Interestingly, they all seem to have picked a particular school of thought in deep learning. With time, this pattern is becoming more and more clear. For instance, Facebook AI Research (FAIR) has been championing self-supervised learning (SSL) for quite some time, alongside releasing relevant papers and tech related to computer vision, image, text, video, and audio understanding. Even though many companies and research institutions seem to have their hands on every possible area within deep learning, a clear pattern is emerging. But, of course, all of them have their favourites. In this article, we will explore some of the recent work in their respective niche/popularised areas.

Robots are becoming everyday devices, increasing their interaction with humans. To make human-machine interaction more natural, cognitive features like Visual Voice Activity Detection (VVAD), which can detect whether a person is speaking or not, given visual input of a camera, need to be implemented. Neural networks are state of the art for tasks in Image Processing, Time Series Prediction, Natural Language Processing and other domains. Those Networks require large quantities of labeled data. Currently there are not many datasets for the task of VVAD. In this work we created a large scale dataset called the VVAD-LRS3 dataset, derived by automatic annotations from the LRS3 dataset. The VVAD-LRS3 dataset contains over 44K samples, over three times the next competitive dataset (WildVVAD). We evaluate different baselines on four kinds of features: facial and lip images, and facial and lip landmark features. With a Convolutional Neural Network Long Short Term Memory (CNN LSTM) on facial images an accuracy of 92% was reached on the test set. A study with humans showed that they reach an accuracy of 87.93% on the test set.

Explainability has become a major topic of research in Artificial Intelligence (AI), aimed at increasing trust in models such as Deep Learning (DL) networks. However, trustworthy models cannot be achieved with explainable AI (XAI) methods unless the XAI methods themselves can be trusted. This necessity gave rise to the assessment of XAI methods. To evaluate XAI methods one may assess interpretability, a qualitative measure of how understandable an explanation is to humans Gilpin et al. [2018]. While this is important to guarantee the proper interaction between humans and the model, interpretability generally involves end-users in the process Mohseni et al. [2018], inducing strong biases. In fact, a qualitative evaluation alone cannot guarantee coherency to reality (i.e., model behavior), as false explanations can be more interpretable than accurate ones. To enable trust on XAI methods, we also need quantitative and objective evaluation metrics which validate the relation between the explanations produced by the XAI method and the trained model being assessed. The challenge of quantitatively evaluating XAI methods lies in the absence of a ground truth: we cannot be sure of what a DL method is doing unless we understand the model parametrization itself (at which point we would not need a XAI method). Nonetheless, we still want to validate the faithfulness Selvaraju et al. [2019] of XAI methods w.r.t. the underlying model, as this allows us to discern between accurate and misleading explanations.

Large pre-trained language models have shown promise for few-shot learning, completing text-based tasks given only a few task-specific examples. Will models soon solve classification tasks that have so far been reserved for human research assistants? Existing benchmarks are not designed to measure progress in applied settings, and so don't directly answer this question. The RAFT benchmark (Real-world Annotated Few-shot Tasks) focuses on naturally occurring tasks and uses an evaluation setup that mirrors deployment. Baseline evaluations on RAFT reveal areas current techniques struggle with: reasoning over long texts and tasks with many classes. Human baselines show that some classification tasks are difficult for non-expert humans, reflecting that real-world value sometimes depends on domain expertise. Yet even non-expert human baseline F1 scores exceed GPT-3 by an average of 0.11. The RAFT datasets and leaderboard will track which model improvements translate into real-world benefits at https://raft.elicit.org .

There has been increasing interest in characterising the error behaviour of systems which contain deep learning models before deploying them into any safety-critical scenario. However, characterising such behaviour usually requires large-scale testing of the model that can be extremely computationally expensive for complex real-world tasks. For example, tasks involving compute intensive object detectors as one of their components. In this work, we propose an approach that enables efficient large-scale testing using simplified low-fidelity simulators and without the computational cost of executing expensive deep learning models. Our approach relies on designing an efficient surrogate model corresponding to the compute intensive components of the task under test. We demonstrate the efficacy of our methodology by evaluating the performance of an autonomous driving task in the Carla simulator with reduced computational expense by training efficient surrogate models for PIXOR and CenterPoint LiDAR detectors, whilst demonstrating that the accuracy of the simulation is maintained.

