Search

The interactions between tumor cells and the tumor microenvironment (TME) dictate therapeutic efficacy of radiation and many systemic therapies in breast cancer. However, to date, there is not a widely available method to reproducibly measure tumor and immune phenotypes for each patient's tumor. Given this unmet clinical need, we applied multiple instance learning (MIL) algorithms to assess activity of ten biologically relevant pathways from the hematoxylin and eosin (H&E) slide of primary breast tumors. We employed different feature extraction approaches and state-of-the-art model architectures. Using binary classification, our models attained area under the receiver operating characteristic (AUROC) scores above 0.70 for nearly all gene expression pathways and on some cases, exceeded 0.80. Attention maps suggest that our trained models recognize biologically relevant spatial patterns of cell sub-populations from H&E. These efforts represent a first step towards developing computational H&E biomarkers that reflect facets of the TME and hold promise for augmenting precision oncology., Comment: Paper accepted at the First Workshop on Imageomics (Imageomics-AAAI-24) - Discovering Biological Knowledge from Images using AI (https://sites.google.com/vt.edu/imageomics-aaai-24/home), held as part of the 38th Annual AAAI Conference on Artificial Intelligence (https://aaai.org/aaai-conference/)

Despite decades of research into query optimization, optimizing queries with disjunctive predicate expressions remains a challenge. Solutions employed by existing systems (if any) are often simplistic and lead to much redundant work being performed by the execution engine. To address these problems, we propose a novel form of query execution called tagged execution. Tagged execution groups tuples into subrelations based on which predicates in the query they satisfy (or don't satisfy) and tags them with that information. These tags then provide additional context for query operators to take advantage of during runtime, allowing them to eliminate much of the redundant work performed by traditional engines and realize predicate pushdown optimizations for disjunctive predicates. However, tagged execution brings its own challenges, and the question of what tags to create is a nontrivial one. Careless creation of tags can lead to an exponential blowup in the tag space, with the overhead outweighing the benefits. To address this issue, we present a technique called tag generalization to minimize the space of tags. We implemented the tagged execution model with tag generalization in our system Basilisk, and our evaluation shows an average 2.7x speedup in runtime over the traditional execution model with up to a 19x speedup in certain situations.

Introduction: With the rapid advances in large language models (LLMs), there have been numerous new open source as well as commercial models. While recent publications have explored GPT-4 in its application to extracting information of interest from radiology reports, there has not been a real-world comparison of GPT-4 to different leading open-source models. Materials and Methods: Two different and independent datasets were used. The first dataset consists of 540 chest x-ray reports that were created at the Massachusetts General Hospital between July 2019 and July 2021. The second dataset consists of 500 chest x-ray reports from the ImaGenome dataset. We then compared the commercial models GPT-3.5 Turbo and GPT-4 from OpenAI to the open-source models Mistral-7B, Mixtral-8x7B, Llama2-13B, Llama2-70B, QWEN1.5-72B and CheXbert and CheXpert-labeler in their ability to accurately label the presence of multiple findings in x-ray text reports using different prompting techniques. Results: On the ImaGenome dataset, the best performing open-source model was Llama2-70B with micro F1-scores of 0.972 and 0.970 for zero- and few-shot prompts, respectively. GPT-4 achieved micro F1-scores of 0.975 and 0.984, respectively. On the institutional dataset, the best performing open-source model was QWEN1.5-72B with micro F1-scores of 0.952 and 0.965 for zero- and few-shot prompting, respectively. GPT-4 achieved micro F1-scores of 0.975 and 0.973, respectively. Conclusion: In this paper, we show that while GPT-4 is superior to open-source models in zero-shot report labeling, the implementation of few-shot prompting can bring open-source models on par with GPT-4. This shows that open-source models could be a performant and privacy preserving alternative to GPT-4 for the task of radiology report classification.

Let $p$ be a prime number and let $S=\{x^p+c_1,\dots,x^p+c_r\}$ be a finite set of unicritical polynomials for some $c_1,\dots,c_r\in\mathbb{Z}$. Moreover, assume that $S$ contains at least one irreducible polynomial over $\mathbb{Q}$. Then we construct a large, explicit subset of irreducible polynomials within the semigroup generated by $S$ under composition; in fact, we show that this subset has positive asymptotic density within the full semigroup when we count polynomials by degree. In addition, when $p=2$ or $3$ we construct an infinite family of semigroups that break the local-global principle for irreducibility. To do this, we use a mix of algebraic and arithmetic techniques and results, including Runge's method, the elliptic curve Chabauty method, and Fermat's Last Theorem.

