The development of accurate and transferable machine learning (ML) potentials for predicting molecular energetics is a challenging task. The process of data generation to train such ML potentials is a task neither well understood nor researched in detail. In this work, we present a fully automated approach for the generation of datasets with the intent of training universal ML potentials. It is based on the concept of active learning (AL) via Query by Committee (QBC), which uses the disagreement between an ensemble of ML potentials to infer the reliability of the ensemble's prediction. QBC allows the presented AL algorithm to automatically sample regions of chemical space where the ML potential fails to accurately predict the potential energy. AL improves the overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test cases by mitigating human biases in deciding what new training data to use. AL also reduces the training set size to a fraction of the data required when using naive random sampling techniques. To provide validation of our AL approach we develop the COMP6 benchmark (publicly available on GitHub), which contains a diverse set of organic molecules. Through the AL process, it is shown that the AL-based potentials perform as well as the ANI-1 potential on COMP6 with only 10% of the data, and vastly outperforms ANI-1 with 25% the amount of data. Finally, we show that our proposed AL technique develops a universal ANI potential (ANI-1x) that provides accurate energy and force predictions on the entire COMP6 benchmark. This universal ML potential achieves a level of accuracy on par with the best ML potentials for single molecule or materials, while remaining applicable to the general class of organic molecules comprised of the elements CHNO.
Survival analysis has been developed and applied in the number of areas including manufacturing, finance, economics and healthcare. In healthcare domain, usually clinical data are high-dimensional, sparse and complex and sometimes there exists few amount of time-to-event (labeled) instances. Therefore building an accurate survival model from electronic health records is challenging. With this motivation, we address this issue and provide a new survival analysis framework using deep learning and active learning with a novel sampling strategy. First, our approach provides better representation with lower dimensions from clinical features using labeled (time-to-event) and unlabeled (censored) instances and then actively trains the survival model by labeling the censored data using an oracle. As a clinical assistive tool, we introduce a simple effective treatment recommendation approach based on our survival model. In the experimental study, we apply our approach on SEER-Medicare data related to prostate cancer among African-Americans and white patients. The results indicate that our approach outperforms significantly than baseline models.
We develop an algorithm for minimizing a function using $n$ batched function value measurements at each of $T$ rounds by using classifiers to identify a function's sublevel set. We show that sufficiently accurate classifiers can achieve linear convergence rates, and show that the convergence rate is tied to the difficulty of active learning sublevel sets. Further, we show that the bootstrap is a computationally efficient approximation to the necessary classification scheme. The end result is a computationally efficient derivative-free algorithm requiring no tuning that consistently outperforms other approaches on simulations, standard benchmarks, real-world DNA binding optimization, and airfoil design problems whenever batched function queries are natural.
We present an approach that improves the sample complexity for a variety of curve fitting problems, including active learning for linear regression, polynomial regression, and continuous sparse Fourier transforms. In the active linear regression problem, one would like to estimate the least squares solution $\beta^*$ minimizing $\|X\beta - y\|_2$ given the entire unlabeled dataset $X \in \mathbb{R}^{n \times d}$ but only observing a small number of labels $y_i$. We show that $O(d)$ labels suffice to find a constant factor approximation $\tilde{\beta}$: \[ \mathbb{E}[\|X\tilde{\beta} - y\|_2^2] \leq 2 \mathbb{E}[\|X \beta^* - y\|_2^2]. \] This improves on the best previous result of $O(d \log d)$ from leverage score sampling. We also present results for the \emph{inductive} setting, showing when $\tilde{\beta}$ will generalize to fresh samples; these apply to continuous settings such as polynomial regression. Finally, we show how the techniques yield improved results for the non-linear sparse Fourier transform setting.
In this work, we propose an automated method to identify semantic bugs in student programs, called ATAS, which builds upon the recent advances in both symbolic execution and active learning. Symbolic execution is a program analysis technique which can generate test cases through symbolic constraint solving. Our method makes use of a reference implementation of the task as its sole input. We compare our method with a symbolic execution-based baseline on 6 programming tasks retrieved from CodeForces comprising a total of 23K student submissions. We show an average improvement of over 2.5x over the baseline in terms of runtime (thus making it more suitable for online evaluation), without a significant degradation in evaluation accuracy.
