As a recent graduate of the Flatiron School's Data Science Bootcamp, I've been inundated with advice on how to ace technical interviews. A soft skill that keeps coming to the forefront is the ability to explain complex machine learning algorithms to a non-technical person. This series of posts is me sharing with the world how I would explain all the machine learning topics I come across on a regular basis...to my grandma. Some get a bit in-depth, others less so, but all I believe are useful to a non-Data Scientist. In the upcoming parts of this series, I'll be going over: "a model is like a Vending Machine, which given an input (money), will give you some output (a soda can maybe) . . . An algorithm is what is used to train a model, all the decisions a model is supposed to take based on the given input, to give an expected output. For example, an algorithm will decide based on the dollar value of the money given, and the product you chose, whether the money is enough or not, how much balance you are supposed to get [back], and so on."

Increasingly, the data that is relevant for companies' machine learning efforts will be not just some data, but all of it; anything less risks missing what could conceivably be the critical insight down the road, the answer to questions as yet Chief information officers of companies have a strange predicament in an age of AI: They are meant to solve problems for companies by marshaling the relevant data on customers and transactions, but the data itself is going to raise new, unexpected questions. Increasingly, the data that is relevant for companies' machine learning efforts will be not just some data, but all of it; anything less risks missing what could conceivably be the critical insight down the road, the answer to questions as yet unasked. Until recently, the era of "big data," as it's called, has been about providing only the requisite information to answer some straightforward question, where the "known unknowns" are all that matters. For example, if you're a retailer, you might want to know how many of your customers would be likely to return items they've bought based on patterns of purchases. In fact, a group from Indian online apparel retailer Myntra this summer showed off a machine learning model for just such an application.

We consider first order expansions of convex penalized estimators in high-dimensional regression problems with random designs. Our setting includes linear regression and logistic regression as special cases. For a given penalty function $h$ and the corresponding penalized estimator $\hat\beta$, we construct a quantity $\eta$, the first order expansion of $\hat\beta$, such that the distance between $\hat\beta$ and $\eta$ is an order of magnitude smaller than the estimation error $\|\hat{\beta} - \beta^*\|$. In this sense, the first order expansion $\eta$ can be thought of as a generalization of influence functions from the mathematical statistics literature to regularized estimators in high-dimensions. Such first order expansion implies that the risk of $\hat{\beta}$ is asymptotically the same as the risk of $\eta$ which leads to a precise characterization of the MSE of $\hat\beta$; this characterization takes a particularly simple form for isotropic design. Such first order expansion also leads to inference results based on $\hat{\beta}$. We provide sufficient conditions for the existence of such first order expansion for three regularizers: the Lasso in its constrained form, the lasso in its penalized form, and the Group-Lasso. The results apply to general loss functions under some conditions and those conditions are satisfied for the squared loss in linear regression and for the logistic loss in the logistic model.

Predicting human behavior is a difficult and crucial task required for motion planning. It is challenging in large part due to the highly uncertain and multi-modal set of possible outcomes in real-world domains such as autonomous driving. Beyond single MAP trajectory prediction, obtaining an accurate probability distribution of the future is an area of active interest. We present MultiPath, which leverages a fixed set of future state-sequence anchors that correspond to modes of the trajectory distribution. At inference, our model predicts a discrete distribution over the anchors and, for each anchor, regresses offsets from anchor waypoints along with uncertainties, yielding a Gaussian mixture at each time step. Our model is efficient, requiring only one forward inference pass to obtain multi-modal future distributions, and the output is parametric, allowing compact communication and analytical probabilistic queries. We show on several datasets that our model achieves more accurate predictions, and compared to sampling baselines, does so with an order of magnitude fewer trajectories.

We present a new method which generalizes subspace learning based on eigenvalue and generalized eigenvalue problems. This method, Roweis Discriminant Analysis (RDA), is named after Sam Roweis to whom the field of subspace learning owes significantly. RDA is a family of infinite number of algorithms where Principal Component Analysis (PCA), Supervised PCA (SPCA), and Fisher Discriminant Analysis (FDA) are special cases. One of the extreme special cases, which we name Double Supervised Discriminant Analysis (DSDA), uses the labels twice; it is novel and has not appeared elsewhere. We propose a dual for RDA for some special cases. We also propose kernel RDA, generalizing kernel PCA, kernel SPCA, and kernel FDA, using both dual RDA and representation theory. Our theoretical analysis explains previously known facts such as why SPCA can use regression but FDA cannot, why PCA and SPCA have duals but FDA does not, why kernel PCA and kernel SPCA use kernel trick but kernel FDA does not, and why PCA is the best linear method for reconstruction. Roweisfaces and kernel Roweisfaces are also proposed generalizing eigenfaces, Fisherfaces, supervised eigenfaces, and their kernel variants. We also report experiments showing the effectiveness of RDA and kernel RDA on some benchmark datasets.

