Extreme Learning Machine (ELM) is a noniterative training method suited for Single Layer Feed Forward Neural Networks (SLFF-NN). Typically, a hardware neural network is trained before implementation in order to avoid ...
详细信息
Extreme Learning Machine (ELM) is a noniterative training method suited for Single Layer Feed Forward Neural Networks (SLFF-NN). Typically, a hardware neural network is trained before implementation in order to avoid additional on-chip occupation, delay and performance degradation. However, ELM provides fixed-time learning capability and simplifies the process of re-training a neural network once implemented in hardware. This is an important issue in many applications where input data are continuously changing and a new training process must be launched very often, providing self-adaptation. This work describes a general SLFF-NN design environment to assist in the definition of neural network hardware implementation parameters including real-time ELM training. The software design environment uses initial user-provided input data with information about the type of problem: sample dataset and validated results, input fields, accuracy; and, together with simulation tools, recommends the optimum configuration for the neural topology and automatically generates synthesizable code for the hardware implementation tool. This is possible due to the design of parameter-dependent synthesis code and optimal hardware architecture design for both neural network and ELM training. Results show all the steps required to follow a successful design flow from the software tool to the final running device and, as an application example, the FPGA implementation for realtime detection of brain area in electrode positioning during a Deep Brain Stimulation (DBS) surgery is shown.
Quality evaluation is a fundamental problem in the field of linguistic description of data. In this work, we analyze the concept of quality and study different approaches to measure quality. Although most of the appro...
详细信息
ISBN:
(纸本)9781467374293
Quality evaluation is a fundamental problem in the field of linguistic description of data. In this work, we analyze the concept of quality and study different approaches to measure quality. Although most of the approaches considered focused on time series data, that are one of the most frequent datasets in real application domains, they can be used for quality assessment of linguistic descriptions generated for any type of data.
In this paper, we introduce a new Gaussian Process (GP) classification method for multisensory data. The proposed approach can deal with noisy and missing data. It is also capable of estimating the contribution of eac...
详细信息
In this paper, we introduce a new Gaussian Process (GP) classification method for multisensory data. The proposed approach can deal with noisy and missing data. It is also capable of estimating the contribution of each sensor towards the classification task. We use Bayesian modeling to build a GP-based classifier which combines the information provided by all sensors and approximates the posterior distribution of the GP using variational Bayesian inference. During its training phase, the algorithm estimates each sensor's weight and then uses this information to assign a label to each new sample. In the experimental section, we evaluate the classiication performance of the proposed method on both synthetic and real data and show its applicability to different scenarios.
To overcome the inability of Description Logics (DLs) to represent vague or imprecise information, several fuzzy extensions have been proposed in the literature. In this context, an important family of reasoning algor...
详细信息
To overcome the inability of Description Logics (DLs) to represent vague or imprecise information, several fuzzy extensions have been proposed in the literature. In this context, an important family of reasoning algorithms for fuzzy DLs is based on a combination of tableau algorithms and Operational Research (OR) problems, specifically using Mixed Integer Linear Programming (MILP). In this paper, we present a MILP-based tableau procedure that allows to reason within fuzzy ALCB, i.e., ALC with individual value restrictions. Interestingly, unlike classical tableau procedures, our tableau algorithm is deterministic, in the sense that it defers the inherent non-determinism in ALCB to a MILP solver.
This paper addresses the RF-design of absorptive bandstop filters (ABSFs) that feature narrow fractional bandwidth (FBW) and quality factors (Qs) of the order of a thousand. The aforementioned functionality is realize...
详细信息
This paper addresses the RF-design of absorptive bandstop filters (ABSFs) that feature narrow fractional bandwidth (FBW) and quality factors (Qs) of the order of a thousand. The aforementioned functionality is realized by means of a mixed implementation scheme in which one-port acoustic wave (AW) resonators and lumped-element impedance inverters are effectively combined in a compact geometry. The non-ideal effects-fabrication tolerances-of the utilized lumped-element components on the ABSF performance-theoretically infinite isolation and input reflection-are analyzed. A method to maintain the large stopband attenuation by adjusting the effective Q of each resonator with the aid of a tunable resistor is also reported. A filter prototype that makes use of commercially available surface acoustic wave (SAW) resonators and surface-mounted devices (SMDs) has been built and tested at 418 MHz for verification purposes. It exhibits a 0.02% 3-dB FBW and a maximum stopband isolation of 62 dB.
Reputation systems provide reputation values of rated parties to users. These reputation values, typically aggregations of individual user ratings, shall be reliable, i.e. should enable a realistic assessment of the p...
