Some colleagues of mine are working with survey responses, and are attempting to predict behaviors with demographic data. So, the plan is to define a dependent variable from some combination of responses to the survey questions, and then use a regression technique to model this dependent variable using other characteristics of the respondents. We all agree on the 5 or so questions that will define the dependent variable, but we disagree on how to specify the definition. I want to look at the actual questions being answered, and create a "score" as a weighted count of the'yeses' to the questions (weights based on how "on point" each question is to the behavior we are trying to define). My colleagues thought that this was too imprecise, and particularly criticised the'intuitive' weight assignment.

One of Canada's largest banks and a multinational software firm are both announcing expanded investment in artificial intelligence (AI) that will include partnerships with Canadian universities, according to announcements made Wednesday morning. Microsoft Corp. just announced its acquisition of Montreal-based AI startup Maluuba Jan. 13 and now it says it wants to double the size of that team over the next two years. That investment will be paired with a $6 million gift to the Universite de Montreal and $1 million to McGill University to be spent on AI research over five years. The announcement was made by Microsoft President Brad Smith from the World Economic Forum in Davos, where he was joined by Navdeep Bains, Canada's Minister of Innovation, Science and Economic Development. "Microsoft's investment is proof of Canadians' world-renowned expertise in artificial intelligence and machine learning," Bains is quoted as saying in a press release.

Want to watch this again later? Report Need to report the video? Report Need to report the video? Need to report the video? This feature is not available right now.

This lecture will cover recurrent neural networks, the key ingredient in the deep learning toolbox for handling sequential computation and modelling sequences. It will start by explaining how gradients can be computed (by considering the time-unfolded graph) and how different architectures can be designed to summarize a sequence, generate a sequence by ancestral sampling in a fully-observed directed model, or learn to map a vector to a sequence, a sequence to a sequence (of the same or different length) or a sequence to a vector. The issue of long-term dependencies, why it arises, and what has been proposed to alleviate it will be core subject of the discussion in this lecture. This includes changes in the architecture and initialization, as well as how to properly characterize the architecture in terms of recurrent or feedforward depth and its ability to create shortcuts or fast propagation of gradients in the unfolded graph. Open questions regarding the limitations of training by maximum likelihood (teacher forcing) and ideas towards towards making learning online (not requiring backprop through time) will also be discussed.

Stability is an important aspect of a classification procedure because unstable predictions can potentially reduce users' trust in a classification system and also harm the reproducibility of scientific conclusions. The major goal of our work is to introduce a novel concept of classification instability, i.e., decision boundary instability (DBI), and incorporate it with the generalization error (GE) as a standard for selecting the most accurate and stable classifier. Specifically, we implement a two-stage algorithm: (i) initially select a subset of classifiers whose estimated GEs are not significantly different from the minimal estimated GE among all the candidate classifiers; (ii) the optimal classifier is chosen as the one achieving the minimal DBI among the subset selected in stage (i). This general selection principle applies to both linear and nonlinear classifiers. Large-margin classifiers are used as a prototypical example to illustrate the above idea. Our selection method is shown to be consistent in the sense that the optimal classifier simultaneously achieves the minimal GE and the minimal DBI. Various simulations and real examples further demonstrate the advantage of our method over several alternative approaches.

In rare disease physician targeting, a major challenge is how to identify physicians who are treating diagnosed or underdiagnosed rare diseases patients. Rare diseases have extremely low incidence rate. For a specified rare disease, only a small number of patients are affected and a fractional of physicians are involved. The existing targeting methodologies, such as segmentation and profiling, are developed under mass market assumption. They are not suitable for rare disease market where the target classes are extremely imbalanced. The authors propose a graphical model approach to predict targets by jointly modeling physician and patient features from different data spaces and utilizing the extra relational information. Through an empirical example with medical claim and prescription data, the proposed approach demonstrates better accuracy in finding target physicians. The graph representation also provides visual interpretability of relationship among physicians and patients. The model can be extended to incorporate more complex dependency structures. This article contributes to the literature of exploring the benefit of utilizing relational dependencies among entities in healthcare industry.

We introduce a new dynamical system for sequentially observed multivariate count data. This model is based on the gamma--Poisson construction---a natural choice for count data---and relies on a novel Bayesian nonparametric prior that ties and shrinks the model parameters, thus avoiding overfitting. We present an efficient MCMC inference algorithm that advances recent work on augmentation schemes for inference in negative binomial models. Finally, we demonstrate the model's inductive bias using a variety of real-world data sets, showing that it exhibits superior predictive performance over other models and infers highly interpretable latent structure.

PixelCNNs are a recently proposed class of powerful generative models with tractable likelihood. Here we discuss our implementation of PixelCNNs which we make available at https://github.com/openai/pixel-cnn. Our implementation contains a number of modifications to the original model that both simplify its structure and improve its performance. 1) We use a discretized logistic mixture likelihood on the pixels, rather than a 256-way softmax, which we find to speed up training. 2) We condition on whole pixels, rather than R/G/B sub-pixels, simplifying the model structure. 3) We use downsampling to efficiently capture structure at multiple resolutions. 4) We introduce additional short-cut connections to further speed up optimization. 5) We regularize the model using dropout. Finally, we present state-of-the-art log likelihood results on CIFAR-10 to demonstrate the usefulness of these modifications.

In general, the interest lies in obtaining informations about the central characteristics of the underlying lifetime distribution (mean lifetime or survival probabilities for instance), often with the objective of comparing results between different conditions under which the lifetime data are acquired. In this work, we will address the problem of inferring about the (upper) tail of the lifetime distribution, for data subject both to random (right) censoring and competing risks. Suppose indeed that we are interested in the lifetimes of n individuals or items, which are subject to K different causes of death or failure, and to random censorship (from the right) as well. We are particularly interested in one of these causes (this main cause will be considered as cause number k thereafter, where k P t1,..., Ku), and we suppose that all causes are exclusive and are likely to be dependent on the others. The censoring time is assumed to be independent of the different causes of death or failure and of the observed lifetime itself.

Random forest (RF) missing data algorithms are an attractive approach for dealing with missing data. They have the desirable properties of being able to handle mixed types of missing data, they are adaptive to interactions and nonlinearity, and they have the potential to scale to big data settings. Currently there are many different RF imputation algorithms but relatively little guidance about their efficacy, which motivated us to study their performance. Using a large, diverse collection of data sets, performance of various RF algorithms was assessed under different missing data mechanisms. Algorithms included proximity imputation, on the fly imputation, and imputation utilizing multivariate unsupervised and supervised splitting---the latter class representing a generalization of a new promising imputation algorithm called missForest. Performance of algorithms was assessed by ability to impute data accurately. Our findings reveal RF imputation to be generally robust with performance improving with increasing correlation. Performance was good under moderate to high missingness, and even (in certain cases) when data was missing not at random.