Comparing and benchmarking [10] is commonly used for self-supervised pre-training, as different DNN models can often be tedious due to the use of well as a benchmarking method for audio event classification [6, 7, different datasets and evaluation protocols. To facilitate the process, 11]. A recently proposed HEAR challenge [12] focuses on evaluating here, we present the Speech Emotion Recognition Adaptation general-purpose audio representations and extends the concept Benchmark (SERAB), a framework for evaluating the performance underlying AudioSet by including additional tasks. In speech representation and generalization capacity of different approaches for utterancelevel learning, NOSS [6] was recently proposed as a platform SER. The benchmark is composed of nine datasets for SER for evaluating speech-specific feature extractors. It includes diverse in six languages. Since the datasets have different sizes and numbers non-semantic speech processing problems, such as speaker and language of emotional classes, the proposed setup is particularly suitable identification, as well as two SER tasks (CREMA-D [13] and for estimating the generalization capacity of pre-trained DNN-based SAVEE [14]).

It is well known that hyperspectral images (HSI) contain rich spatial-spectral contextual information, and how to effectively combine both spectral and spatial information using DNN for HSI classification has become a new research hotspot. Compared with CNN with square kernels, GCN have exhibited exciting potential to model spatial contextual structure and conduct flexible convolution on arbitrarily irregular image regions. However, current GCN only using first-order spectral-spatial signatures can result in boundary blurring and isolated misclassification. To address these, we first designed the graph-based second-order pooling (GSOP) operation to obtain contextual nodes information in non-Euclidean space for GCN. Further, we proposed a novel multi-stream hybridized convolutional network (MSHCNet) with combination of first and second order statistics in Euclidean/non-Euclidean spaces to learn and fuse multi-view complementary information to segment HSIs. Specifically, our MSHCNet adopted four parallel streams, which contained G-stream, utilizing the irregular correlation between adjacent land covers in terms of first-order graph in non-Euclidean space; C-stream, adopting convolution operator to learn regular spatial-spectral features in Euclidean space; N-stream, combining first and second order features to learn representative and discriminative regular spatial-spectral features of Euclidean space; S-stream, using GSOP to capture boundary correlations and obtain graph representations from all nodes in graphs of non-Euclidean space. Besides, these feature representations learned from four different streams were fused to integrate the multi-view complementary information for HSI classification. Finally, we evaluated our proposed MSHCNet on three hyperspectral datasets, and experimental results demonstrated that our method significantly outperformed state-of-the-art eight methods.

In this paper, we provide the first focused study on the discontinuities (aka. When investigating latent holes, existing works are exclusively centred around the encoder network and they merely explore the existence of holes. We tackle these limitations by proposing a highly efficient Tree-based Decoder-Centric (TDC) algorithm for latent hole identification, with a focal point on the text domain. In contrast to past studies, our approach pays attention to the decoder network, as a decoder has a direct impact on the model's output quality. Furthermore, we provide, for the first time, in-depth empirical analysis of the latent hole phenomenon, investigating several important aspects such as how the holes impact VAE algorithms' performance on text generation, and how the holes are distributed in the latent space.

Many practical problems need the output of a machine learning model to satisfy a set of constraints, $K$. Nevertheless, there is no known guarantee that classical neural network architectures can exactly encode constraints while simultaneously achieving universality. We provide a quantitative constrained universal approximation theorem which guarantees that for any non-convex compact set $K$ and any continuous function $f:\mathbb{R}^n\rightarrow K$, there is a probabilistic transformer $\hat{F}$ whose randomized outputs all lie in $K$ and whose expected output uniformly approximates $f$. Our second main result is a "deep neural version" of Berge's Maximum Theorem (1963). The result guarantees that given an objective function $L$, a constraint set $K$, and a family of soft constraint sets, there is a probabilistic transformer $\hat{F}$ that approximately minimizes $L$ and whose outputs belong to $K$; moreover, $\hat{F}$ approximately satisfies the soft constraints. Our results imply the first universal approximation theorem for classical transformers with exact convex constraint satisfaction. They also yield that a chart-free universal approximation theorem for Riemannian manifold-valued functions subject to suitable geodesically convex constraints.

Engineering design problems often involve large state and action spaces along with highly sparse rewards. Since an exhaustive search of those spaces is not feasible, humans utilize relevant domain knowledge to condense the search space. Previously, deep learning agents (DLAgents) were introduced to use visual imitation learning to model design domain knowledge. This note builds on DLAgents and integrates them with one-step lookahead search to develop goal-directed agents capable of enhancing learned strategies for sequentially generating designs. Goal-directed DLAgents can employ human strategies learned from data along with optimizing an objective function. The visual imitation network from DLAgents is composed of a convolutional encoder-decoder network, acting as a rough planning step that is agnostic to feedback. Meanwhile, the lookahead search identifies the fine-tuned design action guided by an objective. These design agents are trained on an unconstrained truss design problem that is modeled as a sequential, action-based configuration design problem. The agents are then evaluated on two versions of the problem: the original version used for training and an unseen constrained version with an obstructed construction space. The goal-directed agents outperform the human designers used to train the network as well as the previous objective-agnostic versions of the agent in both scenarios. This illustrates a design agent framework that can efficiently use feedback to not only enhance learned design strategies but also adapt to unseen design problems.

