Cornelius Glackin

A comparison of fuzzy strategies for corporate acquisition analysis

Analysing all prospective companies for acquisition in large market sectors is an onerous task. A strategy that results in a shortlist of companies that meet certain basic criteria is required. The short-listed companies can then be further investigated in more detail later if desired. Fuzzy logic systems (FLSs) imbued with the expertise of a focal organisations financial experts can be of great assistance in this process. In this paper an investigation into the suitability of FLSs for acquisition analysis is presented. The nuances of training and tuning are discussed. In particular, the difficulty of obtaining suitable amounts of expert data is a recurring theme throughout the paper. A strategy for circumventing this issue is presented that relies on the design of a conventional fuzzy logic rule base with the assistance of a financial expert. With the rule base created, various scenarios such as the simulation of multiple experts and the creation of expert training data are investigated. In particular, two scenarios for the creation of simulated expert data are presented. In the first the responses from the different experts are averaged, and in the second scenario the responses from all the different experts are preserved in the training data. This paper builds on previous work with scalable membership functions, however, the use of fuzzy C-means clustering and backpropagation training, are new developments. Additionally, a type-2 FLS is developed and its potential advantages are discussed for this application. The type-2 system facilitates the inclusion of the opinions of multiple experts. Both the type-1 and type-2 FLSs were trained using the backpropagation algorithm with early stopping and verified with five-fold cross-validation. Multiple runs of the five-fold method were conducted with different random orderings of the data. For this particular application, the type-1 system performed comparably with the type-2 system despite the considerable amount of variation in the expert training data. The training results have proven the methods to be capable of efficient tuning of parameters, and of reliable ranking of prospective companies.

Feature extraction from spectro-temporal signals using dynamic synapses, recurrency, and lateral inhibition

This paper presents a spiking neural network-based investigation of the issues associated with extraction of onset, offset, and coincidental firing features from spectro-temporal data. Speech samples containing spoken isolated digits from the TI46 database are employed to demonstrate the way in which these features can be extracted using leaky integrate-and-fire spiking neurons with dynamic synapses. The flexibility that the additional synaptic parameters in the neuron model provides, is demonstrated to be essential for onset, offset and coincidental firing extraction. Recurrency and the interaction between excitation and inhibition together with latency is demonstrated to be a viable means of extracting offset features. The effects of lateral inhibition and in particular its ability to induce transient synchrony in spike firing is evaluated. In particular, by defining a connection length parameter, and hence a neighbourhood size, synchronous firing is shown to gradually develop as connection length and neighbourhood size increases. Finally, the implications for this connectivity in spiking neural networks and its potential for learning spectral and spatio-temporal patterns via the formation of receptive fields is discussed.

Receptive field optimisation and supervision of a fuzzy spiking neural network

This paper presents a supervised training algorithm that implements fuzzy reasoning on a spiking neural network. Neuron selectivity is facilitated using receptive fields that enable individual neurons to be responsive to certain spike train firing rates and behave in a similar manner as fuzzy membership functions. The connectivity of the hidden and output layers in the fuzzy spiking neural network (FSNN) is representative of a fuzzy rule base. Fuzzy C-Means clustering is utilised to produce clusters that represent the antecedent part of the fuzzy rule base that aid classification of the feature data. Suitable cluster widths are determined using two strategies; subjective thresholding and evolutionary thresholding respectively. The former technique typically results in compact solutions in terms of the number of neurons, and is shown to be particularly suited to small data sets. In the latter technique a pool of cluster candidates is generated using Fuzzy C-Means clustering and then a genetic algorithm is employed to select the most suitable clusters and to specify cluster widths. In both scenarios, the network is supervised but learning only occurs locally as in the biological case. The advantages and disadvantages of the network topology for the Fisher Iris and Wisconsin Breast Cancer benchmark classification tasks are demonstrated and directions of current and future work are discussed.

Implementing Fuzzy Reasoning on a Spiking Neural Network

This paper presents a supervised training algorithm that implements fuzzy reasoning on a spiking neural network. Neuron selectivity is facilitated using receptive fields that enable individual neurons to be responsive to certain spike train frequencies. The receptive fields behave in a similar manner as fuzzy membership functions. The network is supervised but learning only occurs locally as in the biological case. The connectivity of the hidden and output layers is representative of a fuzzy rule base. The advantages and disadvantages of the network topology for the IRIS classification task are demonstrated and directions of current and future work are discussed.

