Understanding human movement is essential for the better design of human-machine interface. Researchers have made great efforts to develop cheap and easy-to-use wearable sensors and algorithms to achieve this goal. In this talk, I will present my previous works in this field, including movement intention detection and perception and cognition recognition using bio-signals collected by wearable sensors. First, I will talk about detection of arm and leg movement intention using ForceMyography, ElectroMyography, and Inertial Measurement Units. The applications of this line of studies are for human-machine interaction, prosthetic limb (robotic arm) control, and rehabilitation and assistive device. Second, I will discuss how to measure visual perception and cognition of surgeons using eye-tracking technology. Results gained can help us to assess mental workloads, team cognition of surgical team, and help to detect the moment of performance difficulty under AR/VR environment.
For decades, our collective view of how software is built was strongly influenced by unscientific, non-empirical "best practices" and experience reports. In recent years, empirical research has invalidated many common beliefs about software development, and generated many new theories, concepts, tools and techniques. In this talk, Dr. Ralph will summarize over a decade of empirical research at the intersection of software engineering and project management, including seminal research on decision-making, waste-reduction, software engineering success, product backlogs, cognitive biases in requirements engineering, and surviving disruption.
Model interpretability is a requirement in many applications in which crucial decisions are made by users relying on a model's outputs. The recent movement for ``algorithmic fairness" also stipulates explainability, and therefore interpretability of learning models. The most notable is ``a right to explanation" enforced in the widely-discussed provision of the European Union General Data Privacy Regulation (GDPR). And yet the most successful contemporary Machine Learning approaches, the Deep Neural Networks, produce models that are highly non-interpretable. Deep Neural Networks have achieved huge success at a wide spectrum of applications from language modeling and computer vision to speech recognition. However, nowadays, good performance alone is not sufficient to satisfy the needs of practical deployment where interpretability is demanded for cases involving ethics and mission critical applications. The complex models of Deep Neural Networks make it hard to understand and
reason the predictions, which hinders its further progress.
In this thesis proposal, we attempt to address this challenge by presenting two methodologies that demonstrate superior interpretability results on experimental data.
The first methodology is named as CNN-INTE. It interprets deep Convolutional Neural Networks (CNN) via meta-learning. In this work, we interpret a specific hidden layer of the deep CNN model on the MNIST image dataset. We use a clustering algorithm in a two-level structure to find the meta-level training data and Random Forest as base learning algorithms to generate the meta-level test data. The interpretation results are displayed visually via diagrams, which clearly indicate how a specific test instance is classified. Our method achieves global interpretability for all the test instances on the hidden layers without sacrificing the accuracy obtained by the original deep CNN model. This means our model is faithful to the original deep CNN model, which leads to reliable interpretations.
In the second methodology, we apply the Knowledge Distillation technique to distill Deep Neural Networks into decision trees in order to attain good performance and interpretability simultaneously. We formulate the problem at hand as a multi-output regression problem and the experiments demonstrate that the student model achieves significantly better accuracy performance (about 1% to 5%) than vanilla decision trees at the same level of tree depth. The experiments are implemented on the TensorFlow platform to make it scalable to big datasets. To the best of our knowledge, we are the first to distill Deep Neural Networks into vanilla decision trees on multi-class datasets.
In the end, we propose a visualization technique for future work.
Dr. Stan Matwin - Faculty of Computer Science (Supervisor)
Dr. Thomas Trappenberg - Faculty of Computer Science (Reader)
Dr. Sageev Oore - Faculty of Computer Science (Reader)
Dr. Fernando Paulovich - Faculty of Computer Science (External Examiner)
Public art festivals provide a unique opportunity for Human-Computer
Interaction (HCI) research. They attract a diverse population interested
in engaging in novel experiences, foster a lively crowd dynamic, provide
a variety of interesting public settings, and can be a remarkably
efficient means of collecting participant data. Yet pitfalls abound: for
example, the desire to provide a good experience can trump scientific
objectives, onsite testing can be difficult, and small mishaps can have
disastrous consequences for data collection. This talk reviews six
studies conducted at art festivals, highlighting successes and failures
in each, and then offers a set of hard-won recommendations, useful for
researchers who might consider a similar approach: have concrete
research objectives as well as higher level interests, use agile
approaches to system building, balance audience engagement and feature
simplicity, articulate backup plans when things go wrong but remain
nimble, consider research as spectacle.
Come join us for an evening of great presentations, coverage of what’s happening in the blockchain ecosystem, food and drinks and an all around exhilarating evening.
