Multimodal sentiment analysis is a trending area of research, and the multimodal fusion is one of its most active topic. Acknowledging humans communicate through a variety of channels (i.e visual, acoustic, linguistic), multimodal systems aim at integrating different unimodal representations into a synthetic one. So far, a consequent effort has been made on developing complex architectures allowing the fusion of these modalities. However, such systems are mainly trained by minimising simple losses such as $L_1$ or cross-entropy. In this work, we investigate unexplored penalties and propose a set of new objectives that measure the dependency between modalities. We demonstrate that our new penalties lead to a consistent improvement (up to $4.3$ on accuracy) across a large variety of state-of-the-art models on two well-known sentiment analysis datasets: \texttt{CMU-MOSI} and \texttt{CMU-MOSEI}. Our method not only achieves a new SOTA on both datasets but also produces representations that are more robust to modality drops. Finally, a by-product of our methods includes a statistical network which can be used to interpret the high dimensional representations learnt by the model.

While there has been a growing interest in pretraining for dialog A crucial step in conversational AI is the identification (Mehri et al., 2019; Zhang et al., 2019d), the focus of underlying information of the user's utterance has mainly been on English datasets. Thus, these (e.g communicative intent or dialog acts, and works can not be directly applied to our multilingual emotions). This requires modeling utterance-level setting. Additionally, available multilingual information (Mitkov, 2014; Williams et al., 2014), pretraining objectives (Lample and Conneau, 2019; to capture immediate nuances of the user utterance; Liu et al., 2020; Xue et al., 2020; Qi et al., 2021) and discourse-level features (Thornbury and Slade, face two main limitations when applied to dialog 2006), to capture patterns over long ranges of the modeling: (1) they are a generalization of monolingual conversation. An added difficulty to this modeling objectives that use flat input text, whereas problem is that most people in the world are bilingual hierarchy has been shown to be a powerful prior (Grosjean and Li, 2013): therefore, progress for dialog modeling. This is a reflection of a dialog on these systems is limited by their inability to process itself, for example, context plays an essential role more than one language (English being the in the labeling of dialog acts.

A new metric \texttt{BaryScore} to evaluate text generation based on deep contextualized embeddings e.g., BERT, Roberta, ELMo) is introduced. This metric is motivated by a new framework relying on optimal transport tools, i.e., Wasserstein distance and barycenter. By modelling the layer output of deep contextualized embeddings as a probability distribution rather than by a vector embedding; this framework provides a natural way to aggregate the different outputs through the Wasserstein space topology. In addition, it provides theoretical grounds to our metric and offers an alternative to available solutions e.g., MoverScore and BertScore). Numerical evaluation is performed on four different tasks: machine translation, summarization, data2text generation and image captioning. Our results show that \texttt{BaryScore} outperforms other BERT based metrics and exhibits more consistent behaviour in particular for text summarization.

Probabilistic generative models are attractive for scientific modeling because their inferred parameters can be used to generate hypotheses and design experiments. This requires that the learned model provide an accurate representation of the input data and yield a latent space that effectively predicts outcomes relevant to the scientific question. Supervised Variational Autoencoders (SVAEs) have previously been used for this purpose, where a carefully designed decoder can be used as an interpretable generative model while the supervised objective ensures a predictive latent representation. Unfortunately, the supervised objective forces the encoder to learn a biased approximation to the generative posterior distribution, which renders the generative parameters unreliable when used in scientific models. This issue has remained undetected as reconstruction losses commonly used to evaluate model performance do not detect bias in the encoder. We address this previously-unreported issue by developing a second order supervision framework (SOS-VAE) that influences the decoder to induce a predictive latent representation. This ensures that the associated encoder maintains a reliable generative interpretation. We extend this technique to allow the user to trade-off some bias in the generative parameters for improved predictive performance, acting as an intermediate option between SVAEs and our new SOS-VAE. We also use this methodology to address missing data issues that often arise when combining recordings from multiple scientific experiments. We demonstrate the effectiveness of these developments using synthetic data and electrophysiological recordings with an emphasis on how our learned representations can be used to design scientific experiments.

Meta-learning empowers artificial intelligence to increase its efficiency by learning how to learn. Unlocking this potential involves overcoming a challenging meta-optimisation problem that often exhibits ill-conditioning, and myopic meta-objectives. We propose an algorithm that tackles these issues by letting the meta-learner teach itself. The algorithm first bootstraps a target from the meta-learner, then optimises the meta-learner by minimising the distance to that target under a chosen (pseudo-)metric. Focusing on meta-learning with gradients, we establish conditions that guarantee performance improvements and show that the improvement is related to the target distance. Thus, by controlling curvature, the distance measure can be used to ease meta-optimization, for instance by reducing ill-conditioning. Further, the bootstrapping mechanism can extend the effective meta-learning horizon without requiring backpropagation through all updates. The algorithm is versatile and easy to implement. We achieve a new state-of-the art for model-free agents on the Atari ALE benchmark, improve upon MAML in few-shot learning, and demonstrate how our approach opens up new possibilities by meta-learning efficient exploration in a Q-learning agent.

