United States

Multimodal processing has attracted much attention lately especially with the success of pre-training. However, the exploration has mainly focused on vision-language pre-training, as introducing more modalities can greatly complicate model design and optimization. In this paper, we extend the state-of-the-art Vision-Language model CLIP to accommodate the audio modality for Vision-Language-Audio multimodal processing. Specifically, we apply inter-modal and intra-modal
contrastive learning to explore the correlation between audio and other modalities in addition to the inner characteristics of the audio modality. Moreover, we further design an audio type token to dynamically learn different audio information
type for different scenarios, as both verbal and nonverbal heterogeneous information is conveyed in general audios. Our proposed CLIP4VLA model is validated in different downstream tasks including video retrieval and video captioning, and achieves the state-of-the-art performance on the benchmark datasets of MSR-VTT, VATEX, and Audiocaps.

AAAI 2023

Accommodating Audio Modality in CLIP for Multimodal Processing

multi-modal pre-training

video understanding

representation learning

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-23 is the Thirty-Seventh AAAI Conference on Artificial Intelligence. The theme of this conference is to create collaborative bridges within and beyond AI. Like previous AAAI conferences, AAAI-23 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and two new activities: a Bridge Program and a Lab Program. Many of these activities are tailored to the theme of bridges and all are selected according to the highest standards, with additional programs for students and young researchers. 
AAAI is providing you with a conference planner, which you can use to help organize your itinerary of activities. This includes talks to attend in person, talks to attend remotely, breaks with colleagues and your site seeing activities. To access this conference planner, please go to [https://aaai-2023.takemobi.io](https://aaai-2023.takemobi.io).

In order to access this site, you need to register. If you haven't already, please register [here](https://aaai.org/Conferences/AAAI-23/registration/).


The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines.

poster

The Area Under the ROC Curve (AUC) is an important model metric for evaluating binary classifiers, and many algorithms have been proposed to optimize AUC approximately. It raises the question of whether the generally insignificant gains observed by previous studies are due to inherent limitations of the metric or the inadequate quality of optimization.

To better understand the value of optimizing for AUC, we present an efficient algorithm, namely AUC-opt, to find the provably optimal AUC linear classifier in $\mathbb{R}^2$, which runs in $\mathcal{O}(n_+ n_- \log (n_+ n_-))$ where $n_+$ and $n_-$ are the number of positive and negative samples respectively. Furthermore, it can be naturally extended to $\mathbb{R}^d$ in $\mathcal{O}((n_+n_-)^{d-1}\log (n_+n_-))$ by calling AUC-opt in lower-dimensional spaces recursively. We prove the problem is NP-complete when $d$ is not fixed, reducing from the \textit{open hemisphere problem}.

Experiments show that compared with other methods, AUC-opt achieves statistically significant improvements on between 17 to 40 in $\mathbb{R}^2$ and on between 4 to 42 in $\mathbb{R}^3$ of 50 t-SNE training datasets. However, generally the gain proves insignificant on most testing datasets compared to the best standard classifiers. Similar observations are found for nonlinear AUC methods under real-world datasets.

Does it pay to optimize AUC?

Evaluating an explanation's faithfulness is desired for many reasons such as trust, interpretability and diagnosing the sources of model's errors. In this work, which focuses on the NLI task, we introduce the methodology of Faithfulness-through-Counterfactuals, which first generates a counterfactual hypothesis based on the logical predicates expressed in the explanation, and then evaluates if the model's prediction on the counterfactual is consistent with that expressed logic (i.e. if the new formula is \textit{logically satisfiable}). In contrast to existing approaches, this does not require any explanations for training a separate verification model. We first validate the efficacy of automatic counterfactual hypothesis generation, leveraging on the few-shot priming paradigm. Next, we show that our proposed metric distinguishes between human-model agreement and disagreement on new counterfactual input. In addition, we conduct a sensitivity analysis to validate that our metric is sensitive to unfaithful explanations.  

