More Like this

1. 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 produce 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. 

   more » « less
2. The Centrality of Black Identity for Black Students in Engineering

   Mobley, Catherine ; Brawner, Catherine E. ; Brent, Rebecca ; Orr, Marisa K. ( January 2021 , Collaborative Network for Engineering and Computing Diversity (CoNECD) Conference)

   Introduction and Theoretical Frameworks Our study draws upon several theoretical foundations to investigate and explain the educational experiences of Black students majoring in ME, CpE, and EE: intersectionality, critical race theory, and community cultural wealth theory. Intersectionality explains how gender operates together with race, not independently, to produce multiple, overlapping forms of discrimination and social inequality (Crenshaw, 1989; Collins, 2013). Critical race theory recognizes the unique experiences of marginalized groups and strives to identify the micro- and macro-institutional sources of discrimination and prejudice (Delgado & Stefancic, 2001). Community cultural wealth integrates an asset-based perspective to our analysis of engineering education to assist in the identification of factors that contribute to the success of engineering students (Yosso, 2005). These three theoretical frameworks are buttressed by our use of Racial Identity Theory, which expands understanding about the significance and meaning associated with students’ sense of group membership. Sellers and colleagues (1997) introduced the Multidimensional Model of Racial Identity (MMRI), in which they indicated that racial identity refers to the “significance and meaning that African Americans place on race in defining themselves” (p. 19). The development of this model was based on the reality that individuals vary greatly in the extent to which they attach meaning to being a member of the Black racial group. Sellers et al. (1997) posited that there are four components of racial identity: 1. Racial salience: “the extent to which one’s race is a relevant part of one’s self-concept at a particular moment or in a particular situation” (p. 24). 2. Racial centrality: “the extent to which a person normatively defines himself or herself with regard to race” (p. 25). 3. Racial regard: “a person’s affective or evaluative judgment of his or her race in terms of positive-negative valence” (p. 26). This element consists of public regard and private regard. 4. Racial ideology: “composed of the individual’s beliefs, opinions and attitudes with respect to the way he or she feels that the members of the race should act” (p. 27). The resulting 56-item inventory, the Multidimensional Inventory of Black Identity (MIBI), provides a robust measure of Black identity that can be used across multiple contexts. Research Questions Our 3-year, mixed-method study of Black students in computer (CpE), electrical (EE) and mechanical engineering (ME) aims to identify institutional policies and practices that contribute to the retention and attrition of Black students in electrical, computer, and mechanical engineering. Our four study institutions include historically Black institutions as well as predominantly white institutions, all of which are in the top 15 nationally in the number of Black engineering graduates. We are using a transformative mixed-methods design to answer the following overarching research questions: 1. Why do Black men and women choose and persist in, or leave, EE, CpE, and ME? 2. What are the academic trajectories of Black men and women in EE, CpE, and ME? 3. In what way do these pathways vary by gender or institution? 4. What institutional policies and practices promote greater retention of Black engineering students? Methods This study of Black students in CpE, EE, and ME reports initial results from in-depth interviews at one HBCU and one PWI. We asked students about a variety of topics, including their sense of belonging on campus and in the major, experiences with discrimination, the impact of race on their experiences, and experiences with microaggressions. For this paper, we draw on two methodological approaches that allowed us to move beyond a traditional, linear approach to in-depth interviews, allowing for more diverse experiences and narratives to emerge. First, we used an identity circle to gain a better understanding of the relative importance to the participants of racial identity, as compared to other identities. The identity circle is a series of three concentric circles, surrounding an “inner core” representing one’s “core self.” Participants were asked to place various identities from a provided list that included demographic, family-related, and school-related identities on the identity circle to reflect the relative importance of the different identities to participants’ current engineering education experiences. Second, participants were asked to complete an 8-item survey which measured the “centrality” of racial identity as defined by Sellers et al. (1997). Following Enders’ (2018) reflection on the MMRI and Nigrescence Theory, we chose to use the measure of racial centrality as it is generally more stable across situations and best “describes the place race holds in the hierarchy of identities an individual possesses and answers the question ‘How important is race to me in my life?’” (p. 518). Participants completed the MIBI items at the end of the interview to allow us to learn more about the participants’ identification with their racial group, to avoid biasing their responses to the Identity Circle, and to avoid potentially creating a stereotype threat at the beginning of the interview. This paper focuses on the results of the MIBI survey and the identity circles to investigate whether these measures were correlated. Recognizing that Blackness (race) is not monolithic, we were interested in knowing the extent to which the participants considered their Black identity as central to their engineering education experiences. Combined with discussion about the identity circles, this approach allowed us to learn more about how other elements of identity may shape the participants’ educational experiences and outcomes and revealed possible differences in how participants may enact various points of their identity. Findings For this paper, we focus on the results for five HBCU students and 27 PWI students who completed the MIBI and identity circle. The overall MIBI average for HBCU students was 43 (out of a possible 56) and the overall MIBI scores ranged from 36-51; the overall MIBI average for the PWI students was 40; the overall MIBI scores for the PWI students ranged from 24-51. Twenty-one students placed race in the inner circle, indicating that race was central to their identity. Five placed race on the second, middle circle; three placed race on the third, outer circle. Three students did not place race on their identity circle. For our cross-case qualitative analysis, we will choose cases across the two institutions that represent low, medium and high MIBI scores and different ranges of centrality of race to identity, as expressed in the identity circles. Our final analysis will include descriptive quotes from these in-depth interviews to further elucidate the significance of race to the participants’ identities and engineering education experiences. The results will provide context for our larger study of a total of 60 Black students in engineering at our four study institutions. Theoretically, our study represents a new application of Racial Identity Theory and will provide a unique opportunity to apply the theories of intersectionality, critical race theory, and community cultural wealth theory. Methodologically, our findings provide insights into the utility of combining our two qualitative research tools, the MIBI centrality scale and the identity circle, to better understand the influence of race on the education experiences of Black students in engineering. 

