Joy Buolamwini was conducting research at MIT on how computers recognized people's faces, when she started experiencing something weird. Whenever she sat before a system's front-facing camera, it wouldn't recognize her face, even after working for her lighter-skinned friends. But when she put on a simple white mask, the face-tracking animation suddenly lit up the screen. Suspecting a more widespread problem, she carried out a study on the AI-powered facial recognition systems of Microsoft, IBM and Face, a Chinese startup that has raised more than $500 million from investors. Buolamwini showed the systems 1,000 faces, and told them to identify each as male or female.

PSD2 stands for second payment standards directive, and was issued by the EU with the intent to open access to data and services previously only available to banks. In theory, PSD2 went into effect on January 13, and should provide a clear framework for implementation. The difference between theory and practice is small in theory, but big in practice. PSD2 is defined in a set of documents published by the EU, which are to be subsequently implemented as legislation by parliaments in EU countries and enforced by regulating bodies. PSD2 will effectively force financial institutions active in the EU to open up data and functionality previously only available to them to other parties.

NatWest is testing an artificial intelligence-powered "digital human" called Cora that will converse with customers in branches – raising fears that bank tellers could be replaced by avatars. Cora, described by the bank as "highly lifelike", is the result of a link-up with a New Zealand tech company whose co-founder was involved with creating digital characters in the blockbuster films Avatar, King Kong and Spider-Man 2. Cora is currently able to answer basic verbal questions such as "How do I login to online banking?", "How do I apply for a mortgage?" NatWest said it could help cut down on waiting times because it would be able to deal with simple problems, adding that Cora's AI skills would eventually expand to answering hundreds of different questions, even detecting human emotions and reacting verbally and physically with facial expressions. The bank implied that a digital human could improve on the real thing by "providing consistent, accurate answers all the time in a way that humans can't always do". While some will be excited by the news that the "AI revolution" is coming to the UK's high streets, others will fear that the 71% taxpayer-owned bank is investing in this technology in order to replace branches or staff.

A patient's own molecular data can be used to identify, with the use of artificial intelligence methods, the best combinatorial multi-drug therapy for that patient. Network modeling plays a major role in this line of work. The ongoing research on artificial intelligence methods for precision cancer medicine at the Computational Biomodeling (Combio) Laboratory of Åbo Akademi and Turku Centre for Computer Science (TUCS) got a major boost with renewed funding from Business Finland. The concept of this project is that a patient's own molecular data can be used to identify, with the use of artificial intelligence methods, the best combinatorial multi-drug therapy for that patient. Network modeling plays a major role in this line of work, integrating genome-scale patient data into detailed interaction networks, that can be analyzed by Combio's recently developed algorithms to identify combinations of drugs and inhibitors that are likely to be therapeutically effective.

The concept of artificial intelligence (AI) has been floating around for years. At first it was just an idea discussed at universities by stereotypical academics. But is has emerged, first test cases have been developed and then, in the form of intelligent personal assistants (Siri, Cortana, Alexa), developed into something that we are confronted with in everyday life. There are new developments almost every month – in January 2018 AI delivered,for the first time, a better reading performance than a human. It is only a matter of time, until AI will move into manufacturing.

We propose a Multi-Instance-Learning (MIL) approach for weakly-supervised learning problems, where a training set is formed by bags (sets of feature vectors or instances) and only labels at bag-level are provided. Specifically, we consider the Multi-Instance Dynamic-Ordinal-Regression (MI-DOR) setting, where the instance labels are naturally represented as ordinal variables and bags are structured as temporal sequences. To this end, we propose Multi-Instance Dynamic Ordinal Random Fields (MI-DORF). In this framework, we treat instance-labels as temporally-dependent latent variables in an Undirected Graphical Model. Different MIL assumptions are modelled via newly introduced high-order potentials relating bag and instance-labels within the energy function of the model. We also extend our framework to address the Partially-Observed MI-DOR problems, where a subset of instance labels are available during training. We show on the tasks of weakly-supervised facial behavior analysis, Facial Action Unit (DISFA dataset) and Pain (UNBC dataset) Intensity estimation, that the proposed framework outperforms alternative learning approaches. Furthermore, we show that MIDORF can be employed to reduce the data annotation efforts in this context by large-scale.

Vector quantization aims to form new vectors/matrices with shared values close to the original. It could compress data with acceptable information loss, and could be of great usefulness in areas like Image Processing, Pattern Recognition and Machine Learning. In recent years, the importance of quantization has been soaring as it has been discovered huge potentials in deploying practical neural networks, which is among one of the most popular research topics. Conventional vector quantization methods usually suffer from their own flaws: hand-coding domain rules quantization could produce poor results when encountering complex data, and clustering-based algorithms have the problem of inexact solution and high time consumption. In this paper, we explored vector quantization problem from a new perspective of sparse least square optimization and designed multiple algorithms with their program implementations. Specifically, deriving from a sparse form of coefficient matrix, three types of sparse least squares, with $l_0$, $l_1$, and generalized $l_1 + l_2$ penalizations, are designed and implemented respectively. In addition, to produce quantization results with given amount of quantized values(instead of penalization coefficient $\lambda$), this paper proposed a cluster-based least square quantization method, which could also be regarded as an improvement of information preservation of conventional clustering algorithm. The algorithms were tested on various data and tasks and their computational properties were analyzed. The paper offers a new perspective to probe the area of vector quantization, while the algorithms proposed could provide more appropriate options for quantization tasks under different circumstances.

Semidefinite programs (SDP) are important in learning and combinatorial optimization with numerous applications. In pursuit of low-rank solutions and low complexity algorithms, we consider the Burer--Monteiro factorization approach for solving SDPs. We show that all approximate local optima are global optima for the penalty formulation of appropriately rank-constrained SDPs as long as the number of constraints scales sub-quadratically with the desired rank of the optimal solution. Our result is based on a simple penalty function formulation of the rank-constrained SDP along with a smoothed analysis to avoid worst-case cost matrices. We particularize our results to two applications, namely, Max-Cut and matrix completion.

In the recent literature the important role of depth in deep learning has been emphasized. In this paper we argue that sufficient width of a feedforward network is equally important by answering the simple question under which conditions the decision regions of a neural network are connected. It turns out that for a class of activation functions including leaky ReLU, neural networks having a pyramidal structure, that is no layer has more hidden units than the input dimension, produce necessarily connected decision regions. This implies that a sufficiently wide layer is necessary to produce disconnected decision regions. We discuss the implications of this result for the construction of neural networks, in particular the relation to the problem of adversarial manipulation of classifiers.

In this paper, we describe a novel shape classification method which is embedded in the Bayesian paradigm. We discuss the modelling and the resulting shape classification algorithm for two and three dimensional data shapes. We conclude by evaluating the efficiency and efficacy of the proposed algorithm on the Kimia shape database for the two dimensional case.