Associate / Full Professor of Theoretical Biophysics and Machine Learning

A world from which we demand more and more requires people who can make a contribution. Critical thinkers who will take a closer look at what is really important. As a Professor, you will perform leading research and teach students in the area of theoretical biophysics and physics-based machine learning, to strengthen the role and visibility of the international Theoretical Biophysics landscape.

As a successful candidate you will join the Department of Biophysics at the Donders Center for Neuroscience (DCN) and perform internationally leading theoretical research in an area of theoretical biophysics or physics-based machine learning. You are interested in applications of theoretical biophysics methods to neuroscience problems studied in the DCN, and you will engage actively in interdisciplinary research collaborations with other physicists in the Faculty of Science and with external partners. You will contribute to the teaching and the innovation of Radboud's popular theoretical machine learning and biophysics courses, and possibly contribute to other core undergraduate physics subjects taught at the Faculty of Science. You will supervise students' research projects at the Bachelor's, Master's and PhD levels. Finally, you will contribute to the effective administration of Radboud University and the acquisition of research funding, and will strengthen the role and visibility of Radboud University in the international Theoretical Biophysics landscape.

Profile

We are

The Donders Institute for Brain, Cognition and Behaviour of Radboud University seeks to appoint a Professor of Theoretical Biophysics and Machine Learning. The Donders Institute is a world-class research institute, housing more than 700 researchers devoted to understanding the mechanistic underpinnings of the human mind/brain. Research at the Donders Institute focuses on four themes:

Language and Communication

Perception, Action, and Decision-making

Development and Lifelong Plasticity

Natural Computing and Neurotechnology.

We have excellent and state-of-the-art research facilities available for a broad range of neuroscience research. The Donders Institute fosters a collaborative, multidisciplinary, supportive research environment with a diverse international staff. English is the lingua franca at the Institute.

You will join the academic staff of the Donders Center for Neuroscience (DCN) - one of the four Donders Centers at Radboud University's Faculty of Science. The Biophysics Department is part of the DCN. Neurophysicists at DCN mainly conduct experimental, theoretical and computational research into the principles of information processing by the brain, with particular focus on the mammalian auditory and visual systems. The Physics of Machine Learning and Complex Systems Group studies a broad range of theoretical topics, ranging from physics-based machine learning paradigms and quantum machine learning, via Bayesian inference and applications of statistical mechanics techniques in medical statistics, to network theory and the modelling of heterogeneous many-variable processes in physics and biology. The group engages in multiple national and international research collaborations, and participates in several multidisciplinary initiatives that support theoretical biophysics and machine learning research and teaching at Radboud University.

Radboud University actively supports equality, diversity and inclusion, and encourages applications from all sections of society. The university offers customised facilities to better align work and private life. Parents are entitled to partly paid parental leave and Radboud University employees enjoy flexibility in the way they structure their work. The university highly values the career development of its staff, which is facilitated by a variety of programmes. The Faculty of Science is an equal opportunity employer, committed to building a culturally diverse intellectual community, and as such encourages applications from women and minorities.

Radboud University

We want to get the best out of science, others and ourselves. Why? Because this is what the world around us desperately needs. Leading research and education make an indispensable contribution to a healthy, free world with equal opportunities for all. This is what unites the more than 24,000 students and 5,600 employees at Radboud University. And this requires even more talent, collaboration and lifelong learning. You have a part to play!

We offer

Additional employment conditions

Work and science require good employment practices. This is reflected in Radboud University's primary and secondary employment conditions. You can make arrangements for the best possible work-life balance with flexible working hours, various leave arrangements and working from home. You are also able to compose part of your employment conditions yourself, for example, exchange income for extra leave days and receive a reimbursement for your sports subscription. And of course, we offer a good pension plan. You are given plenty of room and responsibility to develop your talents and realise your ambitions. Therefore, we provide various training and development schemes.

Would you like more information?

For questions about the position, please contact Ton Coolen, Professor at +31 24 361 42 45 or ton.coolen@donders.ru.nl.

Practical information and applications

You can apply until 25 February 2022, exclusively using the button below. Kindly address your application to Ton Coolen. Please fill in the application form and attach the following documents:

The first round of interviews will take place around the end of March. You would preferably begin employment on 1 September 2022.

This vacancy was also published in a slightly modified form in 2021. Applicants who were rejected at that time are kindly requested not to apply again.

