A nonlinear supervised learning model, the Specialized Mappings Architecture (SMA), is described ... more A nonlinear supervised learning model, the Specialized Mappings Architecture (SMA), is described and applied to the estimation of human body pose from monocular images. The SMA consists of several specialized forward mapping functions and an inverse mapping function. Each specialized function maps certain domains of the input space (image features) onto the output space (body pose parameters). The key algorithmic problems faced are those of learning the specialized domains and mapping functions in an optimal way, as well as performing inference given inputs and knowledge of the inverse function. Solutions to these problems employ the EM algorithm and alternating choices of conditional independence assumptions. Performance of the approach is evaluated with synthetic and real video sequences of human motion.
Multi-class problems have a richer structure than binary classification problems. Thus, they can ... more Multi-class problems have a richer structure than binary classification problems. Thus, they can potentially improve their performance by exploiting the relationship among class labels. While for the purposes of providing an automated classification result this class structure does not need to be explicitly unveiled, for human level analysis or interpretation this is valuable. We develop a multi-class large margin classifier that extracts and takes advantage of class relationships. We provide a bi-convex formulation that explicitly learns a matrix that captures these class relationships and is de-coupled from the feature weights. Our representation can take advantage of the class structure to compress the model by reducing the number of classifiers employed, maintaining high accuracy even with large compression. In addition, we present an efficient formulation in terms of speed and memory.
Hidden Markov models have been used to restore recorded signals of single ion channels buried in ... more Hidden Markov models have been used to restore recorded signals of single ion channels buried in background noise. Parameter estimation and signal restoration are usually carried out through likelihood maximization by using variants of the Baum-Welch forward-backward procedures. This paper presents an alternative approach for dealing with this inferential task. The inferences are made by using a combination of the framework provided by Bayesian statistics and numerical methods based on Markov chain Monte Carlo stochastic simulation. The reliability of this approach is tested by using synthetic signals of known characteristics. The expectations of the model parameters estimated here are close to those calculated using the Baum-Welch algorithm, but the present methods also yield estimates of their errors. Comparisons of the results of the Bayesian Markov Chain Monte Carlo approach with those obtained by filtering and thresholding demonstrate clearly the superiority of the new methods.
This article describes the development of an instrument to measure the multiple dimensions of stu... more This article describes the development of an instrument to measure the multiple dimensions of student-faculty interactions. The sample consisted of 318 students (114 males, 203 females; 58% White, 16% African American, 9% Hispanic Americans) who completed the Student-Professor Interaction Scale (SPIS). Eight dimensions were identified, with Cronbach alphas ranging from 51 to .92. Dimensions ofstudent-faculty interactions were related to academic motivation and academic self-conceptfor the majority White sample, but only academic self-concept for the ethnic minority sample. African American and Hispanic American students reported feeling less connected with professors, perceived their experiences with faculty as more negative, and perceived faculty as less respectful when compared to White students. Implications for student affairs research andpractice are discussed. The importance of student-faculty interactions in facilitating the intellectual and personal growth of college students cannot be overstated. Wlodkowski and Ginsberg (1995) make the following observation: "People who feel unsafe, unconnected, and disrespected are unlikely to be motivated to learn. This is as true in college as it is in elementary school" (p. 2). The most utilized assessment of student-faculty interactions, operationalized by 10 items on the College Student Experiences Questionnaire (CSEQ), focuses on the frequency of interactions with faculty in different situations. While frequency of interactions is certainly one important aspect of student-faculty interaction, it does not include other dimensions that we believe are central in fully conceptualizing and understanding the construct. Therefore, there is a need to develop an instrument that assesses different dimensions of student-faculty interactions. In their influential book Education and Identity, Chickering and Reisser (1993) state that next to peer relations, relationships with faculty are among
Soldner, JL, & Rosales, R.(June, 2011). Examining behavioral approaches to post-‐secondary teachi... more Soldner, JL, & Rosales, R.(June, 2011). Examining behavioral approaches to post-‐secondary teaching. In A. Foreman (Chair), A comparison of interteaching and classroom lecture in rehabilitation education. Symposium conducted at the annual meeting of the Association for Behavior Analysis International, Denver, CO.
