. ICASSP, Atlanta, GA, p.2678-81, 1996) to provide a low-cost and high-performance blind adaptive interference suppression scheme for CDMA systems. Simulation results confirm that the proposed algorithm outperforms several existing algorithms, even under the most severe cases of near-far and multipath conditions.

Use of a neural network to predict stone growth after shock wave lithotripsy

OBJECTIVES To determine whether a neural network is superior to standard computational methods in predicting stone regrowth after shock wave lithotripsy (SWL) and to determine whether the presence of residual fragments, as an independent variable, increases risk. METHODS We reviewed the records of 98 patients with renal or ureteral calculi treated by primary SWL at a single institution and followed up for at least 1 year; residual stone fragment growth or new stone occurrence was determined from abdominal radiographs. A neural network was programmed and trained to predict an increased stone volume over time utilizing input variables, including previous stone events, metabolic abnormality, directed medical therapy, infection, caliectasis, and residual fragments after SWL. Patient data were partitioned into a training set of 65 examples and a test set of 33. The neural network did not encounter the test set until training was complete. RESULTS The average follow-up period was 3.5 years (range 1 to 10). Of 98 patients, 47 had residual stone fragments 3 months after SWL; of these 47, 8 had increased stone volume at last follow-up visit. Of 51 patients stone free after SWL, 4 had stone recurrence. Coexisting risk factors were incorporated into a neural computational model to determine which of the risk factors was individually predictive of stone growth. The classification accuracy of the neural model in the test set was 91%, with a sensitivity of 91%, a specificity of 92%, and a receiver operating characteristic curve area of 0.964, results significantly better than those yielded by linear and quadratic discriminant function analysis. CONCLUSIONS A computational tool was developed to predict accurately the risk of future stone activity in patients treated by SWL. Use of the neural network demonstrates that none of the risk factors for stone growth, including the presence of residual fragments, is individually predictive of continuing stone formation.