详细信息
ISBN:
(纸本)9781450324694
Reputation systems provide reputation values of rated parties to users. These reputation values, typically aggregations of individual user ratings, shall be reliable, i.e. should enable a realistic assessment of the probability that the rated party behaves as expected in a transaction. In order for the reputation values to stay reliable and, thus, for the reputation system to provide a benefit, the system needs to be resistant against manipulations by users, the rated parties trying to improve their reputation values, and even against competitors trying to worsen a reputation value. At the same time, a reputation system shall provide privacy protection for users: rated parties shall not be able to learn who provided a certain rating. Otherwise users might not take part in the system as they fear bad feedback in revenge for bad ratings, or users do not want to be connected to certain transactions based on their provided ratings. In this paper we come up with a solution that provides both, reliability of reputation values on the one hand, and privacy protection for users on the other hand. In contrast to related work, our solution only makes use of a single reputation provider that needs to be trusted (to a certain extent) and does not require any bulletin boards to be present in the system. We make use of the Paillier cryptosystem to provide an aggregation of individual user ratings in a way that no party can learn which user provided a certain rating. Copyright 2014 ACM.
This paper presents a new feature descriptor suitable to the task of matching features points between images with nonlinear intensity variations. This includes image pairs with significant illuminations changes, multi...
详细信息
This paper presents a new feature descriptor suitable to the task of matching features points between images with nonlinear intensity variations. This includes image pairs with significant illuminations changes, multi-modal image pairs and multi-spectral image pairs. The proposed method describes the neighbourhood of feature points combining frequency and spatial information using multi-scale and multi-oriented Log-Gabor filters. Experimental results show the validity of the proposed approach and also the improvements with respect to the state of the art.
This paper focuses on the design of highly selective and high quality factor (Q) absorptive bandstop filters (ABSFs) that are based on a mixed implementation scheme using acoustic wave (AW) resonators and lumped-eleme...
详细信息
This paper focuses on the design of highly selective and high quality factor (Q) absorptive bandstop filters (ABSFs) that are based on a mixed implementation scheme using acoustic wave (AW) resonators and lumped-element components. The proposed approach enables mobile form-factor notch filters with theoretically infinite attenuation even for stopbands showing 3-dB fractional bandwidths (FBWs) as narrow as 0.013-0.02% (Q=10,000-15,000). Proof-of-concept filter prototypes using commercially-available surface acoustic wave (SAW) resonators and surface-mount components are designed, built, and characterized for an example frequency of 418 MHz. Various stopbands with measured FBWs between 0.017 and 0.078% (i.e., 0.07 and 0.327 MHz in absolute terms) and attenuation levels as high as 83 dB are demonstrated. Absorptive notches with reconfigurable levels of isolation in the range of 10-83 dB are also shown through a fabricated ABSF with controllable attenuation.
This study is focused on developing models for identifying epilepsy convulsions in order to enhance the anamnesis of the patient. A 3D accelerometer built-in wearable device is placed on the dominant wrist to gather d...
详细信息
ISBN:
(纸本)9781467385121
This study is focused on developing models for identifying epilepsy convulsions in order to enhance the anamnesis of the patient. A 3D accelerometer built-in wearable device is placed on the dominant wrist to gather data from participants. Based on the data gathered from the sensor, a Fuzzy Rule Based System is learned. On the one hand, statistical data from a set of patients is used to set up the partition data base; on the other hand, the Fuzzy rule base is learned using Ant Colony Optimization. Results show this approach faster and easier to learn than previous research. Introducing minor changes in the fuzzy reasoning produces even more robust models. The presented study shows a valid research path for the identification of the epilepsy convulsions.
The Business Process Modelling and Notation (BPMN) is a widely-accepted standard for process modelling, which can be used to model the clinical processes contained in guidelines. computer systems based on guidelines n...
详细信息
The Business Process Modelling and Notation (BPMN) is a widely-accepted standard for process modelling, which can be used to model the clinical processes contained in guidelines. computer systems based on guidelines need to embed these clinical processes, e.g. using a computer-Interpretable Guideline (CIG) language. However, encoding guidelines in a CIG language is a difficult task which is usually performed by technical staff. Building on our previous work, the transformation-based refinement of guideline models, in this paper we describe an algorithm to transform BPMN models into the SDA CIG language. The use of BPMN has the potential to empower clinicians in the modelling task. In combination with the transformation algorithm, this can lead to an increased adoption of CIG languages, SDA and others.
暂无评论