The Transform Technology Summits start October 13th with Low-Code/No Code: Enabling Enterprise Agility. Women in the AI field are making research breakthroughs, spearheading vital ethical discussions, and inspiring the next generation of AI professionals. We created the VentureBeat Women in AI Awards to emphasize the importance of their voices, work, and experience and to shine a light on some of these leaders. In this series, publishing Fridays, we're diving deeper into conversations with this year's winners, whom we honored recently at Transform 2021. Check out last week's interview with a winner of our AI responsibility and ethics award.

It's been three years since Google launched the Nest Hello, a wired video doorbell with facial recognition that helped set the standard for smart doorbells. In that time, many competitors have appeared, but few have come close to the quality and reliability of that device. The Nest Doorbell (battery) is a Google-made video doorbell that can run on battery power (it can also operate on wired power, if you have that infrastructure and wish to connect it to your existing doorbell chime). In our tests the device performed excellently, didn't give any problems, and proved itself to be a worthy sister device to the original Nest Hello. A large, round black circle with a camera in its center is at the top of the new Nest doorbell. A small LED below that indicates when the camera is recording or processing video.

Even the most skilled inspector might have an off day. Here it can be useful to outsource mundane or mechanical tasks to'intelligent' machines. Poorly defined quality-control procedures were blamed for one of the most extensive automotive parts recalls in history, involving the airbag manufacturer Takata. The firm's inflators, which contained the chemical ammonium nitrate, were found to be unsafe – leading 19 US carmakers to recall 69m of the products. Similar recalls were issued in Japan, China and Oceania.

Eighteen months ago, Fabio Dores was making good money as a drag queen. Performing under the name Felicity Suxwell, he had a club residency and worked hen nights throughout the UK, attracting enough bookings to quit his day job at a lettings agency. Then lockdown came and everything shut down. Bored at home, he was browsing Facebook and spotted an advertisement for LegionFarm, an online video-game coaching platform that offered to match pro gamers with clients looking to improve their abilities. As a skilled player of battle royale hit Apex Legends, he applied to become a coach.

The Internet of things (IoT) is taking the world by storm, thanks to the proliferation of sensors and actuators embedded in everyday things, coupled with the wide availability of high-speed Internet50 and evolution of the 5th-generation (5G) networks.34 IoT devices are increasingly supplying information about the physical environment (for example, infrastructure, assets, homes, and cars). The advent of IoT is enabling not only the connection and integration of devices that monitor physical world phenomena (for example, temperature, pollution, energy consumption, human activities, and movement), but also data-driven and AI-augmented intelligence. At all levels, synergies from advances in IoT, data analytics, and artificial intelligence (AI) are firmly recognized as strategic priorities for digital transformation.10,41,50 IoT poses two key challenges:36 Communication with things and management of things.41 The service paradigm is a key mechanism to overcome these challenges by transforming IoT devices into IoT services, where they will be treated as first-class objects through the prism of services.9 In a nutshell, services are at a higher level of abstraction than data. Services descriptions consist of two parts: functional and non-functional, such as, Quality of Service (QoS) attributes.27 Services often transform data into an actionable knowledge or achieve physical state changes in the operating context.9 As a result, the service paradigm is the perfect basis for understanding the transformation of data into actionable knowledge, that is, making it useful. Despite the increasing uptake of IoT services, most organizations have not yet mastered the requisite knowledge, skills, or understanding to craft a successful IoT strategy.

Nowadays, deep learning models are widely adopted in web-scale applications such as recommender systems, and online advertising. In these applications, embedding learning of categorical features is crucial to the success of deep learning models. In these models, a standard method is that each categorical feature value is assigned a unique embedding vector which can be learned and optimized. Although this method can well capture the characteristics of the categorical features and promise good performance, it can incur a huge memory cost to store the embedding table, especially for those web-scale applications. Such a huge memory cost significantly holds back the effectiveness and usability of EDRMs. In this paper, we propose a binary code based hash embedding method which allows the size of the embedding table to be reduced in arbitrary scale without compromising too much performance. Experimental evaluation results show that one can still achieve 99\% performance even if the embedding table size is reduced 1000$\times$ smaller than the original one with our proposed method.

