More Like this

1. The Temple University Digital Pathology Corpus: The Breast Tissue Subset

   https://doi.org/10.1109/SPMB52430.2021.9672275  

   Wevodau, Z. ; Doshna, B. ; Jhala, N. ; Akhtar, I. ; Obeid, I. ; Picone, J. ( December 2021 , Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, I. (Ed.)

   The Neural Engineering Data Consortium (NEDC) is developing the Temple University Digital Pathology Corpus (TUDP), an open source database of high-resolution images from scanned pathology samples [1], as part of its National Science Foundation-funded Major Research Instrumentation grant titled “MRI: High Performance Digital Pathology Using Big Data and Machine Learning” [2]. The long-term goal of this project is to release one million images. We have currently scanned over 100,000 images and are in the process of annotating breast tissue data for our first official corpus release, v1.0.0. This release contains 3,505 annotated images of breast tissue including 74 patients with cancerous diagnoses (out of a total of 296 patients). In this poster, we will present an analysis of this corpus and discuss the challenges we have faced in efficiently producing high quality annotations of breast tissue. It is well known that state of the art algorithms in machine learning require vast amounts of data. Fields such as speech recognition [3], image recognition [4] and text processing [5] are able to deliver impressive performance with complex deep learning models because they have developed large corpora to support training of extremely high-dimensional models (e.g., billions of parameters). Other fields that do notmore »have access to such data resources must rely on techniques in which existing models can be adapted to new datasets [6]. A preliminary version of this breast corpus release was tested in a pilot study using a baseline machine learning system, ResNet18 [7], that leverages several open-source Python tools. The pilot corpus was divided into three sets: train, development, and evaluation. Portions of these slides were manually annotated [1] using the nine labels in Table 1 [8] to identify five to ten examples of pathological features on each slide. Not every pathological feature is annotated, meaning excluded areas can include focuses particular to these labels that are not used for training. A summary of the number of patches within each label is given in Table 2. To maintain a balanced training set, 1,000 patches of each label were used to train the machine learning model. Throughout all sets, only annotated patches were involved in model development. The performance of this model in identifying all the patches in the evaluation set can be seen in the confusion matrix of classification accuracy in Table 3. The highest performing labels were background, 97% correct identification, and artifact, 76% correct identification. A correlation exists between labels with more than 6,000 development patches and accurate performance on the evaluation set. Additionally, these results indicated a need to further refine the annotation of invasive ductal carcinoma (“indc”), inflammation (“infl”), nonneoplastic features (“nneo”), normal (“norm”) and suspicious (“susp”). This pilot experiment motivated changes to the corpus that will be discussed in detail in this poster presentation. To increase the accuracy of the machine learning model, we modified how we addressed underperforming labels. One common source of error arose with how non-background labels were converted into patches. Large areas of background within other labels were isolated within a patch resulting in connective tissue misrepresenting a non-background label. In response, the annotation overlay margins were revised to exclude benign connective tissue in non-background labels. Corresponding patient reports and supporting immunohistochemical stains further guided annotation reviews. The microscopic diagnoses given by the primary pathologist in these reports detail the pathological findings within each tissue site, but not within each specific slide. The microscopic diagnoses informed revisions specifically targeting annotated regions classified as cancerous, ensuring that the labels “indc” and “dcis” were used only in situations where a micropathologist diagnosed it as such. Further differentiation of cancerous and precancerous labels, as well as the location of their focus on a slide, could be accomplished with supplemental immunohistochemically (IHC) stained slides. When distinguishing whether a focus is a nonneoplastic feature versus a cancerous growth, pathologists employ antigen targeting stains to the tissue in question to confirm the diagnosis. For example, a nonneoplastic feature of usual ductal hyperplasia will display diffuse staining for cytokeratin 5 (CK5) and no diffuse staining for estrogen receptor (ER), while a cancerous growth of ductal carcinoma in situ will have negative or focally positive staining for CK5 and diffuse staining for ER [9]. Many tissue samples contain cancerous and non-cancerous features with morphological overlaps that cause variability between annotators. The informative fields IHC slides provide could play an integral role in machine model pathology diagnostics. Following the revisions made on all the annotations, a second experiment was run using ResNet18. Compared to the pilot study, an increase of model prediction accuracy was seen for the labels indc, infl, nneo, norm, and null. This increase is correlated with an increase in annotated area and annotation accuracy. Model performance in identifying the suspicious label decreased by 25% due to the decrease of 57% in the total annotated