By definition, diversity implies active and inclusive participation. When it comes to Artificial Intelligence Systems (AIS), it's even more critical to establish strategies and partnerships that can foster genuine and holistic inclusion regarding design for these important technologies. Since 2016, City AI has become the world's leading non-profit connecting the AI community globally and supporting local AI ecosystems to develop further. AI provides one of the greatest opportunities for humanity. Yet the access to its expertise and resources is limited.

Tensor completion refers to the task of estimating the missing data from an incomplete measurement or observation, which is a core problem frequently arising from the areas of big data analysis, computer vision, and network engineering. Due to the multidimensional nature of high-order tensors, the matrix approaches, e.g., matrix factorization and direct matricization of tensors, are often not ideal for tensor completion and recovery. Exploiting the potential periodicity and inherent correlation properties appeared in real-world tensor data, in this paper, we shall incorporate the low-rank and sparse regularization technique to enhance Tucker decomposition for tensor completion. A series of computational experiments on real-world datasets, including internet traffic data, color images, and face recognition, show that our model performs better than many existing state-of-the-art matricization and tensorization approaches in terms of achieving higher recovery accuracy. Naturally, these data would be stored as where rank(F) represents the rank of the underlying matrix higher-order tensor (a.k.a., multidimensional array), which F and M R Following the spirit of matrix completion model [1], [2], [3], [4], to name just a few.

Disentanglement is a highly desirable property of representation due to its similarity with human's understanding and reasoning. This improves interpretability, enables the performance of down-stream tasks, and enables controllable generative models. However, this domain is challenged by the abstract notion and incomplete theories to support unsupervised disentanglement learning. We demonstrate the data itself, such as the orientation of images, plays a crucial role in disentanglement and instead of the factors, and the disentangled representations align the latent variables with the action sequences. We further introduce the concept of disentangling action sequences which facilitates the description of the behaviours of the existing disentangling approaches. An analogy for this process is to discover the commonality between the things and categorizing them. Furthermore, we analyze the inductive biases on the data and find that the latent information thresholds are correlated with the significance of the actions. For the supervised and unsupervised settings, we respectively introduce two methods to measure the thresholds. We further propose a novel framework, fractional variational autoencoder (FVAE), to disentangle the action sequences with different significance step-by-step. Experimental results on dSprites and 3D Chairs show that FVAE improves the stability of disentanglement.

Multimodal neural machine translation (NMT) has become an increasingly important area of research over the years because additional modalities, such as image data, can provide more context to textual data. Furthermore, the viability of training multimodal NMT models without a large parallel corpus continues to be investigated due to low availability of parallel sentences with images, particularly for English-Japanese data. However, this void can be filled with comparable sentences that contain bilingual terms and parallel phrases, which are naturally created through media such as social network posts and e-commerce product descriptions. In this paper, we propose a new multimodal English-Japanese corpus with comparable sentences that are compiled from existing image captioning datasets. In addition, we supplement our comparable sentences with a smaller parallel corpus for validation and test purposes. To test the performance of this comparable sentence translation scenario, we train several baseline NMT models with our comparable corpus and evaluate their English-Japanese translation performance. Due to low translation scores in our baseline experiments, we believe that current multimodal NMT models are not designed to effectively utilize comparable sentence data. Despite this, we hope for our corpus to be used to further research into multimodal NMT with comparable sentences.

The prior work on natural language inference (NLI) debiasing mainly targets at one or few known biases while not necessarily making the models more robust. In this paper, we focus on the model-agnostic debiasing strategies and explore how to (or is it possible to) make the NLI models robust to multiple distinct adversarial attacks while keeping or even strengthening the models' generalization power. We firstly benchmark prevailing neural NLI models including pretrained ones on various adversarial datasets. We then try to combat distinct known biases by modifying a mixture of experts (MoE) ensemble method and show that it's nontrivial to mitigate multiple NLI biases at the same time, and that model-level ensemble method outperforms MoE ensemble method. We also perform data augmentation including text swap, word substitution and paraphrase and prove its efficiency in combating various (though not all) adversarial attacks at the same time. Finally, we investigate several methods to merge heterogeneous training data (1.35M) and perform model ensembling, which are straightforward but effective to strengthen NLI models.

The Levenshtein distance is a text similarity metric that measures the distance between 2 words. It has a number of applications, including text autocompletion and autocorrection. For either of these use cases, the word entered by a user is compared to words in a dictionary to find the closest match, at which point a suggestion(s) is made. The dictionary may contain thousands of words, and thus the response of the application for comparing 2 words will likely take a few milliseconds. The Levenshtein distance is usually calculated by preparing a matrix of size (M 1)x(N 1)--where M and N are the lengths of the 2 words--and looping through said matrix using 2 for loops, performing some calculations within each iteration.

In gradient boosting, predictions are made from an ensemble of weak learners. Unlike a random forest that creates a decision tree for each sample, in gradient boosting, trees are created one after the other. Previous trees in the model are not altered. Results from the previous tree are used to improve the next one. In this piece, we'll take a closer look at a gradient boosting library called CatBoost.

Scientists have mapped 1.8 billion individual tree canopies across millions of kilometres of the Sahel and Sahara regions of West Africa. It is the first time ever that trees have been mapped in detail over such a large area. So how was it possible? They employed neural networks which are able to recognise objects, like trees, based on their shapes and colours. To train it, the AI system was shown satellite images where trees had been manually traced.

About a quarter of people haven't ruled out the idea of dating a robot, according to a new survey, and the Dutch are the most accepting of the idea of artificial amour. Researchers from the University of Twente used data from the EU-backed SIENNA project that studies ethics and opinions surrounding cutting edge technology. They surveyed 11,000 people and found 27 per cent either supported the idea of dating a robot or hadn't completely ruled it out, and 72 per cent were completely opposed to the idea of a digital dalliance. In the Netherlands support for someone having a robotic boyfriend or girlfriend went up to 53 per cent, the highest of the 11 countries involved in the survey. The multinational telephone survey by the Dutch research team also found that people were uncomfortable with robots that look and behave like humans. About a quarter of people haven't ruled out the idea of dating a robot, according to a new survey, and the Dutch are the most accepting of the idea of artificial amour We are getting used to interacting with intelligent machines, from robot vacuum cleaners, smart speakers that can control our lights and AI assistants in our phones.

Low-rank tensor decomposition is one of the most effective approaches to reduce the memory and computing requirements of large-size neural networks, enabling their efficient deployment on various hardware platforms. While post-training tensor compression can greatly reduce the cost of inference, uncompressed training still consumes excessive hardware resources, run-time and energy. It is highly desirable to directly train a compact low-rank tensorized model from scratch with a low memory and computational cost. However, this is a very challenging task because it is hard to determine a proper tensor rank a priori, which controls the model complexity and compression ratio in the training process. This paper presents a novel end-to-end framework for low-rank tensorized training of neural networks. We first develop a flexible Bayesian model that can handle various low-rank tensor formats (e.g., CP, Tucker, tensor train and tensor-train matrix) that compress neural network parameters in training. This model can automatically determine the tensor ranks inside a nonlinear forward model, which is beyond the capability of existing Bayesian tensor methods. We further develop a scalable stochastic variational inference solver to estimate the posterior density of large-scale problems in training. Our work provides the first general-purpose rank-adaptive framework for end-to-end tensorized training. Our numerical results on various neural network architectures show orders-of-magnitude parameter reduction and little accuracy loss (or even better accuracy) in the training process.