The study was published in Oncotarget on 22nd of December, 2016. The study represents the proof of concept for applying Generative Adversarial Networks (GANs) to drug discovery. The authors significantly extended this model to generate new leads according to multiple requested characteristics and plan to launch a comprehensive GAN-based drug discovery engine producing promising therapeutic treatments to significantly accelerate pharmaceutical R&D and improve the success rates in clinical trials. Since 2010 deep learning systems demonstrated unprecedented results in image, voice and text recognition, in many cases surpassing human accuracy and enabling autonomous driving, automated creation of pleasant art and even composition of pleasant music. GAN is a fresh direction in deep learning invented by Ian Goodfellow in 2014.

We adopted an approach based on an LSTM neural network to monitor and detect faults in industrial multivariate time series data. To validate the approach we created a Modelica model of part of a real gasoil plant. By introducing hacks into the logic of the Modelica model, we were able to generate both the roots and causes of fault behavior in the plant. Having a self-consistent data set with labeled faults, we used an LSTM architecture with a forecasting error threshold to obtain precision and recall quality metrics. The dependency of the quality metric on the threshold level is considered. An appropriate mechanism such as "one handle" was introduced for filtering faults that are outside of the plant operator field of interest.

It has long been recognized that the invariance and equivariance properties of a representation are critically important for success in many vision tasks. In this paper we present Steerable Convolutional Neural Networks, an efficient and flexible class of equivariant convolutional networks. We show that steerable CNNs achieve state of the art results on the CIFAR image classification benchmark. The mathematical theory of steerable representations reveals a type system in which any steerable representation is a composition of elementary feature types, each one associated with a particular kind of symmetry. We show how the parameter cost of a steerable filter bank depends on the types of the input and output features, and show how to use this knowledge to construct CNNs that utilize parameters effectively.

Multivariate Pattern (MVP) classification holds enormous potential for decoding visual stimuli in the human brain by employing task-based fMRI data sets. There is a wide range of challenges in the MVP techniques, i.e. decreasing noise and sparsity, defining effective regions of interest (ROIs), visualizing results, and the cost of brain studies. In overcoming these challenges, this paper proposes a novel model of neural representation, which can automatically detect the active regions for each visual stimulus and then utilize these anatomical regions for visualizing and analyzing the functional activities. Therefore, this model provides an opportunity for neuroscientists to ask this question: what is the effect of a stimulus on each of the detected regions instead of just study the fluctuation of voxels in the manually selected ROIs. Moreover, our method introduces analyzing snapshots of brain image for decreasing sparsity rather than using the whole of fMRI time series. Further, a new Gaussian smoothing method is proposed for removing noise of voxels in the level of ROIs. The proposed method enables us to combine different fMRI data sets for reducing the cost of brain studies. Experimental studies on 4 visual categories (words, consonants, objects and nonsense photos) confirm that the proposed method achieves superior performance to state-of-the-art methods.

Many real-world systems can be studied in terms of pattern recognition tasks, so that proper use (and understanding) of machine learning methods in practical applications becomes essential. While a myriad of classification methods have been proposed, there is no consensus on which methods are more suitable for a given dataset. As a consequence, it is important to comprehensively compare methods in many possible scenarios. In this context, we performed a systematic comparison of 7 well-known clustering methods available in the R language. In order to account for the many possible variations of data, we considered artificial datasets with several tunable properties (number of classes, separation between classes, etc). In addition, we also evaluated the sensitivity of the clustering methods with regard to their parameters configuration. The results revealed that, when considering the default configurations of the adopted methods, the spectral approach usually outperformed the other clustering algorithms. We also found that the default configuration of the adopted implementations was not accurate. In these cases, a simple approach based on random selection of parameters values proved to be a good alternative to improve the performance. All in all, the reported approach provides subsidies guiding the choice of clustering algorithms.

