Mustafa Mustafa
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Mustafa Mustafa has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
- Abul Hayat
University of Arkansas 72701 Fayetteville AR
2 shared publications - George Stein
Canadian Institute for Theoretical Astrophysics, University of Toronto, 60 St George Street, Toronto, ON M5S 3H8, Canada
2 shared publications Zarija Lukić
2 shared publications- Amir Gholami
University of California Berkeley
1 shared publication - Azmi Levent Sağnak
Univ Hlth Sci Turkey, Diskapi Yildirim Beyazit Training & Res Hosp, Dept Urol, Ankara, Turkey
1 shared publication Deborah Bard
Lawrence Berkeley National Laboratory, Berkeley, USA
1 shared publication
Latest Publications
- 4 January 2026
- Published by Medical Research Center in AlQalam Journal of Medical and Applied Sciences
- p. 24
Recent Insights into Cytomegalovirus Serostatus among Pregnant Women in Tripoli, Libya
Cytomegalovirus (CMV) is a common viral infection that frequently impacts pregnant women globally and causes congenital infections. During pregnancy, primary or recent CMV infection may result in congenital malformations. Although certain Libyan studies have been published on maternal CMV infection, comprehensive data on its seroprevalence during pregnancy remain insufficient. Thus, this study aims to explore the current status of the seroprevalence and examine factors associated with CMV infection among Libyan pregnant women. A cross-sectional study was conducted from September to December 2025, recruiting 202 pregnant women aged 18-47 years who attended antenatal clinics in Tripoli, Libya. Out of 202 pregnant women, 75 were subjected to a concise questionnaire. Venous blood samples were collected, and serum was analyzed. CMV-specific IgG and IgM antibodies were detected using Electrochemiluminescence Immunoassay (ECLIA) on the Roche Cobas e 411 analyzer. Data were analyzed using Python software, and results were expressed as frequencies, percentages, and geometric mean titres (GMTs) with 95% confidence intervals (95% CI). The overall seroprevalence of CMV IgG was 64.85%, while the seropositivity rate for IgM was 27.23%. High IgG seropositivity was observed across all age groups, whereas IgM positivity varied, with the highest rate observed among women aged 30–34 years. Furthermore, we showed that both the first and second trimesters exhibited an increase in the IgG seroprevalence compared to the third trimester. Likewise, women residing in rural areas and those from larger households demonstrated higher CMV IgG levels. Miscarriage history showed minimal association with CMV antibody levels. There is a high prevalence of CMV infection among pregnant women in Tripoli, with most cases reflecting past exposure rather than recent infection. Our findings indicate a high level of immunity to CMV among Libyan pregnant women. Antibodies distribution was influenced by age, gestational period, residence, and family size, while miscarriage history had a limited effect. Our findings urgently call for early screening of CMV during pregnancy, along with increased awareness of preventive hygiene practices and the establishment of local guidelines aimed at reducing the risks of maternal and congenital infections.
- 4 January 2026
- Published by Medical Research Center in AlQalam Journal of Medical and Applied Sciences
- p. 24
Recent Insights into Cytomegalovirus Serostatus among Pregnant Women in Tripoli, Libya
Cytomegalovirus (CMV) is a common viral infection that frequently impacts pregnant women globally and causes congenital infections. During pregnancy, primary or recent CMV infection may result in congenital malformations. Although certain Libyan studies have been published on maternal CMV infection, comprehensive data on its seroprevalence during pregnancy remain insufficient. Thus, this study aims to explore the current status of the seroprevalence and examine factors associated with CMV infection among Libyan pregnant women. A cross-sectional study was conducted from September to December 2025, recruiting 202 pregnant women aged 18-47 years who attended antenatal clinics in Tripoli, Libya. Out of 202 pregnant women, 75 were subjected to a concise questionnaire. Venous blood samples were collected, and serum was analyzed. CMV-specific IgG and IgM antibodies were detected using Electrochemiluminescence Immunoassay (ECLIA) on the Roche Cobas e 411 analyzer. Data were analyzed using Python software, and results were expressed as frequencies, percentages, and geometric mean titres (GMTs) with 95% confidence intervals (95% CI). The overall seroprevalence of CMV IgG was 64.85%, while the seropositivity rate for IgM was 27.23%. High IgG seropositivity was observed across all age groups, whereas IgM positivity varied, with the highest rate observed among women aged 30–34 years. Furthermore, we showed that both the first and second trimesters exhibited an increase in the IgG seroprevalence compared to the third trimester. Likewise, women residing in rural areas and those from larger households demonstrated higher CMV IgG levels. Miscarriage history showed minimal association with CMV antibody levels. There is a high prevalence of CMV infection among pregnant women in Tripoli, with most cases reflecting past exposure rather than recent infection. Our findings indicate a high level of immunity to CMV among Libyan pregnant women. Antibodies distribution was influenced by age, gestational period, residence, and family size, while miscarriage history had a limited effect. Our findings urgently call for early screening of CMV during pregnancy, along with increased awareness of preventive hygiene practices and the establishment of local guidelines aimed at reducing the risks of maternal and congenital infections.