   more » « less
3. Multi-Tenant In-memory Key-Value Cache Partitioning Using Efficient Random Sampling-Based LRU Model

   https://doi.org/10.1109/TCC.2023.3300889  

   Wang, Yuchen ; Yang, Junyao ; Wang, Zhenlin ( August 2023 , IEEE Transactions on Cloud Computing)

   In-memory key-value caches are widely used as a performance-critical layer in web applications, disk-based storage, and distributed systems. The Least Recently Used (LRU) replacement policy has become the de facto standard in those systems since it exploits workload locality well. However, the LRU implementation can be costly due to the rigid data structure in maintaining object priority, as well as the locks for object order updating. Redis as one of the most effective and prevalent deployed commercial systems adopts an approximated LRU policy, where the least recently used item from a small, randomly sampled set of items is chosen to evict. This random sampling-based policy is lightweight and shows its flexibility. We observe that there can exist a significant miss ratio gap between exact LRU and random sampling-based LRU under different sampling size $K$ s. Therefore existing LRU miss ratio curve (MRC) construction techniques cannot be directly applied without loss of accuracy. In this paper, we introduce a new probabilistic stack algorithm named KRR to accurately model random sampling based-LRU, and extend it to handle both fixed and variable objects in key-value caches. We present an efficient stack update algorithm that reduces the expected running time of KRR significantly. To improve the performance of the in-memory multi-tenant key-value cache that utilizes random sampling-based replacement, we propose kRedis, a reference locality- and latency-aware memory partitioning scheme. kRedis guides the memory allocation among the tenants and dynamically customizes $K$ to better exploit the locality of each individual tenant. Evaluation results over diverse workloads show that our model generates accurate miss ratio curves for both fixed and variable object size workloads, and enables practical, low-overhead online MRC prediction. Equipped with KRR, kRedis delivers up to a 50.2% average access latency reduction, and up to a 262.8% throughput improvement compared to Redis. Furthermore, by comparing with pRedis, a state-of-the-art design of memory allocation in Redis, kRedis shows up to 24.8% and 61.8% improvements in average access latency and throughput, respectively. 