area described by this label. A summary of the model performance is given in Table 4, which shows the new prediction accuracy and the absolute change in error rate compared to Table 3. The breast tissue subset we are developing includes 3,505 annotated breast pathology slides from 296 patients. The average size of a scanned SVS file is 363 MB. The annotations are stored in an XML format. A CSV version of the annotation file is also available which provides a flat, or simple, annotation that is easy for machine learning researchers to access and interface to their systems. Each patient is identified by an anonymized medical reference number. Within each patient’s directory, one or more sessions are identified, also anonymized to the first of the month in which the sample was taken. These sessions are broken into groupings of tissue taken on that date (in this case, breast tissue). A deidentified patient report stored as a flat text file is also available. Within these slides there are a total of 16,971 total annotated regions with an average of 4.84 annotations per slide. Among those annotations, 8,035 are non-cancerous (normal, background, null, and artifact,) 6,222 are carcinogenic signs (inflammation, nonneoplastic and suspicious,) and 2,714 are cancerous labels (ductal carcinoma in situ and invasive ductal carcinoma in situ.) The individual patients are split up into three sets: train, development, and evaluation. Of the 74 cancerous patients, 20 were allotted for both the development and evaluation sets, while the remain 34 were allotted for train. The remaining 222 patients were split up to preserve the overall distribution of labels within the corpus. This was done in hope of creating control sets for comparable studies. Overall, the development and evaluation sets each have 80 patients, while the training set has 136 patients. In a related component of this project, slides from the Fox Chase Cancer Center (FCCC) Biosample Repository (https://www.foxchase.org/research/facilities/genetic-research-facilities/biosample-repository -facility) are being digitized in addition to slides provided by Temple University Hospital. This data includes 18 different types of tissue including approximately 38.5% urinary tissue and 16.5% gynecological tissue. These slides and the metadata provided with them are already anonymized and include diagnoses in a spreadsheet with sample and patient ID. We plan to release over 13,000 unannotated slides from the FCCC Corpus simultaneously with v1.0.0 of TUDP. Details of this release will also be discussed in this poster. Few digitally annotated databases of pathology samples like TUDP exist due to the extensive data collection and processing required. The breast corpus subset should be released by November 2021. By December 2021 we should also release the unannotated FCCC data. We are currently annotating urinary tract data as well. We expect to release about 5,600 processed TUH slides in this subset. We have an additional 53,000 unprocessed TUH slides digitized. Corpora of this size will stimulate the development of a new generation of deep learning technology. In clinical settings where resources are limited, an assistive diagnoses model could support pathologists’ workload and even help prioritize suspected cancerous cases. ACKNOWLEDGMENTS This material is supported by the National Science Foundation under grants nos. CNS-1726188 and 1925494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. REFERENCES [1] N. Shawki et al., “The Temple University Digital Pathology Corpus,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York City, New York, USA: Springer, 2020, pp. 67 104. https://www.springer.com/gp/book/9783030368432. [2] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning.” Major Research Instrumentation (MRI), Division of Computer and Network Systems, Award No. 1726188, January 1, 2018 – December 31, 2021. https://www. isip.piconepress.com/projects/nsf_dpath/. [3] A. Gulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH), 2020, pp. 5036-5040. https://doi.org/10.21437/interspeech.2020-3015. [4] C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in Proceedings of the IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. https://ieeexplore.ieee.org/document/8675201. [5] I. Caswell and B. Liang, “Recent Advances in Google Translate,” Google AI Blog: The latest from Google Research, 2020. [Online]. Available: https://ai.googleblog.com/2020/06/recent-advances-in-google-translate.html. [Accessed: 01-Aug-2021]. [6] V. Khalkhali, N. Shawki, V. Shah, M. Golmohammadi, I. Obeid, and J. Picone, “Low Latency Real-Time Seizure Detection Using Transfer Deep Learning,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2021, pp. 1 7. https://www.isip. piconepress.com/publications/conference_proceedings/2021/ieee_spmb/eeg_transfer_learning/. [7] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning,” Philadelphia, Pennsylvania, USA, 2020. https://www.isip.piconepress.com/publications/reports/2020/nsf/mri_dpath/. [8] I. Hunt, S. Husain, J. Simons, I. Obeid, and J. Picone, “Recent Advances in the Temple University Digital Pathology Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2019, pp. 1–4. https://ieeexplore.ieee.org/document/9037859. [9] A. P. Martinez, C. Cohen, K. Z. Hanley, and X. (Bill) Li, “Estrogen Receptor and Cytokeratin 5 Are Reliable Markers to Separate Usual Ductal Hyperplasia From Atypical Ductal Hyperplasia and Low-Grade Ductal Carcinoma In Situ,” Arch. Pathol. Lab. Med., vol. 140, no. 7, pp. 686–689, Apr. 2016. https://doi.org/10.5858/arpa.2015-0238-OA.« less
2. Issues in the Reproducibility of Deep Learning Results