Recently, machine learning (ML) has introduced advanced solutions to many domains. Since ML models provide business advantage to model owners, protecting intellectual property (IP) of ML models has emerged as an important consideration. Confidentiality of ML models can be protected by exposing them to clients only via prediction APIs. However, model extraction attacks can steal the functionality of ML models using the information leaked to clients through the results returned via the API. In this work, we question whether model extraction is a serious threat to complex, real-life ML models. We evaluate the current state-of-the-art model extraction attack (the Knockoff attack) against complex models. We reproduced and confirm the results in the Knockoff attack paper. But we also show that the performance of this attack can be limited by several factors, including ML model architecture and the granularity of API response. Furthermore, we introduce a defense based on distinguishing queries used for Knockoff attack from benign queries. Despite the limitations of the Knockoff attack, we show that a more realistic adversary can effectively steal complex ML models and evade known defenses.

Viral sequence classification is an important task in pathogen detection, epidemiological surveys and evolutionary studies. Statistical learning methods are widely used to classify and identify viral sequences in samples from environments. These methods face several challenges associated with the nature and properties of viral genomes such as recombination, mutation rate and diversity. Also, new generations of sequencing technologies rise other difficulties by generating massive amounts of fragmented sequences. While linear classifiers are often used to classify viruses, there is a lack of exploration of the accuracy space of existing models in the context of alignment free approaches. In this study, we present an exhaustive assessment procedure exploring the power of linear classifiers in genotyping and subtyping partial and complete genomes. It is applied to the Hepatitis C viruses (HCV). Several variables are considered in this investigation such as classifier types (generative and discriminative) and their hyper-parameters (smoothing value and penalty function), the classification task (genotyping and subtyping), the length of the tested sequences (partial and complete) and the length of k-mer words. Overall, several classifiers perform well given a set of precise combination of the experimental variables mentioned above. Finally, we provide the procedure and benchmark data to allow for more robust assessment of classification from virus genomes.

This paper fortifies the recently introduced hierarchical-optimization recursive least squares (HO-RLS) against outliers which contaminate infrequently linear-regression models. Outliers are modeled as nuisance variables and are estimated together with the linear filter/system variables via a sparsity-inducing (non-)convexly regularized least-squares task. The proposed outlier-robust HO-RLS builds on steepest-descent directions with a constant step size (learning rate), needs no matrix inversion (lemma), accommodates colored nominal noise of known correlation matrix, exhibits small computational footprint, and offers theoretical guarantees, in a probabilistic sense, for the convergence of the system estimates to the solutions of a hierarchical-optimization problem: Minimize a convex loss, which models a-priori knowledge about the unknown system, over the minimizers of the classical ensemble LS loss. Extensive numerical tests on synthetically generated data in both stationary and non-stationary scenarios showcase notable improvements of the proposed scheme over state-of-the-art techniques.

Random features has been widely used for kernel approximation in large-scale machine learning. A number of recent studies have explored data-dependent sampling of features, modifying the stochastic oracle from which random features are sampled. While proposed techniques in this realm improve the approximation, their application is limited to a specific learning task. In this paper, we propose a general scoring rule for sampling random features, which can be employed for various applications with some adjustments. We first observe that our method can recover a number of data-dependent sampling methods (e.g., leverage scores and energy-based sampling). Then, we restrict our attention to a ubiquitous problem in statistics and machine learning, namely Canonical Correlation Analysis (CCA). We provide a principled guide for finding the distribution maximizing the canonical correlations, resulting in a novel data-dependent method for sampling features. Numerical experiments verify that our algorithm consistently outperforms other sampling techniques in the CCA task.

Glioma is one of the most common types of brain tumors arising in the glial cells in the human brain and spinal cord. In addition to the threat of death, glioma treatment is also very costly. Hence, automatic and accurate segmentation and measurement from the early stages are critical in order to prolong the survival rates of the patients and to reduce the costs of health care. In the present work, we propose a novel end-to-end cascaded network for semantic segmentation that utilizes the hierarchical structure of the tumor sub-regions with ResNet-like blocks and Squeeze-and-Excitation modules after each convolution and concatenation block. By utilizing cross-validation, an average ensemble technique, and a simple post-processing technique, we obtained dice scores of 90.34, 81.12, and 78.42 and Hausdorff Distances (95th percentile) of 4.32, 6.28, and 3.70 for the whole tumor, tumor core, and enhancing tumor, respectively, on the online validation set.