- Open Access
ADAHESSIAN: An Adaptive Second Order Optimizer for Machine Learning
We introduce ADAHESSIAN, a second order stochastic optimization algorithm which dynamically incorporates the curvature of the loss function via ADAptive estimates of the HESSIAN. Second order algorithms are among the most powerful optimization algorithms with superior convergence properties as compared to first order methods such as SGD and Adam. The main disadvantage of traditional second order methods is their heavier per iteration computation and poor accuracy as compared to first order methods. To address these, we incorporate several novel approaches in ADAHESSIAN, including: (i) a fast Hutchinson based method to approximate the curvature matrix with low computational overhead; (ii) a root-mean-square exponential moving average to smooth out variations of the Hessian diagonal across different iterations; and (iii) a block diagonal averaging to reduce the variance of Hessian diagonal elements. We show that ADAHESSIAN achieves new state-of-the-art results by a large margin as compared to other adaptive optimization methods, including variants of Adam. In particular, we perform extensive tests on CV, NLP, and recommendation system tasks and find that ADAHESSIAN: (i) achieves 1.80%/1.45% higher accuracy on ResNets20/32 on Cifar10, and 5.55% higher accuracy on ImageNet as compared to Adam; (ii) outperforms AdamW for transformers by 0.13/0.33 BLEU score on IWSLT14/WMT14 and 2.7/1.0 PPL on PTB/Wikitext-103; (iii) outperforms AdamW for SqueezeBert by 0.41 points on GLUE; and (iv) achieves 0.032% better score than Adagrad for DLRM on the Criteo Ad Kaggle dataset. Importantly, we show that the cost per iteration of ADAHESSIAN is comparable to first order methods, and that it exhibits robustness towards its hyperparameters.
Open AccessADAHESSIAN: An Adaptive Second Order Optimizer for Machine Learning
We introduce ADAHESSIAN, a second order stochastic optimization algorithm which dynamically incorporates the curvature of the loss function via ADAptive estimates of the HESSIAN. Second order algorithms are among the most powerful optimization algorithms with superior convergence properties as compared to first order methods such as SGD and Adam. The main disadvantage of traditional second order methods is their heavier per iteration computation and poor accuracy as compared to first order methods. To address these, we incorporate several novel approaches in ADAHESSIAN, including: (i) a fast Hutchinson based method to approximate the curvature matrix with low computational overhead; (ii) a root-mean-square exponential moving average to smooth out variations of the Hessian diagonal across different iterations; and (iii) a block diagonal averaging to reduce the variance of Hessian diagonal elements. We show that ADAHESSIAN achieves new state-of-the-art results by a large margin as compared to other adaptive optimization methods, including variants of Adam. In particular, we perform extensive tests on CV, NLP, and recommendation system tasks and find that ADAHESSIAN: (i) achieves 1.80%/1.45% higher accuracy on ResNets20/32 on Cifar10, and 5.55% higher accuracy on ImageNet as compared to Adam; (ii) outperforms AdamW for transformers by 0.13/0.33 BLEU score on IWSLT14/WMT14 and 2.7/1.0 PPL on PTB/Wikitext-103; (iii) outperforms AdamW for SqueezeBert by 0.41 points on GLUE; and (iv) achieves 0.032% better score than Adagrad for DLRM on the Criteo Ad Kaggle dataset. Importantly, we show that the cost per iteration of ADAHESSIAN is comparable to first order methods, and that it exhibits robustness towards its hyperparameters.
- Open Access
Self-Supervised Representation Learning for Astronomical Images
Sky surveys are the largest data generators in astronomy, making automated tools for extracting meaningful scientific information an absolute necessity. We show that, without the need for labels, self-supervised learning recovers representations of sky survey images that are semantically useful for a variety of scientific tasks. These representations can be directly used as features, or fine-tuned, to outperform supervised methods trained only on labeled data. We apply a contrastive learning framework on multi-band galaxy photometry from the Sloan Digital Sky Survey (SDSS) to learn image representations. We then use them for galaxy morphology classification, and fine-tune them for photometric redshift estimation, using labels from the Galaxy Zoo 2 dataset and SDSS spectroscopy. In both downstream tasks, using the same learned representations, we outperform the supervised state-of-the-art results, and we show that our approach can achieve the accuracy of supervised models while using 2-4 times fewer labels for training.