Question Answering (QA) has been a long-standing research topic in AI and NLP fields, and a wealth of studies have been conducted to attempt to equip QA systems with human-level reasoning capability. To approximate the complicated human reasoning process, state-of-the-art QA systems commonly use pre-trained language models (LMs) to access knowledge encoded in LMs together with elaborately designed modules based on Graph Neural Networks (GNNs) to perform reasoning over knowledge graphs (KGs). However, many problems remain open regarding the reasoning functionality of these GNN-based modules. Can these GNN-based modules really perform a complex reasoning process? Are they under- or over-complicated for QA? To open the black box of GNN and investigate these problems, we dissect state-of-the-art GNN modules for QA and analyze their reasoning capability. We discover that even a very simple graph neural counter can outperform all the existing GNN modules on CommonsenseQA and OpenBookQA, two popular QA benchmark datasets which heavily rely on knowledge-aware reasoning. Our work reveals that existing knowledge-aware GNN modules may only carry out some simple reasoning such as counting. It remains a challenging open problem to build comprehensive reasoning modules for knowledge-powered QA.

Transforming large deep neural network (DNN) models into the multi-exit architectures can overcome the overthinking issue and distribute a large DNN model on resource-constrained scenarios (e.g. IoT frontend devices and backend servers) for inference and transmission efficiency. Nevertheless, intellectual property (IP) protection for the multi-exit models in the wild is still an unsolved challenge. Previous efforts to verify DNN model ownership mainly rely on querying the model with specific samples and checking the responses, e.g., DNN watermarking and fingerprinting. However, they are vulnerable to adversarial settings such as adversarial training and are not suitable for the IP verification for multi-exit DNN models. In this paper, we propose a novel approach to fingerprint multi-exit models via inference time rather than inference predictions. Specifically, we design an effective method to generate a set of fingerprint samples to craft the inference process with a unique and robust inference time cost as the evidence for model ownership. We conduct extensive experiments to prove the uniqueness and robustness of our method on three structures (ResNet-56, VGG-16, and MobileNet) and three datasets (CIFAR-10, CIFAR-100, and Tiny-ImageNet) under comprehensive adversarial settings.

Detection of common events and scenes from audio is useful for extracting and understanding human contexts in daily life. Prior studies have shown that leveraging knowledge from a relevant domain is beneficial for a target acoustic event detection (AED) process. Inspired by the observation that many human-centered acoustic events in daily life involve voice elements, this paper investigates the potential of transferring high-level voice representations extracted from a public speaker dataset to enrich an AED pipeline. Towards this end, we develop a dual-branch neural network architecture for the joint learning of voice and acoustic features during an AED process and conduct thorough empirical studies to examine the performance on the public AudioSet [1] with different types of inputs. Our main observations are that: 1) Joint learning of audio and voice inputs improves the AED performance (mean average precision) for both a CNN baseline (0.292 vs 0.134 mAP) and a TALNet [2] baseline (0.361 vs 0.351 mAP); 2) Augmenting the extra voice features is critical to maximize the model performance with dual inputs.

Recently, emotional speech synthesis has achieved remarkable performance. The emotion strength of synthesized speech can be controlled flexibly using a strength descriptor, which is obtained by an emotion attribute ranking function. However, a trained ranking function on specific data has poor generalization, which limits its applicability for more realistic cases. In this paper, we propose a deep learning based emotion strength assessment network for strength prediction that is referred to as StrengthNet. Our model conforms to a multi-task learning framework with a structure that includes an acoustic encoder, a strength predictor and an auxiliary emotion predictor. A data augmentation strategy was utilized to improve the model generalization. Experiments show that the predicted emotion strength of the proposed StrengthNet are highly correlated with ground truth scores for seen and unseen speech. Our codes are available at: https://github.com/ttslr/StrengthNet.

Transformer-based models have achieved great success in various NLP, vision, and speech tasks. However, the core of Transformer, the self-attention mechanism, has a quadratic time and memory complexity with respect to the sequence length, which hinders applications of Transformer-based models to long sequences. Many approaches have been proposed to mitigate this problem, such as sparse attention mechanisms, low-rank matrix approximations and scalable kernels, and token mixing alternatives to self-attention. We propose a novel Pooling Network (PoNet) for token mixing in long sequences with linear complexity. We design multi-granularity pooling and pooling fusion to capture different levels of contextual information and combine their interactions with tokens. On the Long Range Arena benchmark, PoNet significantly outperforms Transformer and achieves competitive accuracy, while being only slightly slower than the fastest model, FNet, across all sequence lengths measured on GPUs. We also conduct systematic studies on the transfer learning capability of PoNet and observe that PoNet achieves 96.0% of the accuracy of BERT on the GLUE benchmark, outperforming FNet by 4.5% relative. Comprehensive ablation analysis demonstrates effectiveness of the designed multi-granularity pooling and pooling fusion for token mixing in long sequences and efficacy of the designed pre-training tasks for PoNet to learn transferable contextualized language representations.