   https://doi.org/10.1109/SPMB47826.2019.9037840  

   Jean-Paul, S. ; Elseify, T. ; Obeid, I. ; Picone, J. ( December 2019 , IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)

   The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should producemore »identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues.« less
3. Study on U.S. Parents' Divisions of Labor During COVID-19, Waves 1-2

   https://doi.org/10.3886/E183142V2  

   Carlson, Daniel L. ; Petts, Richard J. ( January 2022 )

   Abstract

   The COVID-19 pandemic has dramatically altered family life in the United States. Over the long duration of the pandemic, parents had to adapt to shifting work conditions, virtual schooling, the closure of daycare facilities, and the stress of not only managing households without domestic and care supports but also worrying that family members may contract the novel coronavirus. Reports early in the pandemic suggest that these burdens have fallen disproportionately on mothers, creating concerns about the long-term implications of the pandemic for gender inequality and mothers’ well-being. Nevertheless, less is known about how parents’ engagement in domestic labor and paid work has changed throughout the pandemic, what factors may be driving these changes, and what the long-term consequences of the pandemic may be for the gendered division of labor and gender inequality more generally. <br /><br />The Study on U.S. Parents’ Divisions of Labor During COVID-19 (SPDLC) collects longitudinal survey data from partnered U.S. parents that can be used to assess changes in parents’ divisions of domestic labor, divisions of paid labor, and well-being throughout and after the COVID-19 pandemic. The goal of SPDLC is to understand both the short- and long-term impacts of the pandemic for the gendered division of labor, work-family issues, and broader patterns of gender inequality. <br /><br />Survey data for this study is collected using Prolifc (www.prolific.co), an opt-in online platform designed to facilitate scientific research. The sample is comprised U.S. adults who were residing with a romantic partner and at least one biological child (at the time of entry into the study). In each survey, parents answer questions about both themselves and their partners. Wave 1 of SPDLC was conducted in April 2020, and parents who participated in Wave 1 were asked about their division of labor both prior to (i.e., early March 2020) and one month after the pandemic began. Wave 2 of SPDLC was collected in November 2020. Parents who participated in Wave 1 were invited to participate again in Wave 2, and a new cohort of parents was also recruited to participate in the Wave 2 survey. Wave 3 of SPDLC was collected in October 2021. Parents who participated in either of the first two waves were invited to participate again in Wave 3, and another new cohort of parents was also recruited to participate in the Wave 3 survey. This research design (follow-up survey of panelists and new cross-section of parents at each wave) will continue through 2024, culminating in six waves of data spanning the period from March 2020 through October 2024. An estimated total of approximately 6,500 parents will be surveyed at least once throughout the duration of the study. <br /><br />SPDLC data will be released to the public two years after data is collected; Waves 1 and 2 are currently publicly available. Wave 3 will be publicly available in October 2023, with subsequent waves becoming available yearly. Data will be available to download in both SPSS (.sav) and Stata (.dta) formats, and the following data files will be available: (1) a data file for each individual wave, which contains responses from all participants in that wave of data collection, (2) a longitudinal panel data file, which contains longitudinal follow-up data from all available waves, and (3) a repeated cross-section data file, which contains the repeated cross-section data (from new respondents at each wave) from all available waves. Codebooks for each survey wave and a detailed user guide describing the data are also available.
   