Open AccessSelf-Supervised Representation Learning for Astronomical Images
Sky surveys are the largest data generators in astronomy, making automated tools for extracting meaningful scientific information an absolute necessity. We show that, without the need for labels, self-supervised learning recovers representations of sky survey images that are semantically useful for a variety of scientific tasks. These representations can be directly used as features, or fine-tuned, to outperform supervised methods trained only on labeled data. We apply a contrastive learning framework on multi-band galaxy photometry from the Sloan Digital Sky Survey (SDSS) to learn image representations. We then use them for galaxy morphology classification, and fine-tune them for photometric redshift estimation, using labels from the Galaxy Zoo 2 dataset and SDSS spectroscopy. In both downstream tasks, using the same learned representations, we outperform the supervised state-of-the-art results, and we show that our approach can achieve the accuracy of supervised models while using 2-4 times fewer labels for training.
- Open Access
- 1 April 2021
- Published by American Astronomical Society in The Astrophysical Journal Letters
- Vol. 911 (2) , L33
Self-supervised Representation Learning for Astronomical Images
Sky surveys are the largest data generators in astronomy, making automated tools for extracting meaningful scientific information an absolute necessity. We show that, without the need for labels, self-supervised learning recovers representations of sky survey images that are semantically useful for a variety of scientific tasks. These representations can be directly used as features, or fine-tuned, to outperform supervised methods trained only on labeled data. We apply a contrastive learning framework on multiband galaxy photometry from the Sloan Digital Sky Survey (SDSS), to learn image representations. We then use them for galaxy morphology classification and fine-tune them for photometric redshift estimation, using labels from the Galaxy Zoo 2 data set and SDSS spectroscopy. In both downstream tasks, using the same learned representations, we outperform the supervised state-of-the-art results, and we show that our approach can achieve the accuracy of supervised models while using 2–4 times fewer labels for training. The codes, trained models, and data can be found at https://portal.nersc.gov/project/dasrepo/self-supervised-learning-sdss.
- 1 April 2021
- Published by American Astronomical Society in The Astrophysical Journal Letters
- Vol. 911 (2) , L33
Open AccessSelf-supervised Representation Learning for Astronomical Images
Sky surveys are the largest data generators in astronomy, making automated tools for extracting meaningful scientific information an absolute necessity. We show that, without the need for labels, self-supervised learning recovers representations of sky survey images that are semantically useful for a variety of scientific tasks. These representations can be directly used as features, or fine-tuned, to outperform supervised methods trained only on labeled data. We apply a contrastive learning framework on multiband galaxy photometry from the Sloan Digital Sky Survey (SDSS), to learn image representations. We then use them for galaxy morphology classification and fine-tune them for photometric redshift estimation, using labels from the Galaxy Zoo 2 data set and SDSS spectroscopy. In both downstream tasks, using the same learned representations, we outperform the supervised state-of-the-art results, and we show that our approach can achieve the accuracy of supervised models while using 2–4 times fewer labels for training. The codes, trained models, and data can be found at https://portal.nersc.gov/project/dasrepo/self-supervised-learning-sdss.
- Open Access
- 6 May 2019
- Published by Springer Nature in Computational Astrophysics and Cosmology
- Vol. 6 (1) , 1
CosmoGAN: creating high-fidelity weak lensing convergence maps using Generative Adversarial Networks
Inferring model parameters from experimental data is a grand challenge in many sciences, including cosmology. This often relies critically on high fidelity numerical simulations, which are prohibitively computationally expensive. The application of deep learning techniques to generative modeling is renewing interest in using high dimensional density estimators as computationally inexpensive emulators of fully-fledged simulations. These generative models have the potential to make a dramatic shift in the field of scientific simulations, but for that shift to happen we need to study the performance of such generators in the precision regime needed for science applications. To this end, in this work we apply Generative Adversarial Networks to the problem of generating weak lensing convergence maps. We show that our generator network produces maps that are described by, with high statistical confidence, the same summary statistics as the fully simulated maps.
- 6 May 2019
- Published by Springer Nature in Computational Astrophysics and Cosmology
- Vol. 6 (1) , 1
Open AccessCosmoGAN: creating high-fidelity weak lensing convergence maps using Generative Adversarial Networks
Inferring model parameters from experimental data is a grand challenge in many sciences, including cosmology. This often relies critically on high fidelity numerical simulations, which are prohibitively computationally expensive. The application of deep learning techniques to generative modeling is renewing interest in using high dimensional density estimators as computationally inexpensive emulators of fully-fledged simulations. These generative models have the potential to make a dramatic shift in the field of scientific simulations, but for that shift to happen we need to study the performance of such generators in the precision regime needed for science applications. To this end, in this work we apply Generative Adversarial Networks to the problem of generating weak lensing convergence maps. We show that our generator network produces maps that are described by, with high statistical confidence, the same summary statistics as the fully simulated maps.