   more » « less
4. Predictive performance of multi-model ensemble forecasts of COVID-19 across European nations

   https://doi.org/10.7554/eLife.81916  

   Sherratt, Katharine ; Gruson, Hugo ; Grah, Rok ; Johnson, Helen ; Niehus, Rene ; Prasse, Bastian ; Sandmann, Frank ; Deuschel, Jannik ; Wolffram, Daniel ; Abbott, Sam ; et al ( April 2023 , eLife)

   Short-term forecasts of infectious disease burden can contribute to situational awareness and aid capacity planning. Based on best practice in other fields and recent insights in infectious disease epidemiology, one can maximise the predictive performance of such forecasts if multiple models are combined into an ensemble. Here, we report on the performance of ensembles in predicting COVID-19 cases and deaths across Europe between 08 March 2021 and 07 March 2022.

   We used open-source tools to develop a public European COVID-19 Forecast Hub. We invited groups globally to contribute weekly forecasts for COVID-19 cases and deaths reported by a standardised source for 32 countries over the next 1–4 weeks. Teams submitted forecasts from March 2021 using standardised quantiles of the predictive distribution. Each week we created an ensemble forecast, where each predictive quantile was calculated as the equally-weighted average (initially the mean and then from 26th July the median) of all individual models’ predictive quantiles. We measured the performance of each model using the relative Weighted Interval Score (WIS), comparing models’ forecast accuracy relative to all other models. We retrospectively explored alternative methods for ensemble forecasts, including weighted averages based on models’ past predictive performance.

   Over 52 weeks, we collected forecasts from 48 unique models. We evaluated 29 models’ forecast scores in comparison to the ensemble model. We found a weekly ensemble had a consistently strong performance across countries over time. Across all horizons and locations, the ensemble performed better on relative WIS than 83% of participating models’ forecasts of incident cases (with a total N=886 predictions from 23 unique models), and 91% of participating models’ forecasts of deaths (N=763 predictions from 20 models). Across a 1–4 week time horizon, ensemble performance declined with longer forecast periods when forecasting cases, but remained stable over 4 weeks for incident death forecasts. In every forecast across 32 countries, the ensemble outperformed most contributing models when forecasting either cases or deaths, frequently outperforming all of its individual component models. Among several choices of ensemble methods we found that the most influential and best choice was to use a median average of models instead of using the mean, regardless of methods of weighting component forecast models.

   Our results support the use of combining forecasts from individual models into an ensemble in order to improve predictive performance across epidemiological targets and populations during infectious disease epidemics. Our findings further suggest that median ensemble methods yield better predictive performance more than ones based on means. Our findings also highlight that forecast consumers should place more weight on incident death forecasts than incident case forecasts at forecast horizons greater than 2 weeks.