SciELO - Scientific Electronic Library Online Browse SciELO, SciELO Library Error Detector An une... more SciELO - Scientific Electronic Library Online Browse SciELO, SciELO Library Error Detector An unexpected error occurred in SciELO servers. Click on the SciELO logo to browse the SciELO Library or use the browser "BACK" button to return to the previous page. ...
We present an approach for recovering articulated body pose from single monocular images using th... more We present an approach for recovering articulated body pose from single monocular images using the Special-ized Mappings Architecture (SMA), a non-linear supervised learning architecture. SMA's consist of several specialized foiward (input to output space) mapping ...
We have used the gel retardation and DNase I footprinting assays to investigate the in vitro bind... more We have used the gel retardation and DNase I footprinting assays to investigate the in vitro binding of nuclear proteins to the octamer motif present in domain A of the SV40 enhancer and in other enhancer and promoter elements. Three apparently cell-specific (oct-BlA, oct-BlB and oct-B2) and one ubiquitous (oct-B3) proteins were detected in various lymphoid and non-lymphoid cell extracts. We show that the previously described 'ubiquitous' NF-A1 factor may correspond in fact to two proteins, oct-BlA in HeLa cells and oct-BiB in lymphoid cells. Interestingly, the HeLa cell protein oct-BlA formed a complex with the SV40 octamer, which could be detected in gel retardation, but not in DNase I footprinting assays. This absence of protection from DNase I digestion correlates with the inactivity of the SV40 octamer in HeLa cells in vivo. We have also found that the in vitro interaction between the SV40 octamer motif and the lymphoid cell-specific protein oct-B2 was negatively modulated by a component present in the nuclear extracts from several lymphoid cell lines. The interactions between the multiple octamer-binding proteins and the related octamer motifs present in other promoter and enhancer elements were systematically compared and the possible role of these proteins in the control of transcription is discussed.
Human papillomavirus (HPV) infection is associated with cervical cancer. The E2 and E1 papillomav... more Human papillomavirus (HPV) infection is associated with cervical cancer. The E2 and E1 papillomavirus proteins are expressed at the early stage of infection and regulate DNA replication. The E2 protein activates and represses transcription from different HPVs promoters. At some stage when viral DNA gets integrated into the cellular genome, the E2 gene is disrupted or inactivated. This event leads to a derepression of the E6 and E7 viral oncogenes. These viral proteins are required normally for the maintenance of the malignant phenotype. Therefore, the E2, E6, and E7 proteins are present in all patients infected by papillomavirus. In this study, the association of antibody levels against E2, E6, and E7 proteins of HPV types 16, 18, and 6 was determined in relation to the presence of HPV DNA at the initial stages of HPV infection. Serum samples from 172 women with HPV infection, determined by Papanicolau (Pap) smears and colposcopy, were tested. Elevated antibody titers against E2 protein from the HPV 6 and HPV 16 were detected in 46.42 and 66.96% of the patients, respectively. Antibodies against the E7 and E6 proteins of HPV 16 were found in 51.78 and 36.60% of the patients, respectively. Antibodies against the E6 and E7 proteins of HPV 18 were 35 and 45%, respectively. A statistical difference was found for antibody titers against the E2, E6, and E7 proteins between patients with papillomavirus DNA and controls cases who had no cytological abnormalities and no HPV DNA. Sera titers were 1/500 for patients HPV positive and 1/50 for control individuals. Antibodies titers against E6 and E7 proteins were also examined in patients at 6 and 24 months after cryosurgery. In these patients, a slight decrease in the antibody level against the E2, E6, and E7 proteins was found. No correlation was found between age and number of sexual partners, with serum positivity to the E2, E6, and E7 papillomavirus proteins. These data suggest that antibodies against the E2, E6, and E7 proteins are good candidates for use as markers for monitoring cervical HPV infections.
