We propose Stochastic Weight Averaging in Parallel (SW AP), an algorithm to accelerate DNN training. Our algorithm uses large mini-batches to compute an approximate solution quickly and then refines it by averaging the weights of multiple models computed independently and in parallel. The resulting models generalize equally well as those trained with small mini-batches but are produced in a substantially shorter time. We demonstrate the reduction in training time and the good generalization performance of the resulting models on the computer vision datasets CIFAR10, CIFAR100, and ImageNet. Stochastic gradient descent (SGD) and its variants are the de-facto methods to train deep neural networks (DNNs). Each iteration of SGD computes an estimate of the objective's gradient by sampling a mini-batch of the available training data and computing the gradient of the loss restricted to the sampled data. A popular strategy to accelerate DNN training is to increase the mini-batch size together with the available computational resources. Larger mini-batches produce more precise gradient estimates; these allow for higher learning rates and achieve larger reductions of the training loss per iteration.

The mean shift algorithm is a popular way to find modes of some probability density functions taking a specific kernel-based shape, used for clustering or visual tracking. Since its introduction, it underwent several practical improvements and generalizations, as well as deep theoretical analysis mainly focused on its convergence properties. In spite of encouraging results, this question has not received a clear general answer yet. In this paper we focus on a specific class of kernels, adapted in particular to the distributions clustering applications which motivated this work. We show that a novel Mean Shift variant adapted to them can be derived, and proved to converge after a finite number of iterations. In order to situate this new class of methods in the general picture of the Mean Shift theory, we alo give a synthetic exposure of existing results of this field.

One of the promising methods for the treatment of complex diseases such as cancer is combinational therapy. Due to the combinatorial complexity, machine learning models can be useful in this field, where significant improvements have recently been achieved in determination of synergistic combinations. In this study, we investigate the effectiveness of different compound representations in predicting the drug synergy. On a large drug combination screen dataset, we first demonstrate the use of a promising representation that has not been used for this problem before, then we propose an ensemble on representation-model combinations that outperform each of the baseline models. 1 Scientific Background A drug combination is called synergistic if the effect of the drug combination on the reference cell is greater than the total effect taken from the administration of the individual drugs. If the opposite situation is observed, the drug combination is called antagonistic . Understanding whether a combination is antagonistic or synergistic is a resource and time intensive task.

Supervised learning models, also known as quantitative structure-activity regression (QSAR) models, are increasingly used in assisting the process of preclinical, small molecule drug discovery. The models are trained on data consisting of a finite dimensional representation of molecular structures and their corresponding target specific activities. These models can then be used to predict the activity of previously unmeasured novel compounds. In this work we address two problems related to this approach. The first is to estimate the extent to which the quality of the model predictions degrades for compounds very different from the compounds in the training data. The second is to adjust for the screening dependent selection bias inherent in many training data sets. In the most extreme cases, only compounds which pass an activity-dependent screening are reported. By using a semi-supervised learning framework, we show that it is possible to make predictions which take into account the similarity of the testing compounds to those in the training data and adjust for the reporting selection bias. We illustrate this approach using publicly available structure-activity data on a large set of compounds reported by GlaxoSmithKline (the Tres Cantos AntiMalarial Set) to inhibit in vitro P. falciparum growth.

Hidden Markov Model (HMM) combined with Gaussian Process (GP) emission can be effectively used to estimate the hidden state with a sequence of complex input-output relational observations. Especially when the spectral mixture (SM) kernel is used for GP emission, we call this model as a hybrid HMM-GPSM. This model can effectively model the sequence of time-series data. However, because of a large number of parameters for the SM kernel, this model can not effectively be trained with a large volume of data having (1) long sequence for state transition and 2) a large number of time-series dataset in each sequence. This paper proposes a scalable learning method for HMM-GPSM. To effectively train the model with a long sequence, the proposed method employs a Stochastic Variational Inference (SVI) approach. Also, to effectively process a large number of data point each time-series data, we approximate the SM kernel using Reparametrized Random Fourier Feature (R-RFF). The combination of these two techniques significantly reduces the training time. We validate the proposed learning method in terms of its hidden-sate estimation accuracy and computation time using large-scale synthetic and real data sets with missing values.