In stochastic convex optimization the goal is to minimize a convex function $F(x) \doteq {\mathbf E}_{{\mathbf f}\sim D}[{\mathbf f}(x)]$ over a convex set $\cal K \subset {\mathbb R}^d$ where $D$ is some unknown distribution and each $f(\cdot)$ in the support of $D$ is convex over $\cal K$. The optimization is commonly based on i.i.d.~samples $f^1,f^2,\ldots,f^n$ from $D$. A standard approach to such problems is empirical risk minimization (ERM) that optimizes $F_S(x) \doteq \frac{1}{n}\sum_{i\leq n} f^i(x)$. Here we consider the question of how many samples are necessary for ERM to succeed and the closely related question of uniform convergence of $F_S$ to $F$ over $\cal K$. We demonstrate that in the standard $\ell_p/\ell_q$ setting of Lipschitz-bounded functions over a $\cal K$ of bounded radius, ERM requires sample size that scales linearly with the dimension $d$. This nearly matches standard upper bounds and improves on $\Omega(\log d)$ dependence proved for $\ell_2/\ell_2$ setting by Shalev-Shwartz et al. (2009). In stark contrast, these problems can be solved using dimension-independent number of samples for $\ell_2/\ell_2$ setting and $\log d$ dependence for $\ell_1/\ell_\infty$ setting using other approaches. We further show that our lower bound applies even if the functions in the support of $D$ are smooth and efficiently computable and even if an $\ell_1$ regularization term is added. Finally, we demonstrate that for a more general class of bounded-range (but not Lipschitz-bounded) stochastic convex programs an infinite gap appears already in dimension 2.

A shorter version was published in HBR Online on 21st December 2016. Also, the post does not review a new White House paper on AI and its impact released on 20th December 2016, that cites some posts on this blog.] On October 12, 2016, President Obama's Executive Office published two reports that received less media attention than they might have otherwise because the United States was gripped by the final weeks of a presidential campaign race. In these two reports, the administration laid out its plans for the future of artificial intelligence (AI). Depending on one's view of AI's potential impact, the actions resulting from these reports may be more influential on the long arc of history than the outcome of that election.

It's easy to joke that the internet was invented to give people around the world the opportunity to share pictures of cats. However, according to a new report, those kitty pictures may one day turn out to save your life. That is based on work being done by Dr. Alvin Rajkomar, an assistant professor at the University of California, San Francisco Medical Center. Rajkomar trained a deep learning neural network to be able to automatically detect life-threatening abnormalities in chest X-rays. "When I was a medical resident, I ordered a stat X-ray of a patient who I suspected had a life-threatening pneumothorax -- air outside of his lung compressing his heart -- and happened to be standing next to the digital X-ray machine as it was being taken," he told Digital Trends.

The 2012 publication Race against the Machine makes the case that the digitalization of work activities is proceeding so rapidly as to cause dislocations in the job market beyond anything previously experienced [1]. Unlike past mechanization/automation, which affected lower-skill blue-collar and white-collar work, today's information technology affects workers high in the education and skill distribution. Machines can substitute for brains as well as brawn. On one estimate, about 47% of total US employment is at risk of computerization [2]. If you doubt whether a robot or some other machine equipped with digital intelligence connected to the internet could outdo you or me in our work in the foreseeable future, consider news reports about an IBM program to "create" new food dishes (chefs beware), the battle between anesthesiologists and computer programs/robots that do their job much cheaper, and the coming version of Watson ("twice as powerful as the original") based on computers connected over the internet via IBM's Cloud [3]. On the darker side, you do not have to be paranoid to be paranoid about the potential technologies that the super-secret computers of the US National Security Agency (NSA) have on their digital drawing-boards.

About halfway through a particularly tense game of Go held in Seoul, South Korea, between Lee Sedol, one of the best players of all time, and AlphaGo, an artificial intelligence created by Google, the AI program made a mysterious move that demonstrated an unnerving edge over its human opponent. On move 37, AlphaGo chose to put a black stone in what seemed, at first, like a ridiculous position. It looked certain to give up substantial territory--a rookie mistake in a game that is all about controlling the space on the board. Two television commentators wondered if they had misread the move or if the machine had malfunctioned somehow. In fact, contrary to any conventional wisdom, move 37 would enable AlphaGo to build a formidable foundation in the center of the board. The Google program had effectively won the game using a move that no human would've come up with. One reason that understanding language is so difficult for computers and AI systems is that words often have meanings based on context and even the appearance of the letters and words. In the images that accompany this story, several artists demonstrate the use of a variety of visual clues to convey meanings far beyond the actual letters.