In-vehicle power line communication (PLC) holds great promises as an enabler of in-vehicle commun... more In-vehicle power line communication (PLC) holds great promises as an enabler of in-vehicle communication without increasing weight, volume, or cost of the wiring harnesses. In this paper, we join a few recent works and present channel measurements for a specific compact car. We discuss and compare our measurement results with those obtained in previous campaigns. Furthermore, we evaluate our measurements from the viewpoint of high connectivity. That is, we are interested in the suitability of PLC for applications that require a reliable link from any node to any other node at any time. Perhaps not surprisingly, we find that advanced transmission techniques such as multihop or frequency diversity are necessary to achieve this goal.
Privacy-preserving data mining (PPDM) is an emergent research area that addresses the incorporati... more Privacy-preserving data mining (PPDM) is an emergent research area that addresses the incorporation of privacy preserving concerns to data mining techniques. In this paper we propose a privacy-preserving (PP) Cox model for survival analysis, and consider a real clinical setting where the data is horizontally distributed among different institutions. The proposed model is based on linearly projecting the data to a lower dimensional space through an optimal mapping obtained by solving a linear programming problem. Our approach differs from the commonly used random projection approach since it instead finds a projection that is optimal at preserving the properties of the data that are important for the specific problem at hand. Since our proposed approach produces an sparse mapping, it also generates a PP mapping that not only projects the data to a lower dimensional space but it also depends on a smaller subset of the original features (it provides explicit feature selection). Real data from several European healthcare institutions are used to test our model for survival prediction of non-small-cell lung cancer patients. These results are also confirmed using publicly available benchmark datasets. Our experimental results show that we are able to achieve a near-optimal performance without directly sharing the data across different data sources. This model makes it possible to conduct large-scale multi-centric survival analysis without violating privacy-preserving requirements.
We propose a new approach to test selection based on the discovery of subgroups of patients shari... more We propose a new approach to test selection based on the discovery of subgroups of patients sharing the same optimal test, and present its application to breast cancer diagnosis. Subgroups are defined in terms of background information about the patient. We automatically determine the best t subgroups a patient belongs to, and decide for the test proposed by their majority. We introduce the concept of prediction quality to measure how accurate the test outcome is regarding the disease status. The quality of a subgroup is then the best mean prediction quality of its members (choosing the same test for all). Incorporating the quality computation in the search heuristic enables a significant reduction of the search space. In experiments on breast cancer diagnosis data we showed that it is faster than the baseline algorithm APRIORI-SD while preserving its accuracy.
ABSTRACT Phase locked frequency combs spanning > 1.5 THz have been generated using passive... more ABSTRACT Phase locked frequency combs spanning > 1.5 THz have been generated using passively mode locked quantum dash based lasers emitting at 1.55 µm. These combs are particularly suited for future OFDM transmission systems.
The mode coherence of adjacent and non-adjacent spectral modes of a passively mode locked quantum... more The mode coherence of adjacent and non-adjacent spectral modes of a passively mode locked quantum dash (QDash) semiconductor laser are deduced through radio frequency beat-tone linewidth measurements. A wavelength conversion scheme that uses degenerate four wave mixing in a semiconductor optical amplifier is proposed which considerably extends the mode spacing beyond the limit imposed by conventional fast-photodetection and electrical spectrum analysis of around 100 GHz. Using this scheme, the mode coherence of the QDash laser was measured out to the thirty-first harmonic, or a mode separation of 1.5 THz.
Proceedings of the Third Workshop on Data Mining Case Studies and Practice Prize, Fifteenth Annual SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2009), 2009
Advances in medical technology have greatly increased information density for imaging studies. Th... more Advances in medical technology have greatly increased information density for imaging studies. This may result from increased spatial resolution facilitating greater anatomical detail, increased contrast resolution allowing evaluation of more subtle structures than previously possible, or increased temporal image acquisition rate.