Regularization is an effective way to promote the generalization performance of machine learning models. In this paper, we focus on label smoothing, a form of output distribution regularization that prevents overfitting of a neural network by softening the ground-truth labels in the training data in an attempt to penalize overconfident outputs. Existing approaches typically use cross-validation to impose this smoothing, which is uniform across all training data. In this paper, we show that such label smoothing imposes a quantifiable bias in the Bayes error rate of the training data, with regions of the feature space with high overlap and low marginal likelihood having a lower bias and regions of low overlap and high marginal likelihood having a higher bias. These theoretical results motivate a simple objective function for data-dependent smoothing to mitigate the potential negative consequences of the operation while maintaining its desirable properties as a regularizer. We call this approach Structural Label Smoothing (SLS). We implement SLS and empirically validate on synthetic, Higgs, SVHN, CIFAR-10, and CIFAR-100 datasets. The results confirm our theoretical insights and demonstrate the effectiveness of the proposed method in comparison to traditional label smoothing.

Comparing and aligning large datasets is a pervasive problem occurring across many different knowledge domains. We introduce and study MREC, a recursive decomposition algorithm for computing matchings between data sets. The basic idea is to partition the data, match the partitions, and then recursively match the points within each pair of identified partitions. The matching itself is done using black box matching procedures that are too expensive to run on the entire data set. Using an absolute measure of the quality of a matching, the framework supports optimization over parameters including partitioning procedures and matching algorithms. By design, MREC can be applied to extremely large data sets. We analyze the procedure to describe when we can expect it to work well and demonstrate its flexibility and power by applying it to a number of alignment problems arising in the analysis of single cell molecular data.

The wide adoption of machine learning in the critical domains such as medical diagnosis, law, education had propelled the need for interpretable techniques due to the need for end users to understand the reasoning behind decisions due to learning systems. The computational intractability of interpretable learning led practitioners to design heuristic techniques, which fail to provide sound handles to tradeoff accuracy and interpretability. Motivated by the success of MaxSA T solvers over the past decade, recently MaxSA T -based approach, called MLIC, was proposed that seeks to reduce the problem of learning interpretable rules expressed in Conjunctive Normal Form (CNF) to a MaxSA T query. While MLIC was shown to achieve accuracy similar to that of other state of the art black-box classifiers while generating small interpretable CNF formulas, the runtime performance of MLIC is significantly lagging and renders approach unusable in practice. In this context, authors raised the question: Is it possible to achieve the best of both worlds, i.e., a sound framework for interpretable learning that can take advantage of MaxSAT solvers while scaling to real-world instances? In this paper, we take a step towards answering the above question in affirmation. We propose IMLI: an incremental approach to MaxSA T based framework that achieves scalable runtime performance via partition-based training methodology. Extensive experiments on benchmarks arising from UCI repository demonstrate that IMLI achieves up to three orders of magnitude runtime improvement without loss of accuracy and interpretability.

In this paper, we proposed a novel automated model, called Vulnerability Index for Population at Risk (VIPAR) scores, to identify rare populations for their future shooting victimizations. Likewise, the focused deterrence approach identifies vulnerable individuals and offers certain types of treatments (e.g., outreach services) to prevent violence in communities. The proposed rule-based engine model is the first AI-based model for victim prediction. This paper aims to compare the list of focused deterrence strategy with the VIPAR score list regarding their predictive power for the future shooting victimizations. Drawing on the criminological studies, the model uses age, past criminal history, and peer influence as the main predictors of future violence. Social network analysis is employed to measure the influence of peers on the outcome variable. The model also uses logistic regression analysis to verify the variable selections. Our empirical results show that VIPAR scores predict 25.8% of future shooting victims and 32.2% of future shooting suspects, whereas focused deterrence list predicts 13% of future shooting victims and 9.4% of future shooting suspects. The model outperforms the intelligence list of focused deterrence policies in predicting the future fatal and non-fatal shootings. Furthermore, we discuss the concerns about the presumption of innocence right.

In my previous article, I explained what regression was and showed how it could be used in application. This week, I'm going to go over the majority of common machine learning models used in practice, so that I can spend more time building and improving models rather than explaining the theory behind it. All machine learning models are categorized as either supervised or unsupervised. If the model is a supervised model, it's then sub-categorized as either a regression or classification model. We'll go over what these terms mean and the corresponding models that fall into each category below.