Some members have expressed interest in presenting.
The Groundhog team will be presenting an overview of our product and what we’ve been up to in 2018.
Halifax IIBA are pleased to invite you to spend an evening with Bob the BA, a highly acclaimed business analyst educator and presenter.
Bob will be presenting Influencing without Authority!
Register online at:
Seafloor habitat mapping in an ocean of big data: Development of data analysis approaches for map production
Over the past two decades advances in the field of ocean technology have led to the exponential increase in volumes of oceanographic data. Acoustic remote sensing, autonomous surface and underwater platforms, in-situ sensor platforms, and vessel-deployed instruments are now capable of collecting extremely large, diverse and interconnected oceanographic data sets for a wide range of applications. The technology has reached the state where ocean data are being generated at a rate faster than can be assessed and interpreted using traditional methods. The need for development of analytical tools to process these data sets, coupled with the skilled individuals to undertake these analyses is now paramount.
We present case study examples of applied research activities in the field of integrated ocean mapping at the Nova Scotia Community College (NSCC) where analytical procedures are under development to handle, process and deliver results from large oceanographic data sets in support of ocean sector stakeholders. NSCC works closely with regional and Canadian-based companies, federal and provincial government departments, and other academic institutions to provide innovative applied research solutions in the area of ocean technology and ocean analytics. An overview of various habitat mapping approaches will be provided to illustrate advances in this field of research.
Join Engineers Nova Scotia for this new PD Session which will feature several ethics and safety related case studies. This interactive session will allow participants the opportunity to discuss and explore some common ethical issues encountered today.
The human brain is one of the most complicated biological systems in the world. The brain activities measured by various signals such as electroencephalogram (EEG), electrocorticogram (ECoG), and functional magnetic resonance imaging (fMRI) provide avenues that can help understand the underlying mechanisms of the brain as well as diagnosis brain disorders and the related diseases. However, without the proper techniques to analyze the brain signals, they are of limited value. In this talk, I will discuss the challenges in brain signal analysis and emphasize the role of machine learning techniques in feature extraction and classification of EEG/ECoG signals. From an algorithmic perspective, I will present multitask learning techniques that aim to discover the common structure that is shared across the brain signals from different subjects to improve the learning performances. In addition, I will also discuss some theoretical aspects of multitask learning, and address two fundamental questions: First, compared with single-task learning, why multitask learning can succeed? Second, under what conditions multitask learning can succeed?
t is that time of the year to celebrate! The IIBA Halifax Chapter cordially invites you to our annual Holiday Social. Please join the Board of Directors and fellow IIBA members for an evening containing light refreshments, entertaining company and great conversation. Come and enjoy this excellent opportunity to meet and network with our local CBAP certified professionals. This gathering is a special evening to bring existing, old, new and future members together to enjoy the benefits of our local IIBA Chapter. Do not miss this chance to meet colleagues from the Halifax business analysis, management consulting, service management and IT professional communities. Let's celebrate this year's accomplishments and prepare to make new commitments for the coming year. Chapter and non-chapter members are welcome.
Register in advance for complimentary drink ticket:
Declarative Performance: Automating Performance Ops with Kubernetes
Domenic Rosati, Engineering @ Manifold
Performance management is really hard. There are lots of tools that help like APMs and other observability tools. Ensuring your system is highly available. Setting SLOs and observing SLIs.
Kubernetes helps by providing a self healing platform that reconciles current system state with declared ideal state. Is there a way to declare ideal performance and leverage Kubernetes to attempt to meet those objectives automatically?
Can data science be ethical? And why should I care? On DNA, Gamergate, taxi rides, and sea rescue operations
Ethics is everywhere. We need to get our research plans approved by ethics boards, we now do ethical AI (based on ethics codes), the European Union makes data-science projects have an independent ethics advisor, and management now engages in ethical leadership.
Why is this, and what does ethics actually mean in the context of data science? In this talk, I will give a hands-on introduction to important schools of thought and questions from this fascinating subfield of philosophy. The “hands-on” means that we will, interactively, go through a number of real-life case studies from data science, study what values and rights are at stake and how they were and can be disregarded, respected, protected, and questioned.