Logical Satisfiability of Counterfactuals for Faithful Explanations in NLI

    Owing to the prohibitive costs of generating large amounts of labeled data, weakly supervised learning is a growing paradigm within machine learning. In this setting, users design heuristics, or weak labelers, that label subsets of the data. These weak labels are combined (typically via a graphical model) to form pseudolabels, which are then used to train a downstream machine learning model. In this work, we question a foundational premise of the typical weakly supervised learning pipeline: given that the heuristic provides all ``label" information, why do we need to generate pseudolabels at all? Instead, we propose to directly transform the heuristics themselves into corresponding loss functions that penalize differences between our model and the heuristic. Indeed, by constructing losses directly from the heuristics, we can incorporate more information than is used in the standard weakly supervised pipeline, such as information on how the heuristics make their decisions; e.g, this can inform feature selection or gradient information during the training process. We call our method Losses over Labels (LoL), as it focuses on transforming heuristics directly into loss functions without going through the intermediate step of a label. We show that LoL improves upon existing weak supervision methods, on several benchmark image and text classification tasks.

Losses over Labels: Weakly Supervised Learning via Direct Loss Construction

As a novel distributed learning paradigm, federated learning (FL) faces serious challenges in dealing with massive clients with heterogeneous data distribution and computation and communication resources. Various client-variance-reduction schemes and client sampling strategies have been respectively introduced to improve the robustness of FL. Among others, primal-dual algorithms such as the alternating direction of method multipliers (ADMM) have been found being resilient to data distribution and outperform most of the primal-only FL algorithms. However, the reason behind remains a mystery still. In this paper, we firstly reveal the fact that the federated ADMM is essentially a client-variance-reduced algorithm. While this explains the inherent robustness of federated ADMM, the vanilla version of it lacks the ability to be adaptive to the degree of client heterogeneity. Besides, the global model at the server under client sampling is biased which slows down the practical convergence. To go beyond ADMM, we propose a novel primal-dual FL algorithm, termed FedVRA, that allows one to adaptively control the variance-reduction level and biasness of the global model. In addition, FedVRA unifies several representative FL algorithms in the sense that they are either special instances of FedVRA or are close to it. Extensions of FedVRA to semi/un-supervised learning are also presented. Experiments based on (semi-)supervised image classification tasks demonstrate superiority of FedVRA over the existing schemes in learning scenarios with massive heterogeneous clients and client sampling.

Beyond ADMM: A Unified Client-variance-reduced Adaptive Federated Learning Framework

Training ML models which are fair across different demographic groups is of critical importance due to the increased integration of ML in crucial decision-making scenarios such as healthcare and recruitment. Federated learning has been viewed as a promising solution for collaboratively training machine learning models among multiple parties while maintaining their local data privacy. However, federated learning also poses new challenges in mitigating the potential bias against certain populations (e.g., demographic groups), as this typically requires centralized access to the sensitive information (e.g., race, gender) of each datapoint. Motivated by the importance and challenges of group fairness in federated learning, in this work, we propose FairFed, a novel algorithm for fairness-aware aggregation to enhance group fairness in federated learning. Our proposed approach is server-side and agnostic to the applied local debiasing thus allowing for flexible use of different local debiasing methods across clients. We evaluate FairFed empirically versus common baselines for fair ML and federated learning and demonstrate that it provides fairer models, particularly under highly heterogeneous data distributions across clients. We also demonstrate the benefits of FairFed in scenarios involving naturally distributed real-life data collected from different geographical locations or departments within an organization.

FairFed: Enabling Group Fairness in Federated Learning

Pre-trained language models (PLMs) have recently enabled rapid progress on sentiment classification under the pre-train and fine-tune paradigm, where the fine-tuning phase aims to transfer the factual knowledge learned by PLMs to sentiment classification. However, current fine-tuning methods ignore the risk that PLMs cause the problem of sentiment bias, that is, PLMs tend to inject positive or negative sentiment from the contextual information of certain entities (or aspects) into their word embeddings, leading them to establish spurious correlations with labels. In this paper, we propose an adaptive Gumbel-attacked classifier that immunes sentiment bias from an adversarial-attack perspective. Due to the complexity and diversity of sentiment bias, we construct multiple Gumbel-attack expert networks to generate various noises from mixed Gumbel distribution constrained by mutual information minimization, and design an adaptive training framework to synthesize complex noise by confidence-guided controlling the number of expert networks. Finally, we capture these noises that effectively simulate sentiment bias based on the feedback of the classifier, and then propose a multi-channel parameter updating algorithm to strengthen the classifier to recognize these noises by fusing the parameters between the classifier and each expert network. Experimental results illustrate that our method significantly reduced sentiment bias and improved the performance of sentiment classification.

