In automated machine learning systems, concept drift in input data is one of the main challenges. It deteriorates model performance on new data over time. Previous research on concept drift mostly proposed model retraining after observing performance decreases. However, this approach is suboptimal because the system fixes the problem only after suffering from poor performance on new data. Here, we introduce an adversarial validation approach to concept drift problems in automated machine learning systems. With our approach, the system detects concept drift in new data before making inference, trains a model, and produces predictions adapted to the new data. We show that our approach addresses concept drift effectively with the AutoML3 Lifelong Machine Learning challenge data as well as in Uber's internal automated machine learning system, MaLTA.

Automatic source code analysis in key areas of software engineering, such as code security, can benefit from Machine Learning (ML). However, many standard ML approaches require a numeric representation of data and cannot be applied directly to source code. Thus, to enable ML, we need to embed source code into numeric feature vectors while maintaining the semantics of the code as much as possible. code2vec is a recently released embedding approach that uses the proxy task of method name prediction to map Java methods to feature vectors. However, experimentation with code2vec shows that it learns to rely on variable names for prediction, causing it to be easily fooled by typos or adversarial attacks. Moreover, it is only able to embed individual Java methods and cannot embed an entire collection of methods such as those present in a typical Java class, making it difficult to perform predictions at the class level (e.g., for the identification of malicious Java classes). Both shortcomings are addressed in the research presented in this paper. We investigate the effect of obfuscating variable names during the training of a code2vec model to force it to rely on the structure of the code rather than specific names and consider a simple approach to creating class-level embeddings by aggregating sets of method embeddings. Our results, obtained on a challenging new collection of source-code classification problems, indicate that obfuscating variable names produces an embedding model that is both impervious to variable naming and more accurately reflects code semantics. The datasets, models, and code are shared for further ML research on source code.

Neural nets have been used in an elusive number of scientific disciplines. Nevertheless, their parameterization is largely unexplored. Dense nets are the coordinate transformations of a manifold from which the data is sampled. After processing through a layer, the representation of the original manifold may change. This is crucial for the preservation of its topological structure and should therefore be parameterized correctly. We discuss a method to determine the smallest topology preserving layer considering the data domain as abelian connected Lie group and observe that it is decomposable into $\mathbb{R}^p \times \mathbb {T}^q$. Persistent homology allows us to count its $k$-th homology groups. Using K\"unneth's theorem, we count the $k$-th Betti numbers. Since we know the embedding dimension of $\mathbb{R}^p$ and $\mathcal{S}^1$, we parameterize the bottleneck layer with the smallest possible matrix group, which can represent a manifold with those homology groups. Resnets guarantee smaller embeddings due to the dimension of their state space representation.

In realistic settings, a speaker recognition system needs to identify a speaker given a short utterance, while the utterance used to enroll may be relatively long. However, existing speaker recognition models perform poorly with such short utterances. To solve this problem, we introduce a meta-learning scheme with imbalance length pairs. Specifically, we use a prototypical network and train it with a support set of long utterances and a query set of short utterances. However, since optimizing for only the classes in the given episode is not sufficient to learn discriminative embeddings for other classes in the entire dataset, we additionally classify both support set and query set against the entire classes in the training set to learn a well-discriminated embedding space. By combining these two learning schemes, our model outperforms existing state-of-the-art speaker verification models learned in a standard supervised learning framework on short utterance (1-2 seconds) on VoxCeleb dataset. We also validate our proposed model for unseen speaker identification, on which it also achieves significant gain over existing approaches.

Reinforcement learning (RL) is a powerful framework for learning to take actions to solve tasks. However, in many settings, an agent must winnow down the inconceivably large space of all possible tasks to the single task that it is currently being asked to solve. Can we instead constrain the space of tasks to those that are semantically meaningful? In this work, we introduce a framework for using weak supervision to automatically disentangle this semantically meaningful subspace of tasks from the enormous space of nonsensical "chaff" tasks. We show that this learned subspace enables efficient exploration and provides a representation that captures distance between states. On a variety of challenging, vision-based continuous control problems, our approach leads to substantial performance gains, particularly as the complexity of the environment grows.