- Open Access
- 1 October 2017
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 898 (8) , 082023
STAR Data Reconstruction at NERSC/Cori, an adaptable Docker container approach for HPC
As HPC facilities grow their resources, adaptation of classic HEP/NP workflows becomes a need. Linux containers may very well offer a way to lower the bar to exploiting such resources and at the time, help collaboration to reach vast elastic resources on such facilities and address their massive current and future data processing challenges. In this proceeding, we showcase STAR data reconstruction workflow at Cori HPC system at NERSC. STAR software is packaged in a Docker image and runs at Cori in Shifter containers. We highlight two of the typical end-to-end optimization challenges for such pipelines: 1) data transfer rate which was carried over ESnet after optimizing end points and 2) scalable deployment of conditions database in an HPC environment. Our tests demonstrate equally efficient data processing workflows on Cori/HPC, comparable to standard Linux clusters.
- 1 October 2017
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 898 (8) , 082023
Open AccessSTAR Data Reconstruction at NERSC/Cori, an adaptable Docker container approach for HPC
As HPC facilities grow their resources, adaptation of classic HEP/NP workflows becomes a need. Linux containers may very well offer a way to lower the bar to exploiting such resources and at the time, help collaboration to reach vast elastic resources on such facilities and address their massive current and future data processing challenges. In this proceeding, we showcase STAR data reconstruction workflow at Cori HPC system at NERSC. STAR software is packaged in a Docker image and runs at Cori in Shifter containers. We highlight two of the typical end-to-end optimization challenges for such pipelines: 1) data transfer rate which was carried over ESnet after optimizing end points and 2) scalable deployment of conditions database in an HPC environment. Our tests demonstrate equally efficient data processing workflows on Cori/HPC, comparable to standard Linux clusters.
- Open Access
- 1 June 2011
- Published by University of Toronto Press Inc. (UTPress) in Clinical & Investigative Medicine
- Vol. 34 (3) , E179-E183
Assessment of anti-sperm antibodies in couples after testicular sperm extraction
Purpose: Testicular spermatozoa can be retrieved successfully by the testicular sperm extraction (TESE) procedure and used for intracytoplasmic sperm injection. Disruption in the blood-testis barrier can lead to the production of antisperm antibodies (ASA). The aim of this prospective study was to investigate the frequency of ASA formation in couples after TESE procedure. Methods: Thirty-seven couples were included in the study at the Urology Clinic of the Dr. Zekai Tahir Burak Women’s Health Training and Research Hospital. History, physical examination, spermiogram, and endocrine profiles were obtained for all male patients. All the male patients in this study had been diagnosed with nonobstructive azoospermia (NOA) and underwent microdissection TESE. Secondary and tertiary cases were also included in the study. Serum samples were obtained from all 74 patients before TESE, and at three and 12 months after TESE. Serum ASA levels were determined. ANOVA was performed for statistical analysis for serum Follicle-Stimulating Hormone (FSH), testosterone and testicular volume. P < 0.05 was considered significant. Results: There were no differences in the testicular volumes, serum FSH and testosterone levels before and after TESE. None of the patients or their partners developed significant levels of ASA as a result of the TESE procedure. Conclusion: TESE procedure does not cause ASA production in either males or their female partners.
- 1 June 2011
- Published by University of Toronto Press Inc. (UTPress) in Clinical & Investigative Medicine
- Vol. 34 (3) , E179-E183
Open AccessAssessment of anti-sperm antibodies in couples after testicular sperm extraction
Purpose: Testicular spermatozoa can be retrieved successfully by the testicular sperm extraction (TESE) procedure and used for intracytoplasmic sperm injection. Disruption in the blood-testis barrier can lead to the production of antisperm antibodies (ASA). The aim of this prospective study was to investigate the frequency of ASA formation in couples after TESE procedure. Methods: Thirty-seven couples were included in the study at the Urology Clinic of the Dr. Zekai Tahir Burak Women’s Health Training and Research Hospital. History, physical examination, spermiogram, and endocrine profiles were obtained for all male patients. All the male patients in this study had been diagnosed with nonobstructive azoospermia (NOA) and underwent microdissection TESE. Secondary and tertiary cases were also included in the study. Serum samples were obtained from all 74 patients before TESE, and at three and 12 months after TESE. Serum ASA levels were determined. ANOVA was performed for statistical analysis for serum Follicle-Stimulating Hormone (FSH), testosterone and testicular volume. P < 0.05 was considered significant. Results: There were no differences in the testicular volumes, serum FSH and testosterone levels before and after TESE. None of the patients or their partners developed significant levels of ASA as a result of the TESE procedure. Conclusion: TESE procedure does not cause ASA production in either males or their female partners.