A nonlinear supervised learning model, the Specialized Mappings Architecture (SMA), is described ... more A nonlinear supervised learning model, the Specialized Mappings Architecture (SMA), is described and applied to the estimation of human body pose from monocular images. The SMA consists of several specialized forward mapping functions and an inverse mapping function. Each specialized function maps certain domains of the input space (image features) onto the output space (body pose parameters). The key algorithmic problems faced are those of learning the specialized domains and mapping functions in an optimal way, as well as performing inference given inputs and knowledge of the inverse function. Solutions to these problems employ the EM algorithm and alternating choices of conditional independence assumptions. Performance of the approach is evaluated with synthetic and real video sequences of human motion.
Multi-class problems have a richer structure than binary classification problems. Thus, they can ... more Multi-class problems have a richer structure than binary classification problems. Thus, they can potentially improve their performance by exploiting the relationship among class labels. While for the purposes of providing an automated classification result this class structure does not need to be explicitly unveiled, for human level analysis or interpretation this is valuable. We develop a multi-class large margin classifier that extracts and takes advantage of class relationships. We provide a bi-convex formulation that explicitly learns a matrix that captures these class relationships and is de-coupled from the feature weights. Our representation can take advantage of the class structure to compress the model by reducing the number of classifiers employed, maintaining high accuracy even with large compression. In addition, we present an efficient formulation in terms of speed and memory.
Hidden Markov models have been used to restore recorded signals of single ion channels buried in ... more Hidden Markov models have been used to restore recorded signals of single ion channels buried in background noise. Parameter estimation and signal restoration are usually carried out through likelihood maximization by using variants of the Baum-Welch forward-backward procedures. This paper presents an alternative approach for dealing with this inferential task. The inferences are made by using a combination of the framework provided by Bayesian statistics and numerical methods based on Markov chain Monte Carlo stochastic simulation. The reliability of this approach is tested by using synthetic signals of known characteristics. The expectations of the model parameters estimated here are close to those calculated using the Baum-Welch algorithm, but the present methods also yield estimates of their errors. Comparisons of the results of the Bayesian Markov Chain Monte Carlo approach with those obtained by filtering and thresholding demonstrate clearly the superiority of the new methods.
This article describes the development of an instrument to measure the multiple dimensions of stu... more This article describes the development of an instrument to measure the multiple dimensions of student-faculty interactions. The sample consisted of 318 students (114 males, 203 females; 58% White, 16% African American, 9% Hispanic Americans) who completed the Student-Professor Interaction Scale (SPIS). Eight dimensions were identified, with Cronbach alphas ranging from 51 to .92. Dimensions ofstudent-faculty interactions were related to academic motivation and academic self-conceptfor the majority White sample, but only academic self-concept for the ethnic minority sample. African American and Hispanic American students reported feeling less connected with professors, perceived their experiences with faculty as more negative, and perceived faculty as less respectful when compared to White students. Implications for student affairs research andpractice are discussed. The importance of student-faculty interactions in facilitating the intellectual and personal growth of college students cannot be overstated. Wlodkowski and Ginsberg (1995) make the following observation: "People who feel unsafe, unconnected, and disrespected are unlikely to be motivated to learn. This is as true in college as it is in elementary school" (p. 2). The most utilized assessment of student-faculty interactions, operationalized by 10 items on the College Student Experiences Questionnaire (CSEQ), focuses on the frequency of interactions with faculty in different situations. While frequency of interactions is certainly one important aspect of student-faculty interaction, it does not include other dimensions that we believe are central in fully conceptualizing and understanding the construct. Therefore, there is a need to develop an instrument that assesses different dimensions of student-faculty interactions. In their influential book Education and Identity, Chickering and Reisser (1993) state that next to peer relations, relationships with faculty are among
Soldner, JL, & Rosales, R.(June, 2011). Examining behavioral approaches to post-‐secondary teachi... more Soldner, JL, & Rosales, R.(June, 2011). Examining behavioral approaches to post-‐secondary teaching. In A. Foreman (Chair), A comparison of interteaching and classroom lecture in rehabilitation education. Symposium conducted at the annual meeting of the Association for Behavior Analysis International, Denver, CO.