Community is a common characteristic of networks including social networks, biological networks, computer and information networks, to name a few. Community detection is a basic step for exploring and analysing these network data. Typically, homogenous network is a type of networks which consists of only one type of objects with one type of links connecting them. There has been a large body of developments in models and algorithms to detect communities over it. However, real-world networks naturally exhibit heterogeneous qualities appearing as multiple types of objects with multi-relational links connecting them. Those heterogeneous information could facilitate the community detection for its constituent homogeneous networks, but has not been fully explored. In this paper, we exploit heterogeneous multi-relational networks (HMRNet) and propose an efficient message passing based algorithm to simultaneously detect communities for all homogeneous networks. Specifically, an HMRNet is reorganized into a hierarchical structure with homogeneous networks as its layers and heterogeneous links connecting them. To detect communities in such an HMRNet, the problem is formulated as a maximum a posterior (MAP) over a factor graph. Finally a message passing based algorithm is derived to find a best solution of the MAP problem. Evaluation on both synthetic and real-world networks confirms the effectiveness of the proposed method.

Similarity learning has gained a lot of attention from researches in recent years and tons of successful approaches have been recently proposed. However, the majority of the state-of-the-art similarity learning methods consider only a binary similarity. In this paper we introduce a new loss function called Continuous Histogram Loss (CHL) which generalizes recently proposed Histogram loss to multiple-valued similarities, i.e. allowing the acceptable values of similarity to be continuously distributed within some range. The novel loss function is computed by aggregating pairwise distances and similarities into 2D histograms in a differentiable manner and then computing the probability of condition that pairwise distances will not decrease as the similarities increase. The novel loss is capable of solving a wider range of tasks including similarity learning, representation learning and data visualization.

Stochastic gradient Langevin dynamics (SGLD) is a poweful algorithm for optimizing a non-convex objective, where a controlled and properly scaled Gaussian noise is added to the stochastic gradients to steer the iterates towards a global minimum. SGLD is based on the overdamped Langevin diffusion which is reversible in time. By adding an anti-symmetric matrix to the drift term of the overdamped Langevin diffusion, one gets a non-reversible diffusion that converges to the same stationary distribution with a faster convergence rate. In this paper, we study the non-reversible stochastic gradient Langevin dynamics (NSGLD) which is based on discretization of the non-reversible Langevin diffusion. We provide finite time performance bounds for the global convergence of NSGLD for solving stochastic non-convex optimization problems. Our results lead to non-asymptotic guarantees for both population and empirical risk minimization problems. Numerical experiments for a simple polynomial function optimization, Bayesian independent component analysis and neural network models show that NSGLD can outperform SGLD with proper choices of the anti-symmetric matrix.

I congratulate Profs. Binyan Jiang, Rui Song, Jialiang Li, and Donglin Zeng (JSLZ) for an exciting development in conducting inferences on optimal dynamic treatment regimes (DTRs) learned via empirical risk minimization using the entropy loss as a surrogate. JSLZ's approach leverages a rejection-and-importance-sampling estimate of the value of a given decision rule based on inverse probability weighting (IPW) and its interpretation as a weighted (or cost-sensitive) classification. Their use of smooth classification surrogates enables their careful approach to analyzing asymptotic distributions. However, even for evaluation purposes, the IPW estimate is problematic as it leads to weights that discard most of the data and are extremely variable on whatever remains. In this comment, I discuss an optimization-based alternative to evaluating DTRs, review several connections, and suggest directions forward. This extends the balanced policy evaluation approach of Kallus (2018a) to the longitudinal setting.

Individualized treatment recommendation (ITR) is an important analytic framework for precision medicine. The goal is to assign proper treatments to patients based on their individual characteristics. From the machine learning perspective, the solution to an ITR problem can be formulated as a weighted classification problem to maximize the average benefit that patients receive from the recommended treatments. Several methods have been proposed for ITR in both binary and multicategory treatment setups. In practice, one may prefer a more flexible recommendation with multiple treatment options. This motivates us to develop methods to obtain a set of near-optimal individualized treatment recommendations alternative to each other, called alternative individualized treatment recommendations (A-ITR). We propose two methods to estimate the optimal A-ITR within the outcome weighted learning (OWL) framework. We show the consistency of these methods and obtain an upper bound for the risk between the theoretically optimal recommendation and the estimated one. We also conduct simulation studies, and apply our methods to a real data set for Type 2 diabetic patients with injectable antidiabetic treatments. These numerical studies have shown the usefulness of the proposed A-ITR framework. We develop a R package aitr which can be found at https://github.com/menghaomiao/aitr.