Privacy is a major issue in learning from distributed data. Recently the cryptographic literature has provided several tools for this task. However, these tools either reduce the quality/accuracy of the learning algorithm---e.g., by adding noise---or they incur a high performance penalty and/or involve trusting external authorities. We propose a methodology for {\sl private distributed machine learning from light-weight cryptography} (in short, PD-ML-Lite). We apply our methodology to two major ML algorithms, namely non-negative matrix factorization (NMF) and singular value decomposition (SVD). Our resulting protocols are communication optimal, achieve the same accuracy as their non-private counterparts, and satisfy a notion of privacy---which we define---that is both intuitive and measurable. Our approach is to use lightweight cryptographic protocols (secure sum and normalized secure sum) to build learning algorithms rather than wrap complex learning algorithms in a heavy-cost MPC framework. We showcase our algorithms' utility and privacy on several applications: for NMF we consider topic modeling and recommender systems, and for SVD, principal component regression, and low rank approximation.

Policy evaluation is a key process in reinforcement learning. It assesses a given policy using estimation of the corresponding value function. When using a parameterized function to approximate the value, it is common to optimize the set of parameters by minimizing the sum of squared Bellman Temporal Differences errors. However, this approach ignores certain distributional properties of both the errors and value parameters. Taking these distributions into account in the optimization process can provide useful information on the amount of confidence in value estimation. In this work we propose to optimize the value by minimizing a regularized objective function which forms a trust region over its parameters. We present a novel optimization method, the Kalman Optimization for Value Approximation (KOVA), based on the Extended Kalman Filter. KOVA minimizes the regularized objective function by adopting a Bayesian perspective over both the value parameters and noisy observed returns. This distributional property provides information on parameter uncertainty in addition to value estimates. We provide theoretical results of our approach and analyze the performance of our proposed optimizer on domains with large state and action spaces.

We present an online algorithm that induces decision trees using gradient information as the source of supervision. In contrast to previous approaches to gradient-based tree learning, we do not require soft splits or construction of a new tree for every update. In experiments, our method performs comparably to standard incremental classification trees and outperforms state of the art incremental regression trees. We also show how the method can be used to construct a novel type of neural network layer suited to learning representations from tabular data and find that it increases accuracy of multiclass and multi-label classification.

In unsupervised learning, there is no obvious straightforward loss function which can capture the major factors of variations and similarities. Since natural systems have smooth dynamics, an opportunity is lost if an unsupervised loss function remains static during the training process. The absence of concrete supervision suggests that smooth complex dynamics should be integrated as a substitute to the classical static loss functions to better make use of the gradual and uncertain knowledge acquired through self-supervision. In this paper, we propose Dynamic Autoencoder (DynAE), a new model for deep clustering that allows to solve a clustering-reconstruction trade-off by gradually and smoothly eliminating the reconstruction objective in favor of a construction one while preserving the space topology. Experimental evaluations on benchmark datasets show that our approach achieves state-of-the-art results compared to all the other autoencoder-based clustering methods.

This paper describes some applications of an incremental implementation of the principal component analysis (PCA). The algorithm updates the transformation coefficients matrix on-line for each new sample, without the need to keep all the samples in memory. The algorithm is formally equivalent to the usual batch version, in the sense that given a sample set the transformation coefficients at the end of the process are the same. The implications of applying the PCA in real time are discussed with the help of data analysis examples. In particular we focus on the problem of the continuity of the PCs during an on-line analysis.

This paper proposes a novel generic one-class feature learning method which is based on intra-class splitting. In one-class classification, feature learning is challenging, because only samples of one class are available during training. Hence, state-of-the-art methods require reference multi-class datasets to pretrain feature extractors. In contrast, the proposed method realizes feature learning by splitting the given normal class into typical and atypical normal samples. By introducing closeness loss and dispersion loss, an intra-class joint training procedure between the two subsets after splitting enables the extraction of valuable features for one-class classification. Various experiments on three well-known image classification datasets demonstrate the effectiveness of our method which outperformed other baseline models in 25 of 30 experiments.

One of the challenges in large-scale information retrieval (IR) is to develop fine-grained and domain-specific methods to answer natural language questions. Despite the availability of numerous sources and datasets for answer retrieval, Question Answering (QA) remains a challenging problem due to the difficulty of the question understanding and answer extraction tasks. One of the promising tracks investigated in QA is to map new questions to formerly answered questions that are `similar'. In this paper, we propose a novel QA approach based on Recognizing Question Entailment (RQE) and we describe the QA system and resources that we built and evaluated on real medical questions. First, we compare machine learning and deep learning methods for RQE using different kinds of datasets, including textual inference, question similarity and entailment in both the open and clinical domains. Second, we combine IR models with the best RQE method to select entailed questions and rank the retrieved answers. To study the end-to-end QA approach, we built the MedQuAD collection of 47,457 question-answer pairs from trusted medical sources, that we introduce and share in the scope of this paper. Following the evaluation process used in TREC 2017 LiveQA, we find that our approach exceeds the best results of the medical task with a 29.8% increase over the best official score. The evaluation results also support the relevance of question entailment for QA and highlight the effectiveness of combining IR and RQE for future QA efforts. Our findings also show that relying on a restricted set of reliable answer sources can bring a substantial improvement in medical QA.

The weights of a neural network cannot be calculated using an analytical method. Instead, the weights must be discovered via an empirical optimization procedure called stochastic gradient descent.

If you want to succeed in as data scientist, you need to have knowledge on statistics, linear algebra, data science, machine learning. You want to be be expert you learn about deep learning, Natural language processing, Computer vision and more. How many Types of statistics are there?

Now, what was the Gradient Descent algorithm? Above algorithm says, to perform the GD, we need to calculate the gradient of the cost function J. And to calculate the gradient of the cost function, we need to sum (yellow circle!) the cost of each sample. If we have 3 million samples, we have to loop through 3 million times or use the dot product. Do you see np.dot(X.T, y_hat-y) above?