Embedding learning for categorical features is crucial for the deep learning-based recommendation models (DLRMs). Each feature value is mapped to an embedding vector via an embedding learning process. Conventional methods configure a fixed and uniform embedding size to all feature values from the same feature field. However, such a configuration is not only sub-optimal for embedding learning but also memory costly. Existing methods that attempt to resolve these problems, either rule-based or neural architecture search (NAS)-based, need extensive efforts on the human design or network training. They are also not flexible in embedding size selection or in warm-start-based applications. In this paper, we propose a novel and effective embedding size selection scheme. Specifically, we design an Adaptively-Masked Twins-based Layer (AMTL) behind the standard embedding layer. AMTL generates a mask vector to mask the undesired dimensions for each embedding vector. The mask vector brings flexibility in selecting the dimensions and the proposed layer can be easily added to either untrained or trained DLRMs. Extensive experimental evaluations show that the proposed scheme outperforms competitive baselines on all the benchmark tasks, and is also memory-efficient, saving 60\% memory usage without compromising any performance metrics.

Sparse deep neural networks have proven to be efficient for predictive model building in large-scale studies. Although several works have studied theoretical and numerical properties of sparse neural architectures, they have primarily focused on the edge selection. Sparsity through edge selection might be intuitively appealing; however, it does not necessarily reduce the structural complexity of a network. Instead pruning excessive nodes in each layer leads to a structurally sparse network which would have lower computational complexity and memory footprint. We propose a Bayesian sparse solution using spike-and-slab Gaussian priors to allow for node selection during training. The use of spike-and-slab prior alleviates the need of an ad-hoc thresholding rule for pruning redundant nodes from a network. In addition, we adopt a variational Bayes approach to circumvent the computational challenges of traditional Markov Chain Monte Carlo (MCMC) implementation. In the context of node selection, we establish the fundamental result of variational posterior consistency together with the characterization of prior parameters. In contrast to the previous works, our theoretical development relaxes the assumptions of the equal number of nodes and uniform bounds on all network weights, thereby accommodating sparse networks with layer-dependent node structures or coefficient bounds. With a layer-wise characterization of prior inclusion probabilities, we also discuss optimal contraction rates of the variational posterior. Finally, we provide empirical evidence to substantiate that our theoretical work facilitates layer-wise optimal node recovery together with competitive predictive performance.

Multiclass multilabel classification refers to the task of attributing multiple labels to examples via predictions. Current models formulate a reduction of that multilabel setting into either multiple binary classifications or multiclass classification, allowing for the use of existing loss functions (sigmoid, cross-entropy, logistic, etc.). Empirically, these methods have been reported to achieve good performance on different metrics (F1 score, Recall, Precision, etc.). Theoretically though, the multilabel classification reductions does not accommodate for the prediction of varying numbers of labels per example and the underlying losses are distant estimates of the performance metrics. We propose a loss function, sigmoidF1. It is an approximation of the F1 score that (I) is smooth and tractable for stochastic gradient descent, (II) naturally approximates a multilabel metric, (III) estimates label propensities and label counts. More generally, we show that any confusion matrix metric can be formulated with a smooth surrogate. We evaluate the proposed loss function on different text and image datasets, and with a variety of metrics, to account for the complexity of multilabel classification evaluation. In our experiments, we embed the sigmoidF1 loss in a classification head that is attached to state-of-the-art efficient pretrained neural networks MobileNetV2 and DistilBERT. Our experiments show that sigmoidF1 outperforms other loss functions on four datasets and several metrics. These results show the effectiveness of using inference-time metrics as loss function at training time in general and their potential on non-trivial classification problems like multilabel classification.

Breast histology image classification is a crucial step in the early diagnosis of breast cancer. In breast pathological diagnosis, Convolutional Neural Networks (CNNs) have demonstrated great success using digitized histology slides. However, tissue classification is still challenging due to the high visual variability of the large-sized digitized samples and the lack of contextual information. In this paper, we propose a novel CNN, called Multi-level Context and Uncertainty aware (MCUa) dynamic deep learning ensemble model.MCUamodel consists of several multi-level context-aware models to learn the spatial dependency between image patches in a layer-wise fashion. It exploits the high sensitivity to the multi-level contextual information using an uncertainty quantification component to accomplish a novel dynamic ensemble model.MCUamodelhas achieved a high accuracy of 98.11% on a breast cancer histology image dataset. Experimental results show the superior effectiveness of the proposed solution compared to the state-of-the-art histology classification models.