Background: The construction, evolution and usage of complex artificial intelligence (AI) models demand expensive computational resources. While currently available high-performance computing environments support well this complexity, the deployment of AI models in mobile devices, which is an increasing trend, is challenging. Mobile applications consist of environments with low computational resources and hence imply limitations in the design decisions during the AI-enabled software engineering lifecycle that balance the trade-off between the accuracy and the complexity of the mobile applications. Objective: Our objective is to systematically assess the trade-off between accuracy and complexity when deploying complex AI models (e.g. neural networks) to mobile devices, which have an implicit resource limitation. We aim to cover (i) the impact of the design decisions on the achievement of high-accuracy and low resource-consumption implementations; and (ii) the validation of profiling tools for systematically promoting greener AI. Method: This confirmatory registered report consists of a plan to conduct an empirical study to quantify the implications of the design decisions on AI-enabled applications performance and to report experiences of the end-to-end AI-enabled software engineering lifecycle. Concretely, we will implement both image-based and language-based neural networks in mobile applications to solve multiple image classification and text classification problems on different benchmark datasets. Overall, we plan to model the accuracy and complexity of AI-enabled applications in operation with respect to their design decisions and will provide tools for allowing practitioners to gain consciousness of the quantitative relationship between the design decisions and the green characteristics of study.

Accurate, long-term forecasting of human pedestrian trajectories in highly dynamic and interactive scenes is a long-standing challenge. Recent advances in using data-driven approaches have achieved significant improvements in terms of prediction accuracy. However, the lack of group-aware analysis has limited the performance of forecasting models. This is especially apparent in highly populated scenes, where pedestrians are moving in groups and the interactions between groups are extremely complex and dynamic. In this paper, we present Grouptron, a multi-scale dynamic forecasting framework that leverages pedestrian group detection and utilizes individual-level, group-level, and scene-level information for better understanding and representation of the scenes. Our approach employs spatio-temporal clustering algorithms to identify pedestrian groups, creates spatio-temporal graphs at the individual, group, and scene levels. It then uses graph neural networks to encode dynamics at different scales and incorporates encoding across different scales for trajectory prediction. We carried out extensive comparisons and ablation experiments to demonstrate the effectiveness of our approach. Our method achieves 9.3% decrease in final displacement error (FDE) compared with state-of-the-art methods on ETH/UCY benchmark datasets, and 16.1% decrease in FDE in more crowded scenes where extensive human group interactions are more frequently present.

The problem of phase retrieval, or the algorithmic recovery of lost phase information from measured intensity alone, underlies various imaging methods from astronomy to nanoscale imaging. Traditional methods of phase retrieval are iterative in nature, and are therefore computationally expensive and time consuming. More recently, deep learning (DL) models have been developed to either provide learned priors to iterative phase retrieval or in some cases completely replace phase retrieval with networks that learn to recover the lost phase information from measured intensity alone. However, such models require vast amounts of labeled data, which can only be obtained through simulation or performing computationally prohibitive phase retrieval on hundreds of or even thousands of experimental datasets. Using a 3D nanoscale X-ray imaging modality (Bragg Coherent Diffraction Imaging or BCDI) as a representative technique, we demonstrate AutoPhaseNN, a DL-based approach which learns to solve the phase problem without labeled data. By incorporating the physics of the imaging technique into the DL model during training, AutoPhaseNN learns to invert 3D BCDI data from reciprocal space to real space in a single shot without ever being shown real space images. Once trained, AutoPhaseNN is about one hundred times faster than traditional iterative phase retrieval methods while providing comparable image quality.

Machine learning (ML) systems are rapidly increasing in size, are acquiring new capabilities, and are increasingly deployed in high-stakes settings. As with other powerful technologies, safety for ML should be a leading research priority. In response to emerging safety challenges in ML, such as those introduced by recent large-scale models, we provide a new roadmap for ML Safety and refine the technical problems that the field needs to address. We present four problems ready for research, namely withstanding hazards ("Robustness"), identifying hazards ("Monitoring"), steering ML systems ("Alignment"), and reducing risks to how ML systems are handled ("External Safety"). Throughout, we clarify each problem's motivation and provide concrete research directions.

In the early 1990s, the first signs of life of the CNN era were given: LeCun et al. proposed a CNN model trained by the backpropagation algorithm to classify low-resolution images of handwritten digits. Undoubtedly, it was a breakthrough in the field of computer vision. But with the rise of other classification methods, it fell out fashion. That was until 2012, when Krizhevsky et al. revived the interest in CNNs by exhibiting considerably higher image classification accuracy on the ImageNet challenge. Since then, the complexity of the architectures are exponentially increasing and many structures are rapidly becoming obsolete. Using multistream networks as a base and the feature infusion precept, we explore the proposed LCNN cross-fusion strategy to use the backbones of former state-of-the-art networks on image classification in order to discover if the technique is able to put these designs back in the game. In this paper, we showed that we can obtain an increase of accuracy up to 63.21% on the NORB dataset we comparing with the original structure. However, no technique is definitive. While our goal is to try to reuse previous state-of-the-art architectures with few modifications, we also expose the disadvantages of our explored strategy.