Automaton models are often seen as interpretable models. Interpretability itself is not well defined: it remains unclear what interpretability means without first explicitly specifying objectives or desired attributes. In this paper, we identify the key properties used to interpret automata and propose a modification of a state-merging approach to learn variants of finite state automata. We apply the approach to problems beyond typical grammar inference tasks. Additionally, we cover several use-cases for prediction, classification, and clustering on sequential data in both supervised and unsupervised scenarios to show how the identified key properties are applicable in a wide range of contexts.

Abstract--Dictionary learning is the task of determining a data-dependent transform that yields a sparse representation of some observed data. The dictionary learning problem is non-convex, and usually solved via computationally complex iterative algorithms. Furthermore, the resulting transforms obtained generally lack structure that permits their fast application to data. T o address this issue, this paper develops a framework for learning orthonormal dictionaries which are built from products of a few Householder reflectors. Two algorithms are proposed to learn the reflector coefficients: one that considers a sequential update of the reflectors and one with a simultaneous update of all reflectors that imposes an additional internal orthogonal constraint. The proposed methods have low computational complexity and are shown to converge to local minimum points which can be described in terms of the spectral properties of the matrices involved. Simulations of the proposed algorithms are shown in the image processing setting where well-known fast transforms are available for comparisons. The proposed algorithms have favorable reconstruction error and the advantage of a fast implementation relative to the classical, unstructured, dictionaries. Index Terms--sparsifying transforms, fast transforms, dictionary learning, compressed sensing. Sparsifying transforms [1] allow efficient representation of data when a data-dependent overcomplete dictionary is available. Overcomplete dictionaries are useful in image processing [2], [3], [4], speech processing [5] and wireless communications [6], [7]. Unfortunately, the selection of a sparsifying transform involves solving a non-convex optimization problem for a dictionary matrixD such that a real data set can be represented with a sparse representation matrixX whose sparsity level is constrained. Because direct solution of the optimization method is difficult [8], [9], proposed algorithms seek a suboptimal solution via alternating minimization. Most prior work considers alternating minimization for dictionaries that are overcomplete.

For data sets populated by a very well modeled process and by another process of unknown probability density function (PDF), a desired feature when manipulating the fraction of the unknown process (either for enhancing it or suppressing it) consists in avoiding to modify the kinematic distributions of the well modeled one. A bootstrap technique is used to identify sub-samples rich in the well modeled process, and classify each event according to the frequency of it being part of such sub-samples. Comparisons with general MVA algorithms will be shown, as well as a study of the asymptotic properties of the method, making use of a public domain data set that models a typical search for new physics as performed at hadronic colliders such as the Large Hadron Collider (LHC). The most popular classification algorithms based on supervised learning require a well modeled signal and a well modeled background. For the classifier to learn how to separate the two classes, it is crucial that both models are known. The case in which either signal or background has an unknown PDF is, however, acquiring importance in many classification problems that arise in the realm of particle physics, due to the fact that every passing day more and more known models are ruled out by the data. Two scenarios are mainly interesting for particle physics: a very well known background modeled from simulation, in presence of an unknown rare signal; a very well known signal modeled from simulation, contaminated by a background of origin unclear and/or not simulable. In both scenarios, it is desirable to manipulate the fraction of the unknown process, without modifying the kinematic distributions of the very well known one.

Markov logic networks (MLNs) reconcile two opposing schools in machine learning and artificial intelligence: causal networks, which account for uncertainty extremely well, and first-order logic, which allows for formal deduction. An MLN is essentially a first-order logic template to generate Markov networks. Inference in MLNs is probabilistic and it is often performed by approximate methods such as Markov chain Monte Carlo (MCMC) Gibbs sampling. An MLN has many regular, symmetric structures that can be exploited at both first-order level and in the generated Markov network. We analyze the graph structures that are produced by various lifting methods and investigate the extent to which quantum protocols can be used to speed up Gibbs sampling with state preparation and measurement schemes. We review different such approaches, discuss their advantages, theoretical limitations, and their appeal to implementations. We find that a straightforward application of a recent result yields exponential speedup compared to classical heuristics in approximate probabilistic inference, thereby demonstrating another example where advanced quantum resources can potentially prove useful in machine learning.