SciELO - Scientific Electronic Library Online Browse SciELO, SciELO Library Error Detector An une... more SciELO - Scientific Electronic Library Online Browse SciELO, SciELO Library Error Detector An unexpected error occurred in SciELO servers. Click on the SciELO logo to browse the SciELO Library or use the browser "BACK" button to return to the previous page. ...
We present an approach for recovering articulated body pose from single monocular images using th... more We present an approach for recovering articulated body pose from single monocular images using the Special-ized Mappings Architecture (SMA), a non-linear supervised learning architecture. SMA's consist of several specialized foiward (input to output space) mapping ...
We have used the gel retardation and DNase I footprinting assays to investigate the in vitro bind... more We have used the gel retardation and DNase I footprinting assays to investigate the in vitro binding of nuclear proteins to the octamer motif present in domain A of the SV40 enhancer and in other enhancer and promoter elements. Three apparently cell-specific (oct-BlA, oct-BlB and oct-B2) and one ubiquitous (oct-B3) proteins were detected in various lymphoid and non-lymphoid cell extracts. We show that the previously described 'ubiquitous' NF-A1 factor may correspond in fact to two proteins, oct-BlA in HeLa cells and oct-BiB in lymphoid cells. Interestingly, the HeLa cell protein oct-BlA formed a complex with the SV40 octamer, which could be detected in gel retardation, but not in DNase I footprinting assays. This absence of protection from DNase I digestion correlates with the inactivity of the SV40 octamer in HeLa cells in vivo. We have also found that the in vitro interaction between the SV40 octamer motif and the lymphoid cell-specific protein oct-B2 was negatively modulated by a component present in the nuclear extracts from several lymphoid cell lines. The interactions between the multiple octamer-binding proteins and the related octamer motifs present in other promoter and enhancer elements were systematically compared and the possible role of these proteins in the control of transcription is discussed.
Human papillomavirus (HPV) infection is associated with cervical cancer. The E2 and E1 papillomav... more Human papillomavirus (HPV) infection is associated with cervical cancer. The E2 and E1 papillomavirus proteins are expressed at the early stage of infection and regulate DNA replication. The E2 protein activates and represses transcription from different HPVs promoters. At some stage when viral DNA gets integrated into the cellular genome, the E2 gene is disrupted or inactivated. This event leads to a derepression of the E6 and E7 viral oncogenes. These viral proteins are required normally for the maintenance of the malignant phenotype. Therefore, the E2, E6, and E7 proteins are present in all patients infected by papillomavirus. In this study, the association of antibody levels against E2, E6, and E7 proteins of HPV types 16, 18, and 6 was determined in relation to the presence of HPV DNA at the initial stages of HPV infection. Serum samples from 172 women with HPV infection, determined by Papanicolau (Pap) smears and colposcopy, were tested. Elevated antibody titers against E2 protein from the HPV 6 and HPV 16 were detected in 46.42 and 66.96% of the patients, respectively. Antibodies against the E7 and E6 proteins of HPV 16 were found in 51.78 and 36.60% of the patients, respectively. Antibodies against the E6 and E7 proteins of HPV 18 were 35 and 45%, respectively. A statistical difference was found for antibody titers against the E2, E6, and E7 proteins between patients with papillomavirus DNA and controls cases who had no cytological abnormalities and no HPV DNA. Sera titers were 1/500 for patients HPV positive and 1/50 for control individuals. Antibodies titers against E6 and E7 proteins were also examined in patients at 6 and 24 months after cryosurgery. In these patients, a slight decrease in the antibody level against the E2, E6, and E7 proteins was found. No correlation was found between age and number of sexual partners, with serum positivity to the E2, E6, and E7 papillomavirus proteins. These data suggest that antibodies against the E2, E6, and E7 proteins are good candidates for use as markers for monitoring cervical HPV infections.
