Active learning (AL) integrates data labeling and model training to minimize the labeling cost by prioritizing the selection of high value data that can best improve model performance. Readily-available unlabeled data are used for selection mechanisms, but are not used for model training in most conventional pool-based AL methods. To minimize the labeling cost, we unify unlabeled sample selection and model training based on two principles. First, we exploit both labeled and unlabeled data using semi-supervised learning (SSL) to distill information from unlabeled data that improves representation learning and sample selection. Second, we propose a simple yet effective selection metric that is coherent with the training objective such that the selected samples are effective at improving model performance. Experimental results demonstrate superior performance of our proposed principles for limited labeled data compared to alternative AL and SSL combinations. In addition, we study an important problem -- "When can we start AL?". We propose a measure that is empirically correlated with the AL target loss and can be used to assist in determining the proper start point.
Collecting large-scale data with clean labels for supervised training of neural networks is practically challenging. Although noisy labels are usually cheap to acquire, existing methods suffer severely for training datasets with high noise ratios, making high-cost human labeling a necessity. Here we present a method to train neural networks in a way that is almost invulnerable to severe label noise by utilizing a tiny trusted set. Our method, named IEG, is based on three key insights: (i) Isolation of noisy labels, (ii) Escalation of useful supervision from mislabeled data, and (iii) Guidance from small trusted data. On CIFAR100 with a 40% uniform noise ratio and 10 trusted labeled data per class, our method achieves $80.2{\pm}0.3\%$ classification accuracy, only 1.4% higher error than a neural network trained without label noise. Moreover, increasing the noise ratio to 80%, our method still achieves a high accuracy of $75.5{\pm}0.2\%$, compared to the previous best 47.7%. Finally, our method sets new state of the art on various types of challenging label corruption types and levels and large-scale WebVision benchmarks.
Understanding black-box machine learning models is important towards their widespread adoption. However, developing globally interpretable models that explain the behavior of the entire model is challenging. An alternative approach is to explain black-box models through explaining individual prediction using a locally interpretable model. In this paper, we propose a novel method for locally interpretable modeling - Reinforcement Learning-based Locally Interpretable Modeling (RL-LIM). RL-LIM employs reinforcement learning to select a small number of samples and distill the black-box model prediction into a low-capacity locally interpretable model. Training is guided with a reward that is obtained directly by measuring agreement of the predictions from the locally interpretable model with the black-box model. RL-LIM near-matches the overall prediction performance of black-box models while yielding human-like interpretability, and significantly outperforms state of the art locally interpretable models in terms of overall prediction performance and fidelity.
We propose a novel high-performance interpretable deep tabular data learning network, TabNet. TabNet utilizes a sequential attention mechanism that softly selects features to reason from at each decision step and then aggregates the processed information to make a final prediction decision. By explicitly selecting sparse features, TabNet learns very efficiently as the model capacity at each decision step is fully utilized for the most relevant features, resulting in a high performance model. This sparsity also enables more interpretable decision making through the visualization of feature selection masks. We demonstrate that TabNet outperforms other neural network and decision tree variants on a wide range of tabular data learning datasets and yields interpretable feature attributions and insights into the global model behavior.
Quantifying the value of data is a fundamental problem in machine learning. Data valuation has multiple important use cases: (1) building insights about the learning task, (2) domain adaptation, (3) corrupted sample discovery, and (4) robust learning. To adaptively learn data values jointly with the target task predictor model, we propose a meta learning framework which we name Data Valuation using Reinforcement Learning (DVRL). We employ a data value estimator (modeled by a deep neural network) to learn how likely each datum is used in training of the predictor model. We train the data value estimator using a reinforcement signal of the reward obtained on a small validation set that reflects performance on the target task. We demonstrate that DVRL yields superior data value estimates compared to alternative methods across different types of datasets and in a diverse set of application scenarios. The corrupted sample discovery performance of DVRL is close to optimal in many regimes (i.e. as if the noisy samples were known apriori), and for domain adaptation and robust learning DVRL significantly outperforms state-of-the-art by 14.6% and 10.8%, respectively.
We propose a novel framework, learning to transfer learn (L2TL), to improve transfer learning on a target dataset by judicious extraction of information from a source dataset. Our framework considers joint optimization of strongly-shared weights between models of source and target tasks, and employs adaptive weights for scaling of constituent loss terms. The adaptation of the weights is done using a reinforcement learning (RL)-based policy model, which is guided based on a performance metric on the target validation set. We demonstrate state-of-the-art performance of L2TL given fixed models, consistently outperforming fine-tuning baselines on various datasets. In addition, in the regimes of small-scale target datasets and significant label mismatch between source and target datasets, L2TL outperforms previous methods by a large margin.
We propose a new framework for prototypical learning that bases decision-making on few relevant examples that we call prototypes. Our framework utilizes an attention mechanism that relates the encoded representations to determine the prototypes. This results in a model that: (1) enables interpretability by outputting samples most relevant to the decision-making in addition to outputting the classification results; (2) allows confidence-controlled prediction by quantifying the mismatch across prototype labels; (3) permits detection of distribution mismatch; and (4) improves robustness to label noise. We demonstrate that our model is able to maintain comparable performance to baseline models while enabling all these benefits.