To enable personalized cancer treatment, machine learning models have been developed to predict drug response as a function of tumor and drug features. However, most algorithm development efforts have relied on cross-validation within a single study to assess model accuracy. While an essential first step, cross-validation within a biological data set typically provides an overly optimistic estimate of the prediction performance on independent test sets. To provide a more rigorous assessment of model generalizability between different studies, we use machine learning to analyze five publicly available cell line-based data sets: National Cancer Institute 60, ancer Therapeutics Response Portal (CTRP), Genomics of Drug Sensitivity in Cancer, Cancer Cell Line Encyclopedia and Genentech Cell Line Screening Initiative (gCSI). Based on observed experimental variability across studies, we explore estimates of prediction upper bounds. We report performance results of a variety of machine learning models, with a multitasking deep neural network achieving the best cross-study generalizability. By multiple measures, models trained on CTRP yield the most accurate predictions on the remaining testing data, and gCSI is the most predictable among the cell line data sets included in this study. With these experiments and further simulations on partial data, two lessons emerge: (1) differences in viability assays can limit model generalizability across studies and (2) drug diversity, more than tumor diversity, is crucial for raising model generalizability in preclinical screening.

The move toward exascale computing for scientific simulations is placing new demands on compression techniques. It is expected that the I/O system will not be able to support the volume of data that will be written out. To enable quantitative analysis and scientific discovery, we need techniques that can compress high-dimensional simulation data with near-perfect reconstruction. In this work, we investigate Compressed Sensing (CS) to reduce the size of the data from a fusion simulation of a tokamak in 3 dimensions (Figure (a)). The computational domain of the simulation is a toroid, composed of 32 poloidal planes (shown in blue). Each plane has nearly 600,000 grid points, arranged irregularly, and distributed across multiple processors of a massively parallel system. Since these data are analyzed to understand the behavior of coherent structures (Figure (b)) over time, it is important that these structures remain unchanged after reconstruction using CS. We conducted several experiments to understand how best to apply CS to our data set. We used several metrics to investigate the effects of preprocessing, including scaling to improve the contrast in the data and thresholding to increase the sparsity. To determine the size of the compressed data that would enable near-perfect reconstruction, we evaluated the quality of reconstruction (shown in Figures (c) and (d) using the R2 metric) as we varied the percentage of compression (m/n) for various levels of sparsity (k/n) in the data. We found that a successful application of CS is bounded by the percentage of sparsity in the data - the data have to be sparse enough for compression using CS, but not so sparse that it is more cost effective to just write out the locations and values of the non-zero data points. Our experiments also indicated that scaling the data is very helpful and thresholding helps both with compression and the coherent structure analysis performed on the data.

Wind energy is scheduled on the power grid using 0–6 h ahead forecasts generated from computer simulations or historical data. When the forecasts are inaccurate, control room operators use their expertise, as well as the actual generation from previous days, to estimate the amount of energy to schedule. However, this is a challenge, and it would be useful for the operators to have additional information they can exploit to make better informed decisions. In this paper, we use techniques from time series analysis to determine if there are motifs, or frequently occurring diurnal patterns in wind generation data. We compare two different representations of the data and four different ways of identifying the number of motifs. Using data from wind farms in Tehachapi Pass and mid-Columbia Basin, we describe our findings and discuss how these motifs can be used to guide scheduling decisions.

In this paper, we formulate the problem of predicting wind generation as one of streaming data analysis. We want to understand if it is possible to use the weather data in a time window just before the current time to gain insight into how the wind generation might behave in a time interval just after the current time. Specifically, we use a singular value decomposition of the weather data, and how that the number of singular values and the largest singular value can be used to predict the magnitude of the change in the generation in the near future. The analysis uses an incremental algorithm based on a sliding window for reduced computational costs.

Discrete k-median (DKM) clustering problems arise in many real-life applications that involve time-series data sets, in which nondiscrete clustering methods may not represent the problem domain adequately. In this study, we propose mathematical programming formulations and solution methods to efficiently solve the DKM clustering problem. We develop approximation algorithms from a bilinear formulation of the discrete k-median problem using an uncoupled bilinear program algorithm. This approximation algorithm, which we refer to as DKM-L, is composed of two alternating linear programs, where one can be solved in linear time and the other is a minimum cost assignment problem. We then modify this algorithm by replacing the assignment problem with an efficient sequential algorithm for a faster approximation, which we call DKM-S. We also propose a compact exact integer formulation, DKM-I, and a more efficient network design-based exact mixed-integer formulation, DKM-M. All of our methods use arbitrary pairwise distance matrices as input. We apply our methods to simulated single-variate and multivariate random walk time-series data. We report comparative clustering performances using normalized mutual information (NMI) and solution speeds among the DKM methods we propose. We also compare our methods to other clustering algorithms that can operate with distance matrices, such as hierarchical cluster trees (HCT) and partition around medoids (PAM). We present NMI scores and classification accuracies of our DKM algorithms compared to HCT and PAM using five different distance measures on simluated data, as well as public benchmark and real-life neural time-series data sets. We show that DKM-S is much faster than HCT, PAM, and all other DKM methods and produces consistently good clustering results on all data sets.

