The NeurIPS 2023 Machine Learning for Audio Workshop brings together machine learning (ML) experts from various audio domains. There are several valuable audio-driven ML tasks, from speech emotion recognition to audio event detection, but the community is sparse compared to other ML areas, e.g., computer vision or natural language processing. A major limitation with audio is the available data; with audio being a time-dependent modality, high-quality data collection is time-consuming and costly, making it challenging for academic groups to apply their often state-of-the-art strategies to a larger, more generalizable dataset. In this short white paper, to encourage researchers with limited access to large-datasets, the organizers first outline several open-source datasets that are available to the community, and for the duration of the workshop are making several propriety datasets available. Namely, three vocal datasets, Hume-Prosody, Hume-VocalBurst, an acted emotional speech dataset Modulate-Sonata, and an in-game streamer dataset Modulate-Stream. We outline the current baselines on these datasets but encourage researchers from across audio to utilize them outside of the initial baseline tasks.

Domain generalization (DG) is an important problem that learns a model that can generalize to unseen test domains leveraging one or more source domains, under the assumption of shared label spaces. However, most DG methods assume access to abundant source data in the target label space, a requirement that proves overly stringent for numerous real-world applications, where acquiring the same label space as the target task is prohibitively expensive. For this setting, we tackle the multimodal version of the unsupervised domain generalization (UDG) problem, which uses a large task-agnostic unlabeled source dataset, such as LAION-2B during finetuning. Our framework does not explicitly assume any relationship between the source dataset and target task. Instead, it relies only on the premise that the source dataset can be efficiently searched in a joint vision-language space. For this multimodal UDG setting, we propose a novel method to build a small ($<$100K) subset of the source data in three simple steps: (1) diversified retrieval using label names as queries, (2) rank pseudo-labeling, and (3) clustering to find representative samples. To demonstrate the value of studying the multimodal UDG problem, we compare our results against state-of-the-art source-free DG and zero-shot (ZS) methods on their respective benchmarks and show up to 10% improvement in accuracy on 20 diverse target datasets. Additionally, our multi-stage dataset construction method achieves 3% improvement on average over nearest neighbors retrieval. Code is available: https://github.com/Chris210634/mudg

Recent advances in vision-language models have combined contrastive approaches with generative methods to achieve state-of-the-art (SOTA) on downstream inference tasks like zero-shot image classification. However, a persistent issue of these models for image classification is their out-of-distribution (OOD) generalization capabilities. We first show that when an OOD data point is misclassified, the correct class can be typically found in the Top-K predicted classes. In order to steer the model prediction toward the correct class within the top predicted classes, we propose the Image-Caption Encoding (ICE) method, a straightforward approach that directly enforces consistency between the image-conditioned and caption-conditioned predictions at evaluation time only. Intuitively, we take advantage of unique properties of the generated captions to guide our local search for the correct class label within the Top-K predicted classes. We show that our method can be easily combined with other SOTA methods to enhance Top-1 OOD accuracies by 0.5% on average and up to 3% on challenging datasets. Our code: https://github.com/Chris210634/ice

There is extensive interest in metric learning methods for image retrieval. Many metric learning loss functions focus on learning a correct ranking of training samples, but strongly overfit semantically inconsistent labels and require a large amount of data. To address these shortcomings, we propose a new metric learning method, called contextual loss, which optimizes contextual similarity in addition to cosine similarity. Our contextual loss implicitly enforces semantic consistency among neighbors while converging to the correct ranking. We empirically show that the proposed loss is more robust to label noise, and is less prone to overfitting even when a large portion of train data is withheld. Extensive experiments demonstrate that our method achieves a new state-of-the-art across four image retrieval benchmarks and multiple different evaluation settings. Code is available at: https://github.com/Chris210634/metric-learning-using-contextual-similarity

We present a new piecewise linear regression methodology that utilizes fitting a difference of convex functions (DC functions) to the data. These are functions $f$ that may be represented as the difference $\phi_1 - \phi_2$ for a choice of convex functions $\phi_1, \phi_2$. The method proceeds by estimating piecewise-liner convex functions, in a manner similar to max-affine regression, whose difference approximates the data. The choice of the function is regularised by a new seminorm over the class of DC functions that controls the $\ell_\infty$ Lipschitz constant of the estimate. The resulting methodology can be efficiently implemented via Quadratic programming even in high dimensions, and is shown to have close to minimax statistical risk. We empirically validate the method, showing it to be practically implementable, and to have comparable performance to existing regression/classification methods on real-world datasets.