When linear regression generates a relationship between a (dependent) scalar response and one or multiple independent variables, various datasets providing distinct graphical trends can develop resembling relationships based on the same statistical properties. Advanced statistical approaches, such as neural networks and machine learning methods, are of great necessity to process, characterize, and analyze these degenerate datasets. On the other hand, the accurate creation of purposedly degenerate datasets is essential to test new models in the research and education of applied statistics. In this light, the present study characterizes the famous Anscombe datasets and provides a general algorithm for creating multiple paired datasets of identical statistical properties.

We provide a computational exercise suitable for early introduction in an undergraduate statistics or data science course that allows students to 'play the whole game' of data science: performing both data collection and data analysis. While many teaching resources exist for data analysis, such resources are not as abundant for data collection given the inherent difficulty of the task. Our proposed exercise centers around student use of Google Calendar to collect data with the goal of answering the question 'How do I spend my time?' On the one hand, the exercise involves answering a question with near universal appeal, but on the other hand, the data collection mechanism is not beyond the reach of a typical undergraduate student. A further benefit of the exercise is that it provides an opportunity for discussions on ethical questions and considerations that data providers and data analysts face in today's age of large-scale internet-based data collection.

Database research has always given limited attention to optimizing predicates with disjunctions. What little past work there is, has mostly focused on optimizations for traditional row-oriented databases. However, a key difference between how row-oriented and column-oriented engines evaluate predicates is that while row-oriented engines apply predicates to a single tuple at a time, column-oriented engines apply predicates to sets of tuples, adding another dimension to the problem. As such, row-oriented engines focus only on the best order to apply predicates in to "short-circuit" the overall predicate expression, but column-oriented engines must additionally decide on the input sets of tuples for each predicate application. This is important, since smaller inputs lead to faster runtimes, and nontrivial, since the results of earlier predicates can be used to reduce the inputs to later predicates and predicates may be combined via disjunctions in the predicate expression. In this work, we formally analyze the predicate evaluation problem for column-oriented engines and present BestD/Update, the first ever polynomial-time, provably optimal algorithms to deduce the minimum input sets for each predicate application. BestD/Update's optimality is guaranteed under a wide range of cost models, representing different real-world scenarios. Furthermore, when combined with the predicate ordering algorithm Hanani, BestD/Update reduce into EvalPred, a simple O(n log^2 n) algorithm, which we recommend for practical use and optimal for all predicate expressions of nested depth 2 or less. Our evaluation shows, thanks to its optimality and polynomial planning time, EvalPred outperforms not implementing any disjunction optimizations and exiting optimal algorithms by up to 2.6x and 28x respectively for synthetic workloads and by up to 1.3x and 100x respectively for queries from TPC-H and the CH-benchmark.

We present a programmatic approach to incorporating ethics into an undergraduate major in statistical and data sciences. We discuss departmental-level initiatives designed to meet the National Academy of Sciences recommendation for integrating ethics into the curriculum from top-to-bottom as our majors progress from our introductory courses to our senior capstone course, as well as from side-to-side through co-curricular programming. We also provide six examples of data science ethics modules used in five different courses at our liberal arts college, each focusing on a different ethical consideration. The modules are designed to be portable such that they can be flexibly incorporated into existing courses at different levels of instruction with minimal disruption to syllabi. We connect our efforts to a growing body of literature on the teaching of data science ethics, present assessments of our effectiveness, and conclude with next steps and final thoughts.

The Park City Math Institute (PCMI) 2016 Summer Undergraduate Faculty Program met for the purpose of composing guidelines for undergraduate programs in Data Science. The group consisted of 25 undergraduate faculty from a variety of institutions in the U.S., primarily from the disciplines of mathematics, statistics and computer science. These guidelines are meant to provide some structure for institutions planning for or revising a major in Data Science.