   Methods

   Response Rates: Of the 1,157 parents who participated in Wave 1, 828 (72%) also participated in the Wave 2 study. Presence of Common Scales: The following established scales are included in the survey:<br /><ul><li><b>Self-Efficacy</b>, adapted from Pearlin&#39;s mastery scale (Pearlin et al., 1981) and the Rosenberg self-esteem scale (Rosenberg, 2015) and taken from the American Changing Lives Survey</li><li><b>Communication with Partner</b>, taken from the Marriage and Relationship Survey (Lichter &amp; Carmalt, 2009)<br /></li><li><b>Gender Attitudes</b>, taken from the National Survey of Families and Households (Sweet &amp; Bumpass, 1996)<br /></li><li><b>Depressive Symptoms </b>(CES-D-10)</li><li><b>Stress</b>, measured using Cohen&#39;s Perceived Stress Scale (Cohen, Kamarck, &amp; Mermelstein, 1983)<br /></li></ul>Full details about these scales and all other items included in the survey can be found in the user guide and codebook<br /> The second wave of the SPDLC was fielded in November 2020 in two stages. In the first stage, all parents who participated in W1 of the SPDLC and who continued to reside in the United States were re-contacted and asked to participate in a follow-up survey. The W2 survey was posted on Prolific, and messages were sent via Prolific’s messaging system to all previous participants. Multiple follow-up messages were sent in an attempt to increase response rates to the follow-up survey. Of the 1,157 respondents who completed the W1 survey, 873 at least started the W2 survey. Data quality checks were employed in line with best practices for online surveys (e.g., removing respondents who did not complete most of the survey or who did not pass the attention filters). After data quality checks, 5.2% of respondents were removed from the sample, resulting in a final sample size of 828 parents (a response rate of 72%). <br /><br />In the second stage, a new sample of parents was recruited. New parents had to meet the same sampling criteria as in W1 (be at least 18 years old, reside in the United States, reside with a romantic partner, and be a parent living with at least one biological child). Also similar to the W1 procedures, we oversampled men, Black individuals, individuals who did not complete college, and individuals who identified as politically conservative to increase sample diversity. A total of 1,207 parents participated in the W2 survey. Data quality checks led to the removal of 5.7% of the respondents, resulting in a final sample size of new respondents at Wave 2 of 1,138 parents. <br /><br />In both stages, participants were informed that the survey would take approximately 20 minutes to complete. All panelists were provided monetary compensation in line with Prolific’s compensation guidelines, which require that all participants earn above minimum wage for their time participating in studies. <br /> To be included in SPDLC, respondents had to meet the following sampling criteria at the time they enter the study: (a) be at least 18 years old, (b) reside in the United States, (c) reside with a romantic partner (i.e., be married or cohabiting), and (d) be a parent living with at least one biological child. Follow-up respondents must be at least 18 years old and reside in the United States, but may experience changes in relationship and resident parent statuses. Smallest Geographic Unit: U.S. State <p>This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. In accordance with this license, all users of these data must give appropriate credit to the authors in any papers, presentations, books, or other works that use the data. A suggested citation to provide attribution for these data is included below:             <br /><br />Carlson, Daniel L. and Richard J. Petts. 2022. Study on U.S. Parents’ Divisions of Labor During COVID-19 User Guide: Waves 1-2.  </p>
   