   AA, BH, BL, LWa, MMa, PP, SV funded by National Institutes of Health (NIH) Grant 1R01GM109718, NSF BIG DATA Grant IIS-1633028, NSF Grant No.: OAC-1916805, NSF Expeditions in Computing Grant CCF-1918656, CCF-1917819, NSF RAPID CNS-2028004, NSF RAPID OAC-2027541, US Centers for Disease Control and Prevention 75D30119C05935, a grant from Google, University of Virginia Strategic Investment Fund award number SIF160, Defense Threat Reduction Agency (DTRA) under Contract No. HDTRA1-19-D-0007, and respectively Virginia Dept of Health Grant VDH-21-501-0141, VDH-21-501-0143, VDH-21-501-0147, VDH-21-501-0145, VDH-21-501-0146, VDH-21-501-0142, VDH-21-501-0148. AF, AMa, GL funded by SMIGE - Modelli statistici inferenziali per governare l'epidemia, FISR 2020-Covid-19 I Fase, FISR2020IP-00156, Codice Progetto: PRJ-0695. AM, BK, FD, FR, JK, JN, JZ, KN, MG, MR, MS, RB funded by Ministry of Science and Higher Education of Poland with grant 28/WFSN/2021 to the University of Warsaw. BRe, CPe, JLAz funded by Ministerio de Sanidad/ISCIII. BT, PG funded by PERISCOPE European H2020 project, contract number 101016233. CP, DL, EA, MC, SA funded by European Commission - Directorate-General for Communications Networks, Content and Technology through the contract LC-01485746, and Ministerio de Ciencia, Innovacion y Universidades and FEDER, with the project PGC2018-095456-B-I00. DE., MGu funded by Spanish Ministry of Health / REACT-UE (FEDER). DO, GF, IMi, LC funded by Laboratory Directed Research and Development program of Los Alamos National Laboratory (LANL) under project number 20200700ER. DS, ELR, GG, NGR, NW, YW funded by National Institutes of General Medical Sciences (R35GM119582; the content is solely the responsibility of the authors and does not necessarily represent the official views of NIGMS or the National Institutes of Health). FB, FP funded by InPresa, Lombardy Region, Italy. HG, KS funded by European Centre for Disease Prevention and Control. IV funded by Agencia de Qualitat i Avaluacio Sanitaries de Catalunya (AQuAS) through contract 2021-021OE. JDe, SMo, VP funded by Netzwerk Universitatsmedizin (NUM) project egePan (01KX2021). JPB, SH, TH funded by Federal Ministry of Education and Research (BMBF; grant 05M18SIA). KH, MSc, YKh funded by Project SaxoCOV, funded by the German Free State of Saxony. Presentation of data, model results and simulations also funded by the NFDI4Health Task Force COVID-19 (https://www.nfdi4health.de/task-force-covid-19-2) within the framework of a DFG-project (LO-342/17-1). LP, VE funded by Mathematical and Statistical modelling project (MUNI/A/1615/2020), Online platform for real-time monitoring, analysis and management of epidemic situations (MUNI/11/02202001/2020); VE also supported by RECETOX research infrastructure (Ministry of Education, Youth and Sports of the Czech Republic: LM2018121), the CETOCOEN EXCELLENCE (CZ.02.1.01/0.0/0.0/17-043/0009632), RECETOX RI project (CZ.02.1.01/0.0/0.0/16-013/0001761). NIB funded by Health Protection Research Unit (grant code NIHR200908). SAb, SF funded by Wellcome Trust (210758/Z/18/Z).

    

   more » « less
5. Efficient local locking for massively multithreaded in-memory hash-based operators

   https://doi.org/10.1007/s00778-020-00642-5  

   Romanous, Bashar ; Windh, Skyler ; Absalyamov, Ildar ; Budhkar, Prerna ; Halstead, Robert ; Najjar, Walid ; Tsotras, Vassilis ( May 2021 , The VLDB Journal)

   Abstract The join and group-by aggregation are two memory intensive operators that are affecting the performance of relational databases. Hashing is a common approach used to implement both operators. Recent paradigm shifts in multi-core processor architectures have reinvigorated research into how the join and group-by aggregation operators can leverage these advances. However, the poor spatial locality of the hashing approach has hindered performance on multi-core processor architectures which rely on using large cache hierarchies for latency mitigation. Multithreaded architectures can better cope with poor spatial locality by masking memory latency with many outstanding requests. Nevertheless, the number of parallel threads, even in the most advanced multithreaded processors, such as UltraSPARC, is not enough to fully cover the main memory access latency. In this paper, we explore the hardware re-configurability of FPGAs to enable deeper execution pipelines that maintain hundreds (instead of tens) of outstanding memory requests across four FPGAs-drastically increasing concurrency and throughput. We present two end-to-end in-memory accelerators for the join and group-by aggregation operators using FPGAs. Both accelerators use massive multithreading to mask long memory delays of traversing linked-list data structures, while concurrently managing hundreds of thread states across four FPGAs locally. We explore how content addressable memories can be intermixed within our multithreaded designs to act as a synchronizing cache , which enforces locks and merges jobs together before they are written to memory. Throughput results for our hash-join operator accelerator show a speedup between 2 $$\times $$ × and 3.4 $$\times $$ × over the best multi-core approaches with comparable memory bandwidths on uniform and skewed datasets. The accelerator for the hash-based group-by aggregation operator demonstrates that leveraging CAMs achieves average speedup of 3.3 $$\times $$ × with a best case of 9.4 $$\times $$ × in terms of throughput over CPU implementations across five types of data distributions. 

   more » « less