Reliably predicting potential failure risks of machine learning (ML) systems when deployed with production data is a crucial aspect of trustworthy AI. This paper introduces Risk Advisor, a novel post-hoc meta-learner for estimating failure risks and predictive uncertainties of any already-trained black-box classification model. In addition to providing a risk score, the Risk Advisor decomposes the uncertainty estimates into aleatoric and epistemic uncertainty components, thus giving informative insights into the sources of uncertainty inducing the failures. Consequently, Risk Advisor can distinguish between failures caused by data variability, data shifts and model limitations and advise on mitigation actions (e.g., collecting more data to counter data shift). Extensive experiments on various families of black-box classification models and on real-world and synthetic datasets covering common ML failure scenarios show that the Risk Advisor reliably predicts deployment-time failure risks in all the scenarios, and outperforms strong baselines.

Constructing adversarial perturbations for deep neural networks is an important direction of research. Crafting image-dependent adversarial perturbations using white-box feedback has hitherto been the norm for such adversarial attacks. However, black-box attacks are much more practical for real-world applications. Universal perturbations applicable across multiple images are gaining popularity due to their innate generalizability. There have also been efforts to restrict the perturbations to a few pixels in the image. This helps to retain visual similarity with the original images making such attacks hard to detect. This paper marks an important step which combines all these directions of research. We propose the DEceit algorithm for constructing effective universal pixel-restricted perturbations using only black-box feedback from the target network. We conduct empirical investigations using the ImageNet validation set on the state-of-the-art deep neural classifiers by varying the number of pixels to be perturbed from a meagre 10 pixels to as high as all pixels in the image. We find that perturbing only about 10% of the pixels in an image using DEceit achieves a commendable and highly transferable Fooling Rate while retaining the visual quality. We further demonstrate that DEceit can be successfully applied to image dependent attacks as well. In both sets of experiments, we outperformed several state-of-the-art methods.

Researchers from Poland and Japan, working with Sony, have found evidence that machine learning systems trained on GPUs rather than CPUs may contain fewer errors during the training process, and produce superior results, contradicting the common understanding that GPUs simply perform such operations faster, rather than any better. The research, titled Impact of GPU Uncertainty on the Training of Predictive Deep Neural Networks, comes from the Faculty of Psychology and Cognitive Sciences at Adam Mickiewicz University and two Japanese universities, together with SONY Computer Science Laboratories. The study suggests that'uncertainties' which deep neural networks exhibit in the face of various hardware and software configurations favor more expensive (and increasingly scarce) graphics processing units, and found in tests that a deep neural network trained exclusively on CPU produced higher error rates over the same number of epochs (the number of times that the system reprocesses the training data over the course of a session). In this supplemental example from the paper, we see (bottom two rows), similar result quality obtained from a variety of GPUs, and (first row), the inferior results obtained from a range of otherwise very capable CPUs. These preliminary findings do not apply uniformly across popular machine learning algorithms, and in the case of simple autoencoder architectures, the phenomenon does not appear.

The Transform Technology Summits start October 13th with Low-Code/No Code: Enabling Enterprise Agility. AI-powered coding tools, which generate code using machine learning algorithms, have attracted increasing attention over the last decade. In theory, systems like OpenAI's Codex could reduce the time people spend writing software as well as computational and operational costs. But existing systems have major limitations, leading to undesirable results like errors. In search of a better approach, researchers at Salesforce open-sourced a machine learning system called CodeT5, which can understand and generate code in real time. The team claims that CodeT5 achieves state-of-the-art performance on coding tasks including code defect detection, which predicts whether code is vulnerable to exploits, and clone detection, which predicts whether two code snippets have the same functionality.

In their paper, Tabular Data: Deep Learning is Not All You Need, the authors argue that while deep learning methods have shown tremendous success in the image and text domains, traditional tree-based methods like XGBoost still continue to shine when it comes to tabular data. The authors examined Tabnet, Neural Oblivious Decision Ensembles (NODE), DNF-Net, and 1D-CNN deep learning models and compared their performance on eleven datasets with XGBoost. This is an important paper in the sense that it reiterates that deep learning may not be the silver bullet for solving all machine learning problems. On the other hand, tree-based algorithms have been shown to perform at par or even outperform neural networks for tabular data while being simple to use and comprehend. And there is good news for people who like to work with tree-based models.