We can imagine you're curious about our application procedure. It offers a rough outline of what you can expect during the application process, how we handle your personal data and how we deal with internal and external candidates.

We drafted this vacancy to find and hire our new colleague ourselves. Recruitment agencies are kindly requested to refrain from responding.

The rest is here:
Associate / Full Professor of Theoretical Biophysics and Machine Learning job with RADBOUD UNIVERSITY NIJMEGEN | 278686 - Times Higher Education (THE)

January 24, 2022 - Neurobiologists at Harvard University have entered a joint research agreement with NTT Research, Inc., a division of NTT, to study animal neuro-responses with the hope of informing future artificial intelligence systems. The five-year research project, launched in the fall of 2021, enables researchers at the two organizations to collaboratively study how animals maintain behavioral flexibility, specifically in the task of navigation. Greater understanding of how this challenge is approached in biology may eventually enable the design of new computing machines with similar capabilities. The agreement was coordinated by Harvard Office of Technology Development.

The principal investigator is Venkatesh Murthy, PhD, the Raymond Leo Erikson Life Sciences Professor of Molecular and Cellular Biology at Harvard and the Paul J. Finnegan Family Director of the Center for Brain Science. Murthys counterpart at NTT Research for the joint project is Physics & Informatics (PHI) Lab Research Scientist Gautam Reddy, PhD, who was previously an Independent Post-Doctoral Fellow at Harvards NSF-Simons Center for Mathematical and Statistical Analysis of Biology.

This joint research aims to better elucidate how animals maintain the ability to respond appropriately to a wide variety of complex real-world scenarios. The investigators expect the results from one aspect of the research to be a source of new, biologically inspired ideas for artificial reinforcement learning systems that rely on representation learning. Such ideas have played a major role in recent advances in artificial intelligence. Results from another aspect of the research should provide a quantitative understanding of how animals track trails, as well as identify the basic elements of general behavioral strategies that perform flexibly and reliably in the real world. Murthys lab has a long track record in experimental and computational neurobiology. Expertise relevant to the joint research includes the ability to record from or image many individual neurons in the brain while an animal performs behavioral tasks. This technical expertise will enable the research team to understand what computations are performed by biological neural networks when an animal is navigating in a complex world.

Murthy and Reddy have previously worked together on understanding the computational principles behind olfaction. Their focus was on how the smell receptors in the nose respond to blends of odorous compounds. During his time at Harvards NSF-Simons Center for Mathematical Biology, Reddy worked on the theory behind how animals track scent trails and on developing a computational framework to explain how evolution optimizes organisms.

I am delighted to continue this line of inquiry with Dr. Reddy through the NTT Research PHI Lab, Murthy said. The brain is an example of an extremely efficient computational device, and plenty of phenomena within it remain unexplored and unexplained. We believe the results of these investigations in neurobiology will reveal basic understandings and prove useful in the field of artificial intelligence.

Efficient computation is at the heart of quantum computing and neuroscience. Inspired by neuroscience, recent advances in machine learning have recently begun to change how we process data, said NTTs PHI Lab Director Yoshihisa Yamamoto, PhD. This joint research project could provide a rich source of animal-inspired algorithms that generalize across various research domains within NTT and inspire truly novel interdisciplinary ideas.

Adapted from a press release by NTT Research.

Here is the original post:
Collaboration with NTT Research to advance computational neurobiology - Harvard Office of Technology Development

Snapchat is taking a proactive approach in fighting drug deals taking place on its social media platform. The company shared an update concerning its most recent efforts to halt the push to sell drugs through connections on its app.Snapchat typically makes the news when the social media platform goes dark, sending users in a frenzy wondering when their beloved app will be back up and running. But the extremely popular app, especially among teenagers and young adults, found itself in a different type of spotlight last October when NBC News did a story about the troubling drug deals presumably taking place on the app.

The report examined the death of teens and young adults who were suspected of buying fentanyl-laced drugs using Snapchat. In the report, it spoke about teens and young adults who had bought what they believed to be a prescription pill, but turned out to be a counterfeit pill containing deadly doses of fentanyl. Since that report, Snapchat has been ramping up its efforts to thwart drug deals on its platform.

