Mario Martín

On-line support vector machine regression

This paper describes an on-line method for building e-insensitive support vector machines for regression as described in [12]. The method is an extension of the method developed by [1] for building incremental support vector machines for classification. Machines obtained by using this approach are equivalent to the ones obtained by applying exact methods like quadratic programming, but they are obtained more quickly and allow the incremental addition of new points, removal of existing points and update of target values for existing data. This development opens the application of SVM regression to areas such as on-line prediction of temporal series or generalization of value functions in reinforcement learning.

Learning Generalized Policies from Planning Examples Using Concept Languages

In this paper we are concerned with the problem of learning how to solve planning problems in one domain given a number of solved instances. This problem is formulated as the problem of inferring a function that operates over all instances in the domain and maps states and goals into actions. We call such functions generalized policies and the question that we address is how to learn suitable representations of generalized policies from data. This question has been addressed recently by Roni Khardon (Technical Report TR-09-97, Harvard, 1997). Khardon represents generalized policies using an ordered list of existentially quantified rules that are inferred from a training set using a version of Rivests learning algorithm (Machine Learning, vol. 2, no. 3, pp. 229–246, 1987). Here, we follow Khardons approach but represent generalized policies in a different way using a concept language. We show through a number of experiments in the blocks-world that the concept language yields a better policy using a smaller set of examples and no background knowledge.

A portable method for acquiring information extraction patterns without annotated corpora

The main issue when building Information Extraction (IE) systems is how to obtain the knowledge needed to identify relevant information in a document. Most approaches require expert human intervention in many steps of the acquisition process. In this paper we describe ESSENCE, a new method for acquiring IE patterns that significantly reduces the need for human intervention. The method is based on ELA, a specifically designed learning algorithm for acquiring IE patterns without tagged examples. The distinctive features of ESSENCE and ELA are that (1) they permit the automatic acquisition of IE patterns from unrestricted and untagged text representative of the domain, due to (2) their ability to identify regularities around semantically relevant concept-words for the IE task by (3) using non-domain-specific lexical knowledge tools such as WordNet, and (4) restricting the human intervention to defining the task, and validating and typifying the set of IE patterns obtained. Since ESSENCE does not require a corpus annotated with the type of information to be extracted and it uses a general purpose ontology and widely applied syntactic tools, it reduces the expert effort required to build an IE system and therefore also reduces the effort of porting the method to any domain. The results of the application of ESSENCE to the acquisition of IE patterns in an MUC-like task are shown.

Deep learning for freezing of gait detection in Parkinson’s disease patients in their homes using a waist-worn inertial measurement unit

Among Parkinsons disease (PD) motor symptoms, freezing of gait (FOG) may be the most incapacitating. FOG episodes may result in falls and reduce patients quality of life. Accurate assessment of FOG would provide objective information to neurologists about the patients condition and the symptoms characteristics, while it could enable non-pharmacologic support based on rhythmic cues.This paper is, to the best of our knowledge, the first study to propose a deep learning method for detecting FOG episodes in PD patients. This model is trained using a novel spectral data representation strategy which considers information from both the previous and current signal windows. Our approach was evaluated using data collected by a waist-placed inertial measurement unit from 21 PD patients who manifested FOG episodes. These data were also employed to reproduce the state-of-the-art methodologies, which served to perform a comparative study to our FOG monitoring system.The results of this study demonstrate that our approach successfully outperforms the state-of-the-art methods for automatic FOG detection. Precisely, the deep learning model achieved 90% for the geometric mean between sensitivity and specificity, whereas the state-of-the-art methods were unable to surpass the 83% for the same metric.

Conceptual Connectivity Analysis by Means of Fuzzy Partitions

In this paper we present a set of tools to analyze concepts that describe and explain a set of observations. Due to the inherent vagueness of concepts, that makes hard to decide in a dichotomic base weather an observation is, or is not, a good example for a concept, we consider the concepts associated to fuzzy subsets. Then we study the adequation and coverage of a collection of fuzzy sets to describe a set of observations. In the same way, once a set of concepts has been acepted to describe a set of objects, we study how the concepts are related by means of the observations, and reciprocally, how the objects are related by the concepts. Finally, a short description of the computer program COCOA is given.

A machine learning methodology for the selection and classification of spontaneous spinal cord dorsum potentials allows disclosure of structured (non-random) changes in neuronal connectivity induced by nociceptive stimulation.