Determinantal point processes (DPPs) are an important concept in random matrix theory and combinatorics. They have also recently attracted interest in the study of numerical methods for machine learning, as they offer an elegant "missing link" between independent Monte Carlo sampling and deterministic evaluation on regular grids, applicable to a general set of spaces. This is helpful whenever an algorithm explores to reduce uncertainty, such as in active learning, Bayesian optimization, reinforcement learning, and marginalization in graphical models. To draw samples from a DPP in practice, existing literature focuses on approximate schemes of low cost, or comparably inefficient exact algorithms like rejection sampling. We point out that, for many settings of relevance to machine learning, it is also possible to draw exact samples from DPPs on continuous domains. We start from an intuitive example on the real line, which is then generalized to multivariate real vector spaces. We also compare to previously studied approximations, showing that exact sampling, despite higher cost, can be preferable where precision is needed.
Breast cancer is the second most common malignancy among women and has become a major public health problem in current society. Traditional breast cancer identification requires experienced pathologists to carefully read the breast slice, which is laborious and suffers from inter-observer variations. Consequently, an automatic classification framework for breast cancer identification is worthwhile to develop. Recent years witnessed the development of deep learning technique. Increasing number of medical applications start to use deep learning to improve diagnosis accuracy. In this paper, we proposed a novel training strategy, namely reversed active learning (RAL), to train network to automatically classify breast cancer images. Our RAL is applied to the training set of a simple convolutional neural network (CNN) to remove mislabeled images. We evaluate the CNN trained with RAL on publicly available ICIAR 2018 Breast Cancer Dataset (IBCD). The experimental results show that our RAL increases the slice-based accuracy of CNN from 93.75% to 96.25%.
We present an active learning architecture that allows a robot to actively learn which data collection strategy is most efficient for acquiring motor skills to achieve multiple outcomes, and generalise over its experience to achieve new outcomes. The robot explores its environment both via interactive learning and goal-babbling. It learns at the same time when, who and what to actively imitate from several available teachers, and learns when not to use social guidance but use active goal-oriented self-exploration. This is formalised in the framework of life-long strategic learning. The proposed architecture, called Socially Guided Intrinsic Motivation with Active Choice of Teacher and Strategy (SGIM-ACTS), relies on hierarchical active decisions of what and how to learn driven by empirical evaluation of learning progress for each learning strategy. We illustrate with an experiment where a simulated robot learns to control its arm for realising two kinds of different outcomes. It has to choose actively and hierarchically at each learning episode: 1) what to learn: which outcome is most interesting to select as a goal to focus on for goal-directed exploration; 2) how to learn: which data collection strategy to use among self-exploration, mimicry and emulation; 3) once he has decided when and what to imitate by choosing mimicry or emulation, then he has to choose who to imitate, from a set of different teachers. We show that SGIM-ACTS learns significantly more efficiently than using single learning strategies, and coherently selects the best strategy with respect to the chosen outcome, taking advantage of the available teachers (with different levels of skills).
Credit card fraud detection is a very challenging problem because of the specific nature of transaction data and the labeling process. The transaction data is peculiar because they are obtained in a streaming fashion, they are strongly imbalanced and prone to non-stationarity. The labeling is the outcome of an active learning process, as every day human investigators contact only a small number of cardholders (associated to the riskiest transactions) and obtain the class (fraud or genuine) of the related transactions. An adequate selection of the set of cardholders is therefore crucial for an efficient fraud detection process. In this paper, we present a number of active learning strategies and we investigate their fraud detection accuracies. We compare different criteria (supervised, semi-supervised and unsupervised) to query unlabeled transactions. Finally, we highlight the existence of an exploitation/exploration trade-off for active learning in the context of fraud detection, which has so far been overlooked in the literature.
