Deep learning research in medicine is a bit like the Wild West at the moment; sometimes you find gold, sometimes a giant steampunk spider-bot causes a ruckus. This has derailed my series on whether AI will be replacing doctors soon, as I have felt the need to focus a bit more on how to assess the quality of medical AI research. I wanted to start closing out my series on the role of AI in medicine. What has happened instead is that several papers have claimed to beat doctors, and have failed to justify these claims. Despite this, and despite not going through peer review, the groups involved have issued press releases about their achievements, marketing the results direct to the public and the media. I don't think this is malicious. I think there is a cultural divide between the machine learning and the medical communities, a different way of doing research, and a different level of evidence required for making strong claims. If I have to be honest, I think the machine learning community has a fair bit to learn from medical research in this regard. Last time I made a set of three rules about how to assess medical AI research, and the third was a glib recommendation to "actually read the paper".

One of the things I am trying to do this year is some more technical posts (following up on some issues I have noticed at the intersection between medicine and machine learning). This is the first in a little mini-series on performance testing. Medical research has a different way of doing things, being more cautious about making claims and a bit more rigorous in justifying them, both of which are useful ideas to apply more broadly in machine learning (particularly at the applied end). While performance testing is often considered basic knowledge, one of my supervisors/colleagues is a bit of a ROC expert so I hope I can pass on some new ways of looking at things that are interesting even for some of the more knowledgeable folks around here.

In this webinar, Jan 11, DataRobot will show how automated machine learning can be used to reduce false positive rates, thereby improving the efficiency of AML transaction monitoring and reducing costs.

We can see that Ridge and Lasso is performing far better than Linear Regression when the correlation exists in the dataset. We can tune our penalty parameter further and try to find best value of RMSE. Here, 0.01 is the best value I got for lambda. So, we can use these regression methods when the variables are highly correlated. Hope this article was useful in understanding Bias-Variance trade-off, Lasso and Ridge Regression. Please feel free to comment, give feedback and share this article if you found it useful. You can find the original article on my blog here.

Microscopic interpretation of stained smears is one of the most operator-dependent and time intensive activities in the clinical microbiology laboratory. Here, we investigated application of an automated image acquisition and convolutional neural network (CNN)-based approach for automated Gram stain classification. Using an automated microscopy platform, uncoverslipped slides were scanned with a 40x dry objective, generating images of sufficient resolution for interpretation. We collected 25,488 images from positive blood culture Gram stains prepared during routine clinical workup. These images were used to generate 100,213 crops containing Gram-positive cocci in clusters, Gram-positive cocci in chains/pairs, Gram-negative rods, or background (no cells).

It has been proposed that complex populations, such as those that arise in genomics studies, may exhibit dependencies among observations as well as among variables. This gives rise to the challenging problem of analyzing unreplicated high-dimensional data with unknown mean and dependence structures. Matrix-variate approaches that impose various forms of (inverse) covariance sparsity allow flexible dependence structures to be estimated, but cannot directly be applied when the mean and covariance matrices are estimated jointly. We present a practical method utilizing generalized least squares and penalized (inverse) covariance estimation to address this challenge. We establish consistency and obtain rates of convergence for estimating the mean parameters and covariance matrices. The advantages of our approaches are: (i) dependence graphs and covariance structures can be estimated in the presence of unknown mean structure, (ii) the mean structure becomes more efficiently estimated when accounting for the dependence structure among observations; and (iii) inferences about the mean parameters become correctly calibrated. We use simulation studies and analysis of genomic data from a twin study of ulcerative colitis to illustrate the statistical convergence and the performance of our methods in practical settings. Several lines of evidence show that the test statistics for differential gene expression produced by our methods are correctly calibrated and improve power over conventional methods.

Languages shared by people differ in different regions based on their accents, pronunciation and word usages. In this era sharing of language takes place mainly through social media and blogs. Every second swing of such a micro posts exist which induces the need of processing those micro posts, in-order to extract knowledge out of it. Knowledge extraction differs with respect to the application in which the research on cognitive science fed the necessities for the same. This work further moves forward such a research by extracting semantic information of streaming and batch data in applications like Named Entity Recognition and Author Profiling. In the case of Named Entity Recognition context of a single micro post has been utilized and context that lies in the pool of micro posts were utilized to identify the sociolect aspects of the author of those micro posts. In this work Conditional Random Field has been utilized to do the entity recognition and a novel approach has been proposed to find the sociolect aspects of the author (Gender, Age group).

Zen monks have been using a tool called a'koan' for hundreds of years to assist them in reaching enlightenment. These koans are like riddles or stories that can only be solved by letting go of ones narrowing believes and stories about how things should be. Zen students sit in silent meditation and observe how the koan is working on them, slowly transforming their way of looking at the world and revealing a tiny piece of the path to nirvana, that place of no suffering. You might wonder what that has to do with cyber security. With the increased popularity of deep learning and the omni presence of the term artificial intelligence (AI), a lot of security practitioners are tricked into believing that these approaches are the magic silver bullet we have been waiting for to solve all of our cyber security challenges.

Hospitals and other healthcare providers continue to be prime targets for hackers, and as providers increasingly depend on the constant flow of information through their systems, the potential negative effects of attacks loom larger in 2018. In addition to the substantial financial costs associated with cyberattacks, security breaches frequently disrupt critical medical services--a major problem when hospitals get attacked is that patient record systems can go down or become encrypted. When that happens, doctors are forced to revert to taking and circulating patient notes by hand. They may not be able to access important information regarding patient history, medications, test results and more. With the threat of and severity of attacks becoming more damaging than ever, healthcare organizations are increasingly looking for new security solutions that can better protect them.

The increasing size and complexity of scientific data could dramatically enhance discovery and prediction for basic scientific applications, e.g., neuroscience, genetics, systems biology, etc. Realizing this potential, however, requires novel statistical analysis methods that are both interpretable and predictive. We introduce the Union of Intersections (UoI) method, a flexible, modular, and scalable framework for enhanced model selection and estimation. The method performs model selection and model estimation through intersection and union operations, respectively. We show that UoI can satisfy the bi-criteria of low-variance and nearly unbiased estimation of a small number of interpretable features, while maintaining high-quality prediction accuracy. We perform extensive numerical investigation to evaluate a UoI algorithm ($UoI_{Lasso}$) on synthetic and real data. In doing so, we demonstrate the extraction of interpretable functional networks from human electrophysiology recordings as well as the accurate prediction of phenotypes from genotype-phenotype data with reduced features. We also show (with the $UoI_{L1Logistic}$ and $UoI_{CUR}$ variants of the basic framework) improved prediction parsimony for classification and matrix factorization on several benchmark biomedical data sets. These results suggest that methods based on UoI framework could improve interpretation and prediction in data-driven discovery across scientific fields.