Previous studies aimed to disclose the functional organization of the neuronal networks involved in the generation of the spontaneous cord dorsum potentials (CDPs) generated in the lumbosacral spinal segments used predetermined templates to select specific classes of spontaneous CDPs. Since this procedure was time consuming and required continuous supervision, it was limited to the analysis of two specific types of CDPs (negative CDPs and negative positive CDPs), thus excluding potentials that may reflect activation of other neuronal networks of presumed functional relevance. We now present a novel procedure based in machine learning that allows the efficient and unbiased selection of a variety of spontaneous CDPs with different shapes and amplitudes. The reliability and performance of the present method is evaluated by analyzing the effects on the probabilities of generation of different classes of spontaneous CDPs induced by the intradermic injection of small amounts of capsaicin in the anesthetized cat, a procedure known to induce a state of central sensitization leading to allodynia and hyperalgesia. The results obtained with the selection method presently described allowed detection of spontaneous CDPs with specific shapes and amplitudes that are assumed to represent the activation of functionally coupled sets of dorsal horn neurones that acquire different, structured configurations in response to nociceptive stimuli. These changes are considered as responses tending to adequate transmission of sensory information to specific functional requirements as part of homeostatic adjustments.

A Randomized Algorithm for the Exact Solution of Transductive Support Vector Machines

Random sampling is an efficient method for dealing with constrained optimization problems. In computational geometry, this method has been successfully applied, through Clarkson’s algorithm (Clarkson 1996), to solve a general class of problems called violator spaces. In machine learning, Transductive Support Vector Machines (TSVM) is a learning method used when only a small fraction of labeled data is available, which implies solving a nonconvex optimization problem. Several approximation methods have been proposed to solve it, but they usually find suboptimal solutions. However, a global optimal solution may be obtained by using exact techniques, but at the cost of suffering an exponential time complexity with respect to the number of instances. In this article, an interpretation of TSVM in terms of violator space is given. A randomized method is presented that extends the use of exact methods, thus reducing the time complexity exponentially w.r.t. the number of support vectors of the optimal solution instead of exponentially w.r.t. the number of instances.

Deep learning for detecting freezing of gait episodes in Parkinson’s disease based on accelerometers

Freezing of gait (FOG) is one of the most incapacitating symptoms among the motor alterations of Parkinson’s disease (PD). Manifesting FOG episodes reduce patients’ quality of life and their autonomy to perform daily living activities, while it may provoke falls. Accurate ambulatory FOG assessment would enable non-pharmacologic support based on cues and would provide relevant information to neurologists on the disease evolution.

Markovian Analysis of the Sequential Behavior of the Spontaneous Spinal Cord Dorsum Potentials Induced by Acute Nociceptive Stimulation in the Anesthetized Cat

In a previous study we developed a Machine Learning procedure for the automatic identification and classification of spontaneous cord dorsum potentials (CDPs). This study further supported the proposal that in the anesthetized cat, the spontaneous CDPs recorded from different lumbar spinal segments are generated by a distributed network of dorsal horn neurons with structured (non-random) patterns of functional connectivity and that these configurations can be changed to other non-random and stable configurations after the noceptive stimulation produced by the intradermic injection of capsaicin in the anesthetized cat. Here we present a study showing that the sequence of identified forms of the spontaneous CDPs follows a Markov chain of at least order one. That is, the system has memory in the sense that the spontaneous activation of dorsal horn neuronal ensembles producing the CDPs is not independent of the most recent activity. We used this markovian property to build a procedure to identify portions of signals as belonging to a specific functional state of connectivity among the neuronal networks involved in the generation of the CDPs. We have tested this procedure during acute nociceptive stimulation produced by the intradermic injection of capsaicin in intact as well as spinalized preparations. Altogether, our results indicate that CDP sequences cannot be generated by a renewal stochastic process. Moreover, it is possible to describe some functional features of activity in the cord dorsum by modeling the CDP sequences as generated by a Markov order one stochastic process. Finally, these Markov models make possible to determine the functional state which produced a CDP sequence. The proposed identification procedures appear to be useful for the analysis of the sequential behavior of the ongoing CDPs recorded from different spinal segments in response to a variety of experimental procedures including the changes produced by acute nociceptive stimulation. They are envisaged as a useful tool to examine alterations of the patterns of functional connectivity between dorsal horn neurons under normal and different pathological conditions, an issue of potential clinical concern.

Approximate policy iteration using regularised Bellman residuals minimisation

In this paper we present an approximate policy iteration (API) method called API‐BRMϵ using a very effective implementation of incremental support vector regression (SVR) to approximate the value function able to generalise in continuous (or large) space reinforcement learning (RL) problems. RL is a methodology able to solve complex and uncertain decision problems usually modelled as Markov decision problems. API-BRMϵ is formalised as a non-parametric regularisation problem based on an outcome of the Bellman residual minimisation (BRM) which is able to minimise the variance of the problem. API-BRMϵ is incremental and can be applied to RL using the on-line agent interaction framework. Based on non-parametric SVR API-BRMϵ is able to find the global solution of the problem with convergence guarantees to the optimal solution. A value function should be defined to find the optimal policy specifying the total reward that an agent might expect in its current state taking one action. Therefore, the agent will use the value function to choose the action to take. Some experimental evidence and performance for well-known RL benchmarks are presented.