Determining if two sets are related - that is, if they have similar values or if one set contains the other - is an important problem with many applications in data cleaning, data integration, and information retrieval. A particularly popular metric that has been proposed is to measure the relatedness of two sets by treating the elements as vertices of a bipartite graph and calculating the score of the maximum matching pairing between elements. Compared to other metrics which require exact matchings between elements, this metric uses a similarity function to compare elements between the two sets, making it robust to small dissimilarities in elements and more useful for real-world, dirty data. Unfortunately, the metric suffers from expensive computational cost, taking O(n^3) time, where n is the number of elements in sets, for each set-to-set comparison. Thus for applications which try to search for all pairings of related sets in a brute-force manner, the runtime becomes unacceptably large. To address this challenge, we developed SilkMoth, a system capable of rapidly discovering related set pairs in collections of sets. Internally, SilkMoth creates a signature for each set, with the property that any other set which is related must match the signature. SilkMoth then uses these signatures to prune the search space, so only sets which match the signatures are left as candidates. Finally, SilkMoth applies the maximum matching metric on remaining candidates to verify which of these candidates are truly related sets. Thus, a contribution of this paper is the characterization of the space of signatures which enable this property. We show that selecting the optimal signature in this space is NP-complete, and based on insights from the characterization of the space, we propose two novel filters which help to prune the candidates further before verification., Comment: VLDB 2017

Existing database systems are not optimized for queries with a LIMIT clause---operating instead in an all-or-nothing manner. In this paper, we propose a fast LIMIT query evaluation engine, called NeedleTail, aimed at letting analysts browse a small sample of the query results on large datasets as quickly as possible, independent of the overall size of the result set. NeedleTail introduces density maps, a lightweight in-memory indexing structure, and a set of efficient algorithms (with desirable theoretical guarantees) to quickly locate promising blocks, trading off locality and density. In settings where the samples are used to compute aggregates, we extend techniques from survey sampling to mitigate the bias in our samples. Our experimental results demonstrate that NeedleTail returns results 4x faster on HDDs and 9x faster on SSDs on average, while occupying up to 23x less memory than existing techniques.

Data visualization is by far the most commonly used mechanism to explore data, especially by novice data analysts and data scientists. And yet, current visual analytics tools are rather limited in their ability to guide data scientists to interesting or desired visualizations: the process of visual data exploration remains cumbersome and time-consuming. We propose zenvisage, a platform for effortlessly visualizing interesting patterns, trends, or insights from large datasets. We describe zenvisage's general purpose visual query language, ZQL ("zee-quel") for specifying the desired visual trend, pattern, or insight - ZQL draws from use-cases in a variety of domains, including biology, mechanical engineering, climate science, and commerce. We formalize the expressiveness of ZQL via a visual exploration algebra, and demonstrate that ZQL is at least as expressive as that algebra. While analysts are free to use ZQL directly, we also expose ZQL via a visual specification interface that we describe in this paper. We then describe our architecture and optimizations, preliminary experiments in supporting and optimizing for ZQL queries in our initial zenvisage prototype, and a user study to evaluate whether data scientists are able to effectively use zenvisage for real applications., Comment: Tech Report

Visualizations are frequently used as a means to understand trends and gather insights from datasets, but often take a long time to generate. In this paper, we focus on the problem of rapidly generating approximate visualizations while preserving crucial visual proper- ties of interest to analysts. Our primary focus will be on sampling algorithms that preserve the visual property of ordering; our techniques will also apply to some other visual properties. For instance, our algorithms can be used to generate an approximate visualization of a bar chart very rapidly, where the comparisons between any two bars are correct. We formally show that our sampling algorithms are generally applicable and provably optimal in theory, in that they do not take more samples than necessary to generate the visualizations with ordering guarantees. They also work well in practice, correctly ordering output groups while taking orders of magnitude fewer samples and much less time than conventional sampling schemes., Comment: Tech Report. 17 pages. Condensed version to appear in VLDB Vol. 8 No. 5

An extension to classical unification, called {\em graded unification} is presented. It is capable of combining contradictory information. An interactive processing paradigm and parser based on this new operator are also presented., Comment: 3 pages, To appear in ACL-94 Student Session, Postscript file; extract with Unix uudecode and uncompress