These methods, known as Mahalanobis metric learning approaches, have been analyzed Classical linear metric learning methods have recently theoretically, are scalable, and usually involve convex optimization been extended along two distinct lines: problems that can be solved globally (Kulis, 2013; deep metric learning methods for learning embeddings Bellet et al., 2015). of the data using neural networks, and Classical metric learning methods have been extended along Bregman divergence learning approaches for extending various axes; two important directions are deep metric learning learning Euclidean distances to more general and Bregman divergence learning. Deep metric learning divergence measures such as divergences over approaches replace the linear mapping learned in Mahalanobis distributions. In this paper, we introduce deep metric learning methods with more general mappings Bregman divergences, which are based on learning that are learned via neural networks (Hoffer & Ailon, and parameterizing functional Bregman divergences 2015; Chopra et al., 2005). On the other hand, Bregman using neural networks, and which unify divergence methods replace the squared Euclidean distance and extend these existing lines of work. We show with arbitrary Bregman divergences (Bregman, 1967), and in particular how deep metric learning formulations, learn the underlying generating function of the Bregman kernel metric learning, Mahalanobis metric divergence via piecewise linear approximators (Siahkamari learning, and moment-matching functions for et al., 2019) or convex combinations of existing basis functions comparing distributions arise as special cases of (Wu et al., 2009).

In this paper we consider the fundamental problem of semi-supervised kernel function learning. We propose a general regularized framework for learning a kernel matrix, and then demonstrate an equivalence between our proposed kernel matrix learning framework and a general linear transformation learning problem. Our result shows that the learned kernel matrices parameterize a linear transformation kernel function and can be applied inductively to new data points. Furthermore, our result gives a constructive method for kernelizing most existing Mahalanobis metric learning formulations. To make our results practical for large-scale data, we modify our framework to limit the number of parameters in the optimization process.

Despite achieving remarkable success in various domains, recent studies have uncovered the vulnerability of deep neural networks to adversarial perturbations, creating concerns on model generalizability and new threats such as prediction-evasive misclassification or stealthy reprogramming. Among different defense proposals, stochastic network defenses such as random neuron activation pruning or random perturbation to layer inputs are shown to be promising for attack mitigation. However, one critical drawback of current defenses is that the robustness enhancement is at the cost of noticeable performance degradation on legitimate data, e.g., large drop in test accuracy. This paper is motivated by pursuing for a better trade-off between adversarial robustness and test accuracy for stochastic network defenses. We propose Defense Efficiency Score (DES), a comprehensive metric that measures the gain in unsuccessful attack attempts at the cost of drop in test accuracy of any defense. To achieve a better DES, we propose hierarchical random switching (HRS), which protects neural networks through a novel randomization scheme. A HRS-protected model contains several blocks of randomly switching channels to prevent adversaries from exploiting fixed model structures and parameters for their malicious purposes. Extensive experiments show that HRS is superior in defending against state-of-the-art white-box and adaptive adversarial misclassification attacks. We also demonstrate the effectiveness of HRS in defending adversarial reprogramming, which is the first defense against adversarial programs. Moreover, in most settings the average DES of HRS is at least 5X higher than current stochastic network defenses, validating its significantly improved robustness-accuracy trade-off.

Metric learning is the problem of learning a task-specific distance function given supervision. Classical linear methods for this problem (known as Mahalanobis metric learning approaches) are well-studied both theoretically and empirically, but are limited to Euclidean distances after learned linear transformations of the input space. In this paper, we consider learning a Bregman divergence, a rich and important class of divergences that includes Mahalanobis metrics as a special case but also includes the KL-divergence and others. We develop a formulation and algorithm for learning arbitrary Bregman divergences based on approximating their underlying convex generating function via a piecewise linear function. We show several theoretical results of our resulting model, including a PAC guarantee that the learned Bregman divergence approximates an arbitrary Bregman divergence with error O_p (m^(-1/(d+2))), where m is the number of training points and d is the dimension of the data. We provide empirical results on using the learned divergences for classification, semi-supervised clustering, and ranking problems.

Existing automatic music generation approaches that feature deep learning can be broadly classified into two types: raw audio models and symbolic models. Symbolic models, which train and generate at the note level, are currently the more prevalent approach; these models can capture long-range dependencies of melodic structure, but fail to grasp the nuances and richness of raw audio generations. Raw audio models, such as DeepMind's WaveNet, train directly on sampled audio waveforms, allowing them to produce realistic-sounding, albeit unstructured music. In this paper, we propose an automatic music generation methodology combining both of these approaches to create structured, realistic-sounding compositions. We consider a Long Short Term Memory network to learn the melodic structure of different styles of music, and then use the unique symbolic generations from this model as a conditioning input to a WaveNet-based raw audio generator, creating a model for automatic, novel music. We then evaluate this approach by showcasing results of this work.