Snap stated that it has a zero tolerance for drug dealing on Snapchat. It says it has made significant operational improvements over the past year toward its goal of completely eradicating drug dealers from its platform. It claims to take a holistic approach, which includes "deploying tools that proactively detect drug-related content, working with law enforcement to support their investigations, and provide in-app information and support to Snapchatters who search for drug-related terms through a new educational portal, Heads Up."

The social media company announced that it is adding two new partners to its Heads Up portal in order to provide important in-app resources to its users. Community Anti-Drug Coalitions of America (CADCA), is a nonprofit organization that focuses on creating safe, healthy and drug-free communities. Truth Initiative is the second addition, and is an organization that strives to steer teens and young adults away from smoking, vaping and nicotine in general. Along with these two new organizations being added, Snap will soon be releasing its next episode of it special Good Luck America series which will focus on fentanyl.

Snapchat is also updating its Quick Add suggestion feature in order to reduce interactions between kids and strangers. The company states, "In order to be discoverable in Quick Add by someone else, users under 18 will need to have a certain number of friends in common with that person." In the past, users would be given a list of recommended friends based on mutual connections, regardless if you knew the person in real life or not. Work is also being done on additional parental tools that it will roll out in the coming months.

Another way Snapchat is looking to deter drug dealers from using its platform, is in its cooperation with law enforcement. It has implemented measures using artificial intelligence (AI) and machine learning to identify drug slang and content on the app, and then works with law enforcement to report potential cases and to comply with information requests. Snapchat has increased its law enforcement operations team by 74% since its creation. And remarkably, Snapchat claims that a whopping 88% of drug related content it uncovers is proactively detected by its AI and machine learning algorithms. That's up from 33% since its previous update.

"When we find drug dealing activity, we promptly ban the account, use technology to block the offender from creating new accounts on Snapchat, and in some cases proactively refer the account to law enforcement for investigation," Snapchat says.

Continued here:
How Snapchat Is Using AI And Machine Learning To Thwart Drug Deals - Hot Hardware

Data source and study population

Data used in this study was drawn from a SingHEART prospective longitudinal cohort study (ClinicalTrials.gov Identifier: NCT02791152). The study is a multi-ethnic population-based study conducted on healthy Asians, aged 2169years old without known diabetes mellitus or prior cardiovascular disease (Ischemic heart disease, stroke, peripheral vascular disease). The study complied with the Declaration of Helsinki and written informed consent were given by participants. The study was approved by the SingHealth Centralized Institutional Review Board.

We included 600 volunteers, aged of 30years with valid calcium score, into the main analysis of this study. Two hundred volunteers under the age of 30years, who did not have a calcium score were excluded, as the calcium score was the main outcome of our analysis.

Subset analysis for activity tracker data was performed on 430 out of the 600 volunteers who had adequate data. Although subjects recruited were issued an activity tracker to be worn over a period of five days with first and last days of the study being partial days, there was inconsistent wearing of the activity. Discounting the partial days, each subject would yield an activity log for three complete tracking days (or equivalent to days with>20 valid hours of steps and sleep data)24,25. For data consistency and quality, subjects with improper activity tracker usage i.e. activity reading log less than five days and/or sleep reading log less than three days were censored.

Coronary artery calcium (CAC) scoring was used as the modelling outcome. The coronary calcium is a specific marker of coronary atherosclerosis, a precursor for coronary artery disease26; it also reflects arterial age under the influence of underlying comorbidities and lifestyle. The CAC score was also regarded as the best marker for risk prediction of cardiovascular events27,28.

This study stratified subjects into two classes of CVD risk. Low risk if their coronary artery calcium score were 0, and high risk if calcium score were 100 and above. Subjects who did not fall into these 2 categories were considered intermediate risk.

The aim of this study is to look at how accurate the machine learning algorithm is in handling different data types, in the task of predicting high risk and low risk patients, based on calcium score.

Table 1 summarizes the data from SingHEART that was used in this study.

Data variables were categorized into four groups; lifestyle survey questionnaires, blood test data, 24-h ambulatory blood pressure, and activity tracking data by commercially available Fitbit Charge HR29.

Data pre-processing, transformation and imputation were performed on the raw data. Variables selected were based on their a priori knowledge from previous publications on cardiovascular risk assessment1,2,3, and physician expert advice. In total, there were 30, 17, 12 and 16 unique variables in the respective groups: survey questionnaire, 24h blood pressure and heart rate monitoring, blood tests and Fitbit data.