We address a question answering task on real-world images that is set up as a Visual Turing Test. By combining latest advances in image representation and natural language processing, we propose Ask Your Neurons, a scalable, jointly trained, end-to-end formulation to this problem. In contrast to previous efforts, we are facing a multi-modal problem where the language output (answer) is conditioned on visual and natural language inputs (image and question). We provide additional insights into the problem by analyzing how much information is contained only in the language part for which we provide a new human baseline. To study human consensus, which is related to the ambiguities inherent in this challenging task, we propose two novel metrics and collect additional answers which extend the original DAQUAR dataset to DAQUAR-Consensus. Moreover, we also extend our analysis to VQA, a large-scale question answering about images dataset, where we investigate some particular design choices and show the importance of stronger visual models. At the same time, we achieve strong performance of our model that still uses a global image representation. Finally, based on such analysis, we refine our Ask Your Neurons on DAQUAR, which also leads to a better performance on this challenging task.

Information on these groups can help optimizing the structure and contents of the site. In this paper we use an approach based on formal concepts for constructing taxonomies of user groups. For decreasing the huge amount of concepts that arise in applications, we employ stability index of a concept, which describes how a group given by a concept extent differs from other such groups. We analyze resulting taxonomies of user groups for three target websites.

Google has previously taught its artificial intelligence to play games, and it's even capable of creating its own encryption. Now, its language translation tool has used machine learning to create a'language' all of its own. In September, the search giant turned on its Google Neural Machine Translation (GNMT) system to help it automatically improve how it translates languages. The machine learning system analyses and makes sense of languages by looking at entire sentences – rather than individual phrases or words. Following several months of testing, the researchers behind the AI have seen it be able to blindly translate languages even if it's never studied one of the languages involved in the translation.

During its wheeled treks on the Red Planet, NASA's Spirit rover may have encountered a potential signature of past life on Mars, report scientists at Arizona State University (ASU). To help make their case, the researchers have contrasted Spirit's study of "Home Plate" -- a plateau of layered rocks that the robot explored during the early part of its third year on Mars -- with features found within active hot spring/geyser discharge channels at a site in northern Chile called El Tatio. The work has resulted in a provocative paper: "Silica deposits on Mars with features resembling hot spring biosignatures at El Tatio in Chile." As reported online last week in the journal Nature Communications, field work in Chile by the ASU team -- Steven Ruff and Jack Farmer of the university's School of Earth and Space Exploration -- shows that the nodular and digitate silica structures at El Tatio that most closely resemble those on Mars include complex sedimentary structures produced by a combination of biotic and abiotic processes. "Although fully abiotic processes are not ruled out for the Martian silica structures, they satisfy an a priori definition of potential biosignatures," the researchers wrote in the study.

Within the Machine Learning for Artists workshop program in Opendot from 21st to 25th of November, we are proud to invite you to the Gene Kogan OpenTalk, on Wednesday 23rd at 7 pm in Opendot lab. A Beautiful and Interesting Game: a lecture by Gene Kogan on creative applications for Machine Learning algorithms This talk examines the rise of machine learning and artificial intelligence through the lens of artistic practice and creative subversion. Recent breakthroughs in scientific research, combined with the proliferation of big data and cheap GPU computing power, have dramatically increased the capacities of machine intelligence in a variety of domains. The tech titans have swiftly integrated them into most of their core services, whilst numerous startups have appeared to capitalize on emerging markets. At the same time, artists, boosted by independent open source implementations, have attempted to subvert and illuminate those same technologies, shedding light on the sometimes beautiful and sometimes dangerous new faculties of these powerful algorithms.

Dogs have a remarkable ability to recall events from the past, in a similar way to humans. That's according to a new study which found evidence canines have a similar'episodic memory' to their human counterparts. Dogs can recall a person's actions even when they do not expect to have their memory tested, says the research. Previously, evidence that animals use episodic memory has been hard to come by, as it's impossible to ask an animal, in this case a dog, what they remember (stock image) Dogs trained using the trick can watch a person perform an action and carry out the action themselves. For example, if the their owner jumps in the air and then gives the command'do it', the dog would jump in the air. However, the researchers also needed to show that the dogs remembered what they just saw a person do, even when they weren't expecting to be asked or rewarded.