Reducing Sentiment Bias in Pre-trained Sentiment Classification via Adaptive Gumbel Attack

Even though data is abundant, it is often subjected to some form of censoring or truncation which inherently creates biases. Removing such biases and performing parameter estimation is a classical challenge in Statistics. In this paper, we focus on the problem of estimating the means of a mixture of two balanced $d$-dimensional Gaussians when the samples are prone to truncation. A recent theoretical study on the performance of the Expectation-Maximization (EM) algorithm for the aforementioned problem showed EM almost surely converges for $d=1$ and exhibits local convergence for $d>1$ to the true means. Nevertheless, the EM algorithm for the case of truncated mixture of two Gaussians is not easy to implement as it requires solving a set of \emph{nonlinear} equations at every iteration which makes the algorithm impractical. In this work, we propose a gradient based variant of the EM algorithm that has global convergence guarantees when $d=1$ and local convergence for $d>1$ to the true means. Moreover, the update rule at every iteration is easy to compute which makes the proposed method practical. We also provide numerous experiments to obtain more insights into the effect of truncation on the convergence to the true parameters in high dimensions.

Mean estimation of truncated mixtures of two Gaussians: A gradient based approach

The white balance methods for sRGB images (sRGB-WB) aim to directly remove their color temperature shifts. Despite achieving promising white balance (WB) performance, the existing methods suffer from WB instability, i.e., their results are inconsistent for images with different color temperatures. We propose a stable white balance network (SWBNet) to alleviate this problem. It learns the color temperature-insensitive features to generate white-balanced images, resulting in consistent WB results. Speciﬁcally, the color temperature-insensitive features are learned by implicitly suppressing low-frequency information sensitive to color temperatures. Then, a color temperature contrastive loss is introduced to facilitate the most information shared among features of the same scene and different color temperatures. This way, features from the same scene are more insensitive to color temperatures regardless of the inputs. We also present a color temperature sensitivity-oriented transformer that globally perceives multiple color temperature shifts within an image and corrects them by different weights. It helps to improve the accuracy of stabilized SWBNet, especially for multi-illumination sRGB images. Experiments indicate that our SWBNet achieves stable and remarkable WB performance.

SWBNet: A Stable White Balance Network for sRGB Images

Pre-trained language models (LMs) store knowledge in their parameters and can generate informative responses when used in conversational systems.
However, LMs suffer from the problem of "hallucination:" they may generate plausible-looking statements that are irrelevant or factually incorrect.
To address this problem, we propose a contrastive learning scheme, namely MixCL.
A novel mixed contrastive objective is proposed to explicitly optimize the implicit knowledge elicitation process of LMs, and thus reduce their hallucination in conversations.
We also examine negative sampling strategies of retrieved hard negatives and model-generated negatives.
We conduct experiments on Wizard-of-Wikipedia, a public open-domain knowledge-grounded dialogue benchmark, and assess the effectiveness of MixCL.
MixCL effectively reduces the hallucination of LMs in conversations and achieves the highest performance among LM-based dialogue agents in terms of relevancy and factuality.
We show that MixCL achieves comparable performance to the state-of-the-art KB-based approaches while enjoying notable advantages in terms of efficiency and scalability.

Contrastive Learning Reduces Hallucination in Conversations

We study competition among contests in a general model that allows for an arbitrary and heterogeneous space of contest design, where the goal of the contest designers is to maximize the contestants' sum of efforts. Our main result shows that optimal contests in the monopolistic setting (i.e., those that maximize the sum of efforts in a model with a single contest) form an equilibrium in the model with competition among contests. Under a very natural assumption these contests are in fact dominant, and the equilibria that they form are unique. Moreover, equilibria with the optimal contests are Pareto-optimal even in cases where other equilibria emerge. In many natural cases, they also maximize the social welfare.

Premium content

Accommodating Audio Modality in CLIP for Multimodal Processing

Downloads

Next from AAAI 2023

Does it pay to optimize AUC?

Similar lecture

Self-supervised Video Representation Learning via Latent Time Navigation