The Framingham 10-year risk score was computed using seven traditional risk factors: gender, age, single timepoint systolic blood pressure, Total Cholesterol (TC), High Density Lipoprotein (HDL), smoking status and presence of diabetes. A Framingham risk score of<10% is consider low risk, while20% is considered high risk30.

Figure1 shows the methodological framework of the present study. Exploratory analysis showed that ensemble MLA classifiers were superior at discriminating low risk individuals while ensemble MLA regressors performed better identifying individuals with high CVD risk. To leverage on the merits of both the classifiers and regressors MLA, we used both approaches for our model.

Modelling flow chart using ensemble MLA for cardiovascular risk prediction.

The ensemble classifiers produce a binary prediction outcome; low or non-low risk. The ensemble regressors makes a numerical prediction on the calcium score for individuals classified as non-low risk, and stratify into three bins of low, high, and intermediate risk. The predicted numerical values may range from negative to positive number. Negative predicted values were first converted to zero and subsequently the continuous predictions were converted to discrete bins using unique value percentile discretization ensuring records with the same numerical prediction are assigned the same risk category. Finally, the prediction outcome resides in a decision node build on a rule-based logic. The decision node assigns an outcome of low risk if classifiers predict an individual to be low in CVD risk, high risk if classifier predicts non-low risk and regressor predicts high risk. Patients with incongruent classifiers and regressor outcomes are considered unclassified.

The ensemble models in both classification and regression phase each fit three base learners (naive bayes (NB), random forest (RF) and support vector classifier (SVC) for classification prediction, and generalized linear regression (GLM), support vector regressor (SVR) and stochastic gradient descent (SGD) for regression prediction). These base learners were chosen based on preliminary analysis, where these models showed efficiency in handling missing values and outliers.

The ensemble model then uses majority vote to determine the class label in classification phase. For the regression phase, the ensemble model averages the normalized predictions from the base regressor models to form a numerical outcome.

All models were trained on a stratified five-fold cross-validation. As SingHEART data had an imbalanced CVD risk distribution of risk based on the calcium score (low risk 63.4%, high risk 8.3%, intermediate risk 18.7%) we oversampled the training set for the minority class labels to allow model to better learn features from the under-represented classes31. The data were first partitioned into five mutually exclusive subsets, with each subset sharing the same proportion of class label as original dataset. At each iteration, the MLAs trained on four parts (80%) and validated on the fifth, the holdout set (20%). The process repeats five times, with five different but overlapping training sets. The resulting metrics from each fold were averaged to produce a single estimate.

To simulate access to the different variable groups as per clinical workflow and ease of information availability, we assessed the performance of individual variable group, and in combination as per the following:

Model 1: Survey Questionnaire.

Model 2: 24h ambulatory blood pressure and heart rate.

Model 3: Clinical blood results.

Model 4: Model 1+Model 2.

Model 5: Model 1+Model 3.

Model 6: Model 1 to Model 3.

Model 6*: Model 1 to Model 3 with feature selection.

Model 7: Physical activity and sleep trackers (exploratory subset analysis).

Variables in model 6* were reduced using SVC recursive feature elimination with cross-validation (SVC-RFECV) method to automatically select the best set of predictors that yield the highest area under Receiver Operating Characteristic curves (AUC). Model 16 were trained using 600 subjects.

We also performed exploratory analysis using MLA on the Fitbit Charge HR data (Model 7). Model 7 was trained on a subset of 430 subjects constrained by availability of valid activity tracking data.

Since no single metric can objectively evaluate the cardiovascular risk prediction, we evaluate the performance of our models at CVD risk class level using a panel of metrics; sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), F1-score and Area under Receiver Operating Characteristic curves (AUC). Overall discriminative ability of the model was described by the area under received operating characteristic curve (ROC). All AUC metrics were accompanied by 95% confidence interval (CI) and standard deviation (SD).

To better understand the relative importanceof different risk factors, we conduct a post-hoc approach to rank the variables by their contribution to CVD risk prediction. Feature importance were obtained from the SVC algorithm where the relative importance was determined by the absolute size of the coefficients in relation to others. All statistical analyses were conducted on Python version 3.7 environment and all MLAs and evaluation metrics were constructed using Scikit-learn libraries.

Continue reading here:
Application of ensemble machine learning algorithms on lifestyle factors and wearables for cardiovascular risk prediction | Scientific Reports -...