We study Bayesian hypernetworks: a framework for approximate Bayesian inference in neural networks. A Bayesian hypernetwork $\h$ is a neural network which learns to transform a simple noise distribution, $p(\vec\epsilon) = \N(\vec 0,\mat I)$, to a distribution $q(\pp) := q(h(\vec\epsilon))$ over the parameters $\pp$ of another neural network (the "primary network")\@. We train $q$ with variational inference, using an invertible $\h$ to enable efficient estimation of the variational lower bound on the posterior $p(\pp | \D)$ via sampling. In contrast to most methods for Bayesian deep learning, Bayesian hypernets can represent a complex multimodal approximate posterior with correlations between parameters, while enabling cheap iid sampling of~$q(\pp)$. In practice, Bayesian hypernets can provide a better defense against adversarial examples than dropout, and also exhibit competitive performance on a suite of tasks which evaluate model uncertainty, including regularization, active learning, and anomaly detection.
modAL is a modular active learning framework for Python, aimed to make active learning research and practice simpler. Its distinguishing features are (i) clear and modular object oriented design (ii) full compatibility with scikit-learn models and workflows. These features make fast prototyping and easy extensibility possible, aiding the development of real-life active learning pipelines and novel algorithms as well. modAL is fully open source, hosted on GitHub at https://github.com/cosmic-cortex/modAL. To assure code quality, extensive unit tests are provided and continuous integration is applied. In addition, a detailed documentation with several tutorials are also available for ease of use. The framework is available in PyPI and distributed under the MIT license.
In this paper, we propose an active learning algorithm and models which can gradually learn individual's preference through pairwise comparisons. The active learning scheme aims at finding individual's most preferred choice with minimized number of pairwise comparisons. The pairwise comparisons are encoded into probabilistic models based on assumptions of choice models and deep Gaussian processes. The next-to-compare decision is determined by a novel acquisition function. We benchmark the proposed algorithm and models using functions with multiple local optima and one public airline itinerary dataset. The experiments indicate the effectiveness of our active learning algorithm and models.
Single-trial classification of event-related potentials in electroencephalogram (EEG) signals is a very important paradigm of brain-computer interface (BCI). Because of individual differences, usually some subject-specific calibration data are required to tailor the classifier for each subject. Transfer learning has been extensively used to reduce such calibration data requirement, by making use of auxiliary data from similar/relevant subjects/tasks. However, all previous research assumes that all auxiliary data have been labeled. This paper considers a more general scenario, in which part of the auxiliary data could be unlabeled. We propose active semi-supervised transfer learning (ASTL) for offline BCI calibration, which integrates active learning, semi-supervised learning, and transfer learning. Using a visual evoked potential oddball task and three different EEG headsets, we demonstrate that ASTL can achieve consistently good performance across subjects and headsets, and it outperforms some state-of-the-art approaches in the literature.
There are many important regression problems in real-world brain-computer interface (BCI) applications, e.g., driver drowsiness estimation from EEG signals. This paper considers offline analysis: given a pool of unlabeled EEG epochs recorded during driving, how do we optimally select a small number of them to label so that an accurate regression model can be built from them to label the rest? Active learning is a promising solution to this problem, but interestingly, to our best knowledge, it has not been used for regression problems in BCI so far. This paper proposes a novel enhanced batch-mode active learning (EBMAL) approach for regression, which improves upon a baseline active learning algorithm by increasing the reliability, representativeness and diversity of the selected samples to achieve better regression performance. We validate its effectiveness using driver drowsiness estimation from EEG signals. However, EBMAL is a general approach that can also be applied to many other offline regression problems beyond BCI.