The Roberta IRONSIDE project: A dialog capable humanoid personal assistant in a wheelchair for dependent persons

With an aging population and the financial difficulties of having a full time caregiver for every dependent person living at home, assistant robots appear to be a solution for advanced countries. However, most of what can be done with a robot can be done without it. So it is difficult to quantify what real value an assistant robot can add. Such a robot should be a real assistant capable of helping a person, whether indoors or outdoors. Additionally, the robot should be a companion for dialoging, as well as a system capable of detecting health problems. The Roberta Ironside project is a robotic evolution, embodying the expertise learned during the development of pure vocal personal assistants for dependent persons during the vAssist project (Sansen et al. 2014). The project proposes a relatively affordable and simplified design of a human-sized humanoid robot that fits the requirements of this analysis. After an overall description of the robot, the justification of the novel choice of a handicapped robot in an electric wheel-chair, this paper emphasizes the technology that is used for the head and the face and the subsequent verbal and non-verbal communication capabilities of the robot, in turn highlighting the characteristics of Embodied Conversational Agents.

Lateral inhibitory networks: Synchrony, edge enhancement, and noise reduction

This paper investigates how layers of spiking neurons can be connected using lateral inhibition in different ways to bring about synchrony, reduce noise, and extract or enhance features. To illustrate the effects of the various connectivity regimes spectro-temporal speech data in the form of isolated digits is employed. The speech samples are preprocessed using the Lyons Passive Ear cochlear model, and then encoded into tonotopically arranged spike arrays using the BSA spiker algorithm. The spike arrays are then subjected to various lateral inhibitory connectivity regimes configured by two connectivity parameters, namely connection length and neighbourhood size. The combination of these parameters are demonstrated to produce various effects such as transient synchrony, reduction of noisy spikes, and sharpening of spectro-temporal features.

Privacy preserving encrypted phonetic search of speech data

This paper presents a strategy for enabling speech recognition to be performed in the cloud whilst preserving the privacy of users. The approach advocates a demarcation of responsibilities between the client and server-side components for performing the speech recognition task. On the client-side resides the acoustic model, which symbolically encodes the audio and encrypts the data before uploading to the server. The server-side then employs searchable encryption to enable the phonetic search of the speech content. Some preliminary results for speech encoding and searchable encryption are presented.

A new secure and lightweight searchable encryption scheme over encrypted cloud data

Searchable Encryption is an emerging cryptographic technique that enables searching capabilities over encrypted data on the cloud. In this paper, a novel searchable encryption scheme for the client-server architecture has been presented. The scheme exploits the properties of the modular inverse to generate a probabilistic trapdoor which facilitates the search over the secure inverted index table. We propose indistinguishability that is achieved by using the property of a probabilistic trapdoor. We design and implement a proof of concept prototype and test our scheme with a real dataset of files. We analyze the performance of our scheme against our claim of the scheme being light weight. The security analysis yields that our scheme assures a higher level of security as compared to other existing schemes.

Convolutional Neural Networks for Phoneme Recognition.

This paper presents a novel application of convolutional neural networks to phoneme recognition. The phonetic transcription of the TIMIT speech corpus is used to label spectrogram segments for training the convolutional neural network. A window of a fixed size slides over the spectrogram of the TIMIT utterances and the resulting spectrogram patches are assigned to the appropriate phone class by parsing TIMIT’s phone transcription. The convolutional neural network is the standard GoogLeNet implementation trained with stochastic gradient descent with mini batches. After training, phonetic rescoring is performed in the usual way to map the TIMIT phone set to the smaller standard set. Benchmark results are presented for comparison to other state-of-the-art approaches. Finally, conclusions and future directions with regard to extending the approach are discussed.

Accelerated encryption algorithms for secure storage and processing in the cloud

The objective of this paper is to outline the design specification, implementation and evaluation of a proposed accelerated encryption framework which deploys both homomorphic and symmetric-key encryptions to serve the privacy preserving processing; in particular, as a sub-system within the Privacy Preserving Speech Processing framework architecture as part of the PPSP-in-Cloud Platform. Following a preliminary study of GPU efficiency gains optimisations benchmarked for AES implementation we have addressed and resolved the Big Integer processing challenges in parallel implementation of bilinear pairing thus enabling the creation of partially homomorphic encryption schemes which facilitates applications such as speech processing in the encrypted domain on the cloud. This novel implementation has been validated in laboratory tests using a standard speech corpus and can be used for other application domains to support secure computation and privacy preserving big data storage/processing in the cloud.