The neural transducer is an end-to-end model for automatic speech recognition (ASR). While the model is well-suited for streaming ASR, the training process remains challenging. During training, the memory requirements may quickly exceed the capacity of state-of-the-art GPUs, limiting batch size and sequence lengths. In this work, we analyze the time and space complexity of a typical transducer training setup. We propose a memory-efficient training method that computes the transducer loss and gradients sample by sample. We present optimizations to increase the efficiency and parallelism of the…Apple Machine Learning Research
Continuous Pseudo-Labeling from the Start
Self-training (ST), or pseudo-labeling has sparked significant interest in the automatic speech recognition (ASR) community recently because of its success in harnessing unlabeled data. Unlike prior semi-supervised learning approaches that relied on iteratively regenerating pseudo-labels (PLs) from a trained model and using them to train a new model, recent state-of-the-art methods perform ‘continuous training’ where PLs are generated using a very recent version of the model being trained. Nevertheless, these approaches still rely on bootstrapping the ST using an initial supervised learning…Apple Machine Learning Research
FastFill: Efficient Compatible Model Update
*= Equal Contributors
In many retrieval systems the original high dimensional data (e.g., images) is mapped to a lower dimensional feature through a learned embedding model. The task of retrieving the most similar data from a gallery set to a given query data is performed through similarity comparison on features. When the embedding model is updated, it might produce features that are not comparable/compatible with features already in the gallery computed with the old model. Subsequently, all features in the gallery need to be re-computed using the new embedding model — a computationally…Apple Machine Learning Research
Loss minimization through the lens of outcome indistinguishability
We present a new perspective on loss minimization and the recent notion of Omniprediction through the lens of Outcome Indistingusihability. For a collection of losses and hypothesis class, omniprediction requires that a predictor provide a loss-minimization guarantee simultaneously for every loss in the collection compared to the best (loss-specific) hypothesis in the class. We present a generic template to learn predictors satisfying a guarantee we call Loss Outcome Indistinguishability. For a set of statistical tests–based on a collection of losses and hypothesis class–a predictor is Loss…Apple Machine Learning Research
A Unifying Theory of Distance from Calibration
We study the fundamental question of how to define and measure the distance from calibration for probabilistic predictors. While the notion of perfect calibration is well-understood, there is no consensus on how to quantify the distance from perfect calibration. Numerous calibration measures have been proposed in the literature, but it is unclear how they compare to each other, and many popular measures such as Expected Calibration Error (ECE) fail to satisfy basic properties like continuity.
We present a rigorous framework for analyzing calibration measures, inspired by the literature on…Apple Machine Learning Research
Improving Human Annotation Effectiveness for Fact Collection by Identifying the Most Relevant Answers
This paper was accepted at the Workshops on Data Science with Human in the Loop at EMNLP 2022
Identifying and integrating missing facts is a crucial task for knowledge graph completion to ensure robustness towards downstream applications such as question answering. Adding new facts to a knowledge graph in real world system often involves human verification effort, where candidate facts are verified for accuracy by human annotators. This process is labor-intensive, time-consuming, and inefficient since only a small number of missing facts can be identified. This paper proposes a simple but…Apple Machine Learning Research
Designing Data: Proactive Data Collection and Iteration for Machine Learning
Lack of diversity in data collection has caused significant failures in machine learning (ML) applications. While ML developers perform post-collection interventions, these are time intensive and rarely comprehensive. Thus, new methods to track and manage data collection, iteration, and model training are necessary for evaluating whether datasets reflect real world variability. We present designing data, an iterative, bias mitigating approach to data collection connecting HCI concepts with ML techniques. Our process includes (1) Pre-Collection Planning, to reflexively prompt and document…Apple Machine Learning Research
Symbol Guided Hindsight Priors for Reward Learning from Human Preferences
This paper was accepted at the “Human in the Loop Learning Workshop” at NeurIPS 2022.
Specification of reward functions for Reinforcement Learning is a challenging task which is bypassed by the framework of Preference Based Learning methods which instead learn from preference labels on trajectory queries. These methods, however, still suffer from high requirements of preference labels and often would still achieve low reward recovery. We present the PRIOR framework that alleviates the issues of impractical number of queries to humans as well as poor reward recovery through computing priors…Apple Machine Learning Research
RangeAugment: Efficient Online Augmentation with Range Learning
State-of-the-art automatic augmentation methods (e.g., AutoAugment and RandAugment) for visual recognition tasks diversify training data using a large set of augmentation operations. The range of magnitudes of many augmentation operations (e.g., brightness and contrast) is continuous. Therefore, to make search computationally tractable, these methods use fixed and manually-defined magnitude ranges for each operation, which may lead to sub-optimal policies. To answer the open question on the importance of magnitude ranges for each augmentation operation, we introduce RangeAugment that allows us…Apple Machine Learning Research
Supervised Training of Conditional Monge Maps
Optimal transport (OT) theory describes general principles to define and select, among many possible choices, the most efficient way to map a probability measure onto another. That theory has been mostly used to estimate, given a pair of source and target probability measures , a parameterized map that can efficiently map onto . In many applications, such as predicting cell responses to treatments, the data measures (features of untreated/treated cells) that define optimal transport problems do not arise in isolation but are associated with a context (the treatment). To account for and…Apple Machine Learning Research