Active learning is a machine learning approach for reducing the data labeling effort. Given a pool of unlabeled samples, it tries to select the most useful ones to label so that a model built from them can achieve the best possible performance. This paper focuses on pool-based sequential active learning for regression (ALR). We first propose three essential criteria that an ALR approach should consider in selecting the most useful unlabeled samples: informativeness, representativeness, and diversity, and compare four existing ALR approaches against them. We then propose a new ALR approach using passive sampling, which considers both the representativeness and the diversity in both the initialization and subsequent iterations. Remarkably, this approach can also be integrated with other existing ALR approaches in the literature to further improve the performance. Extensive experiments on 11 UCI, CMU StatLib, and UFL Media Core datasets from various domains verified the effectiveness of our proposed ALR approaches.
Human labeling of data can be very time-consuming and expensive, yet, in many cases it is critical for the success of the learning process. In order to minimize human labeling efforts, we propose a novel active learning solution that does not rely on existing sources of unlabeled data. It uses a small amount of labeled data as the core set for the synthesis of useful membership queries (MQs) - unlabeled instances generated by an algorithm for human labeling. Our solution uses modification operators, functions that modify instances to some extent. We apply the operators on a small set of instances (core set), creating a set of new membership queries. Using this framework, we look at the instance space as a search space and apply search algorithms in order to generate new examples highly relevant to the learner. We implement this framework in the textual domain and test it on several text classification tasks and show improved classifier performance as more MQs are labeled and incorporated into the training set. To the best of our knowledge, this is the first work on membership queries in the textual domain.
We introduce a conceptually simple and scalable framework for continual learning domains where tasks are learned sequentially. Our method is constant in the number of parameters and is designed to preserve performance on previously encountered tasks while accelerating learning progress on subsequent problems. This is achieved by training a network with two components: A knowledge base, capable of solving previously encountered problems, which is connected to an active column that is employed to efficiently learn the current task. After learning a new task, the active column is distilled into the knowledge base, taking care to protect any previously acquired skills. This cycle of active learning (progression) followed by consolidation (compression) requires no architecture growth, no access to or storing of previous data or tasks, and no task-specific parameters. We demonstrate the progress & compress approach on sequential classification of handwritten alphabets as well as two reinforcement learning domains: Atari games and 3D maze navigation.
When using active learning, smaller batch sizes are typically more efficient from a learning efficiency perspective. However, in practice due to speed and human annotator considerations, the use of larger batch sizes is necessary. While past work has shown that larger batch sizes decrease learning efficiency from a learning curve perspective, it remains an open question how batch size impacts methods for stopping active learning. We find that large batch sizes degrade the performance of a leading stopping method over and above the degradation that results from reduced learning efficiency. We analyze this degradation and find that it can be mitigated by changing the window size parameter of how many past iterations of learning are taken into account when making the stopping decision. We find that when using larger batch sizes, stopping methods are more effective when smaller window sizes are used.
This paper investigates and evaluates support vector machine active learning algorithms for use with imbalanced datasets, which commonly arise in many applications such as information extraction applications. Algorithms based on closest-to-hyperplane selection and query-by-committee selection are combined with methods for addressing imbalance such as positive amplification based on prevalence statistics from initial random samples. Three algorithms (ClosestPA, QBagPA, and QBoostPA) are presented and carefully evaluated on datasets for text classification and relation extraction. The ClosestPA algorithm is shown to consistently outperform the other two in a variety of ways and insights are provided as to why this is the case.
Standard myopic active learning assumes that human annotations are always obtainable whenever new samples are selected. This, however, is unrealistic in many real-world applications where human experts are not readily available at all times. In this paper, we consider the single shot setting: all the required samples should be chosen in a single shot and no human annotation can be exploited during the selection process. We propose a new method, Active Learning through Random Labeling (ALRL), which substitutes single human annotator for multiple, what we will refer to as, pseudo annotators. These pseudo annotators always provide uniform and random labels whenever new unlabeled samples are queried. This random labeling enables standard active learning algorithms to also exhibit the exploratory behavior needed for single shot active learning. The exploratory behavior is further enhanced by selecting the most representative sample via minimizing nearest neighbor distance between unlabeled samples and queried samples. Experiments on real-world datasets demonstrate that the proposed method outperforms several state-of-the-art approaches.