Wind energy is scheduled on the power grid using 0-6 hour ahead forecasts generated from computer simulations or historical data. When the forecasts are inaccurate, control room operators use their expertise, as well as the actual generation from previous days, to estimate the amount of energy to schedule. However, this is a challenge, and it would be useful for the operators to have additional information they can exploit to make better informed decisions. In this paper, we use techniques from time series analysis to determine if there are motifs, or frequently occurring diurnal patterns in wind generation data. Using data from wind farms in Tehachapi Pass and mid-Columbia Basin, we describe our findings and discuss how these motifs can be used to guide scheduling decisions.

There is an urgent need for a quick screening process that could help neurologists diagnose and determine whether a patient is epileptic versus simply demonstrating symptoms linked to epilepsy but actually stemming from a different illness. An inaccurate diagnosis could have fatal consequences, particularly in operating rooms and intensive care units. Electroencephalogram (EEG) has been traditionally used, as a gold standard, to diagnose patients by evaluating those brain functions that might correspond to epilepsy and other brain disorders. This research therefore focuses on developing new classification techniques for multichannel EEG recordings. Two time-series classification techniques, namely, Support Feature Machine (SFM) and Network-Based Support Vector Machine (SVM) (NSVM), are proposed in this paper to predict from EEG readings whether a person is epileptic or nonepileptic. The SFM approach is an optimization model that maximizes classification accuracy by selecting a group of electrodes (features) that has strong class separability based on time-series similarity measures and correctly classifies EEG samples in the training phase. The NSVM approach integrates a new network-based model for multidimensional time-series data with traditional SVMs to exploit both the spatial and temporal characteristics of EEG data. The proposed techniques are tested on two EEG data sets acquired from ten and five patients, respectively. Compared with other commonly used classification techniques such as SVM and decision trees, the proposed SFM and NSVM techniques provide very promising and practical results and require much less time and memory resources than traditional techniques. This study is a necessary application of data mining to advance the diagnosis and treatment of human epilepsy.

In this study, a novel multidimensional time series classification technique, namely support feature machine (SFM), is proposed. SFM is inspired by the optimization model of support vector machine and the nearest neighbor rule to incorporate both spatial and temporal of the multi-dimensional time series data. This paper also describes an application of SFM for detecting abnormal brain activity. Epilepsy is a case in point in this study. In epilepsy studies, electroencephalograms (EEGs), acquired in multidimensional time series format, have been traditionally used as a gold-standard tool for capturing the electrical changes in the brain. From multi-dimensional EEG time series data, SFM was used to identify seizure pre-cursors and detect seizure susceptibility (pre-seizure) periods. The empirical results showed that SFM achieved over 80% correct classification of per-seizure EEG on average in 10 patients using 5-fold cross validation. The proposed optimization model of SFM is very compact and scalable, and can be implemented as an online algorithm. The outcome of this study suggests that it is possible to construct a computerized algorithm used to detect seizure pre-cursors and warn of impending seizures through EEG classification.

Epilepsy is one of the most common brain disorders, but the dynamical transitions to neurological dysfunctions of epilepsy are not well understood in current neuroscience research. Uncontrolled epilepsy poses a significant burden to society due to associated healthcare cost to treat and control the unpredictable and spontaneous occurrence of seizures. The objective of this study is to develop and present a novel classification technique that is used to classify normal and abnormal (epileptic) brain activities through quantitative analyses of electroencephalogram (EEG) recordings. Such technique is based on the integration of sophisticated approaches from data mining and signal processing research (i.e., chaos theory,k-nearest neighbor, and statistical time series analysis). The proposed technique can correctly classify normal and abnormal EEGs with a sensitivity of 81.29% and a specificity of 72.86%, on average, across ten patients. Experimental results suggest that the proposed technique can be used to develop abnormal brain activity classification for detecting seizure precursors. Success of this study demonstrates that the proposed technique can excavate hidden patterns/relationships in EEGs and give greater understanding of brain functions from a system perspective, which will advance current diagnosis and treatment of epilepsy.