In-vehicle power line communication (PLC) holds great promises as an enabler of in-vehicle commun... more In-vehicle power line communication (PLC) holds great promises as an enabler of in-vehicle communication without increasing weight, volume, or cost of the wiring harnesses. In this paper, we join a few recent works and present channel measurements for a specific compact car. We discuss and compare our measurement results with those obtained in previous campaigns. Furthermore, we evaluate our measurements from the viewpoint of high connectivity. That is, we are interested in the suitability of PLC for applications that require a reliable link from any node to any other node at any time. Perhaps not surprisingly, we find that advanced transmission techniques such as multihop or frequency diversity are necessary to achieve this goal.
Privacy-preserving data mining (PPDM) is an emergent research area that addresses the incorporati... more Privacy-preserving data mining (PPDM) is an emergent research area that addresses the incorporation of privacy preserving concerns to data mining techniques. In this paper we propose a privacy-preserving (PP) Cox model for survival analysis, and consider a real clinical setting where the data is horizontally distributed among different institutions. The proposed model is based on linearly projecting the data to a lower dimensional space through an optimal mapping obtained by solving a linear programming problem. Our approach differs from the commonly used random projection approach since it instead finds a projection that is optimal at preserving the properties of the data that are important for the specific problem at hand. Since our proposed approach produces an sparse mapping, it also generates a PP mapping that not only projects the data to a lower dimensional space but it also depends on a smaller subset of the original features (it provides explicit feature selection). Real data from several European healthcare institutions are used to test our model for survival prediction of non-small-cell lung cancer patients. These results are also confirmed using publicly available benchmark datasets. Our experimental results show that we are able to achieve a near-optimal performance without directly sharing the data across different data sources. This model makes it possible to conduct large-scale multi-centric survival analysis without violating privacy-preserving requirements.
We propose a new approach to test selection based on the discovery of subgroups of patients shari... more We propose a new approach to test selection based on the discovery of subgroups of patients sharing the same optimal test, and present its application to breast cancer diagnosis. Subgroups are defined in terms of background information about the patient. We automatically determine the best t subgroups a patient belongs to, and decide for the test proposed by their majority. We introduce the concept of prediction quality to measure how accurate the test outcome is regarding the disease status. The quality of a subgroup is then the best mean prediction quality of its members (choosing the same test for all). Incorporating the quality computation in the search heuristic enables a significant reduction of the search space. In experiments on breast cancer diagnosis data we showed that it is faster than the baseline algorithm APRIORI-SD while preserving its accuracy.
ABSTRACT Phase locked frequency combs spanning > 1.5 THz have been generated using passive... more ABSTRACT Phase locked frequency combs spanning > 1.5 THz have been generated using passively mode locked quantum dash based lasers emitting at 1.55 µm. These combs are particularly suited for future OFDM transmission systems.
The mode coherence of adjacent and non-adjacent spectral modes of a passively mode locked quantum... more The mode coherence of adjacent and non-adjacent spectral modes of a passively mode locked quantum dash (QDash) semiconductor laser are deduced through radio frequency beat-tone linewidth measurements. A wavelength conversion scheme that uses degenerate four wave mixing in a semiconductor optical amplifier is proposed which considerably extends the mode spacing beyond the limit imposed by conventional fast-photodetection and electrical spectrum analysis of around 100 GHz. Using this scheme, the mode coherence of the QDash laser was measured out to the thirty-first harmonic, or a mode separation of 1.5 THz.
Proceedings of the Third Workshop on Data Mining Case Studies and Practice Prize, Fifteenth Annual SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2009), 2009
Advances in medical technology have greatly increased information density for imaging studies. Th... more Advances in medical technology have greatly increased information density for imaging studies. This may result from increased spatial resolution facilitating greater anatomical detail, increased contrast resolution allowing evaluation of more subtle structures than previously possible, or increased temporal image acquisition rate.
Uploads
Papers by Renzie Rosales