   Other

   To help provide estimates that are more representative of U.S. partnered parents, the SPDLC includes sampling weights. Weights can be included in statistical analyses to make estimates from the SPDLC sample representative of U.S. parents who reside with a romantic partner (married or cohabiting) and a child aged 18 or younger based on age, race/ethnicity, and gender. National estimates for the age, racial/ethnic, and gender profile of U.S. partnered parents were obtained using data from the 2020 Current Population Survey (CPS). Weights were calculated using an iterative raking method, such that the full sample in each data file matches the nationally representative CPS data in regard to the gender, age, and racial/ethnic distributions within the data. This variable is labeled <b>CPSweightW2</b> in the Wave 2 dataset, and <b>CPSweightLW2</b> in the longitudinal dataset (which includes Waves 1 and 2). There is not a weight variable included in the W1-W2 repeated cross-section data file.<br />
   More>>
4. QoS-Aware Placement of Deep Learning Services on the Edge with Multiple Service Implementations

   https://doi.org/10.1109/ICCCN52240.2021.9522156  

   Hudson, Nathaniel ; Khamfroush, Hana ; Lucani, Daniel E. ( July 2021 , IEEE ICCCN Big Data and Machine Learning for Networking (BDMLN) Workshop • 2021)

   Mobile edge computing pushes computationally-intensive services closer to the user to provide reduced delay due to physical proximity. This has led many to consider deploying deep learning models on the edge – commonly known as edge intelligence (EI). EI services can have many model implementations that provide different QoS. For instance, one model can perform inference faster than another (thus reducing latency) while achieving less accuracy when evaluated. In this paper, we study joint service placement and model scheduling of EI services with the goal to maximize Quality-of-Servcice (QoS) for end users where EI services have multiple implementations to serve user requests, each with varying costs and QoS benefits. We cast the problem as an integer linear program and prove that it is NP-hard. We then prove the objective is equivalent to maximizing a monotone increasing, submodular set function and thus can be solved greedily while maintaining a (1 – 1/e)-approximation guarantee. We then propose two greedy algorithms: one that theoretically guarantees this approximation and another that empirically matches its performance with greater efficiency. Finally, we thoroughly evaluate the proposed algorithm for making placement and scheduling decisions in both synthetic and real-world scenarios against the optimal solution and some baselines.more »In the real-world case, we consider real machine learning models using the ImageNet 2012 data-set for requests. Our numerical experiments empirically show that our more efficient greedy algorithm is able to approximate the optimal solution with a 0.904 approximation on average, while the next closest baseline achieves a 0.607 approximation on average.« less
5. Building and steering binned template fits with cabinetry

   https://doi.org/10.1051/epjconf/202125103067  

   Cranmer, Kyle ; Held, Alexander ( January 2021 , EPJ Web of Conferences)

   Biscarat, C. ; Campana, S. ; Hegner, B. ; Roiser, S. ; Rovelli, C.I. ; Stewart, G.A. (Ed.)

   The cabinetry library provides a Python-based solution for building and steering binned template fits. It tightly integrates with the pythonic High Energy Physics ecosystem, and in particular with pyhf for statistical inference. cabinetry uses a declarative approach for building statistical models, with a JSON schema describing possible configuration choices. Model building instructions can additionally be provided via custom code, which is automatically executed when applicable at key steps of the workflow. The library implements interfaces for performing maximum likelihood fitting, upper parameter limit determination, and discovery significance calculation. cabinetry also provides a range of utilities to study and disseminate fit results. These include visualizations of the fit model and data, visualizations of template histograms and fit results, ranking of nuisance parameters by their impact, a goodness-of-fit calculation, and likelihood scans. The library takes a modular approach, allowing users to include some or all of its functionality in their workflow.