In the second part of the talk, I will focus on a specific application of ethical questioning: the analysis of vehicle/human trajectory data, and a specific value: privacy. I will discuss two recent examples of the analysis of such data – the New York City taxi rides dataset, and the use of data from the maritime Automatic Information System (AIS) for mapping refugee movements on the Mediterranean Sea. The examples will illustrate a feature that engineers often find very difficult to deal with: the tension between allowing for different (and often mutually incompatible) ethical stances on the one hand, and requiring adherence to certain ethical norms that are considered non-negotiable on the other hand. But these examples will also illustrate why we should care, why it is intellectually stimulating to think about ethics, and why doing so requires us to also question ethics or “ethical” codes, boards and advisors, and certainly leaders.
In this talk, I will present two contrasting approaches to building generative models for audio with deep learning techniques:
1) In the TimbreTron system (developed in collaboration with students and faculty at U of Toronto and Vector), we learn to manipulate the timbre of a sound sample from one instrument to match that of another while preserving musical content such as pitch and rhythm. I will describe how we do this by combining a CycleGAN architecture, appropriate spectral representations and a conditional WaveNet synthesizer.
2) In PerformanceRNN (developed in collaboration with researchers at Google, and with subsequent developments in collaboration with undergraduate students at Dalhousie FCS), we work directly with MIDI data rather than raw audio, which allows us to treat music generation as language-modeling problem. We use a conditional LSTM to generate solo piano music based on a dataset of human performances.
I will also provide overviews as needed throughout the talk of concepts related both to audio generation (e.g. "What is MIDI? What are spectral representations?") as well as to deep learning techniques (e.g. "What is CycleGAN?”).
The panel will discuss the ethical challenges arising as artificial intelligence permeates our lives.
Many people believe that we need to open AI code and "training data" in order to be transparent about the important decisions being made when it comes to AI tools today. But, to achieve that level of transparency, we take the risk that these tools will be manipulated or abused.
If machine learning continues to progress, we'll be able to introduce and enforce ethics rules that regulate the development, training and use of AI tools - or, our "electronic children" - and maybe even reinforce these rules. Would these rules be universally accepted by all? Would groups or communities not bound by these rules break them to gain economic or political advantages?
Join us as we embark on this discussion with panelists:
Dr. Darren Abramson, Department of Philosophy, Dalhousie University
Dr. Fosca Giannotti, Director, Research at the Information Science and Technology Institute “A. Faedo” of the National Research Council, Pisa, Italy
Dr. Stan Matwin, Director, Institute for Big Data Analytics, Dalhousie University
Reasoning on data and algorithmic bias: explaining the network effect in opinion dynamics and the training data bias in machine learning
Data science is creating novel means to study the complexity of our societies and to measure, understand and predict social phenomena. My seminar gives an overview of recent research at the Knowledge Discovery (KDD) Lab in Pisa within the SoBigData.eu research infrastructure, targeted at explaining the effects of data and algorithmic bias in different domains, using both data-driven and model-driven arguments.
First, I introduce a model showing how algorithmic bias instilled in an opinion diffusion process artificially yields increased polarisation, fragmentation and instability in a population. Second, I focus on the urgent open challenge of how to construct meaningful explanations of opaque AI/ML black-box decision systems, introducing the local-to-global framework for the explanation of ML classifiers.
The two cases show how the combination of data-driven and model-driven interdisciplinary research has a huge potential to shed new light on complex phenomena like discrimination and polarisation, as well as to explain how decision making black-boxes, both human and artificial, actually work.
I conclude with an account of the open data science paradigm pursued in SoBigData.eu Research Infrastructure and its importance for interdisciplinary data driven science that impacts societal challenges.
The Institute for Big Data Analytics is turning five. Headquartered at Dalhousie University’s Faculty of Computer Science, the Institute for Big Data Analytics has become an international hub of excellence in big data research – supporting local industry to use big data to make an impact and training the next generation of researchers and practitioners to advance this area of innovation.
As the Institute for Big Data Analytics reaches this milestone, we will acknowledge the progress made in this area for Dalhousie and the region, look back on the research advances made, and look ahead to the next five years. This is an opportunity to not only focus on the direction the Institute for Big Data Analytics will take to align with the university’s and regions’ strategic priorities, but also to discuss the future of big data and the opportunities and challenges we face as a society.
Join us for speeches, student research poster presentations, and refreshments.
Practice and develop your Business Analyst skill set by having a business date with the IIBA Halifax Chapter. Through the evening you and your fellow #HaliBAs will practice some of BABOK techniques that are less often used than the general Brainstorming and Requirements Facilitation Workshops. Come join us for an evening of fun, learning and laughter.
5:30 - 6:00pm: Networking
20 mins: #HaliBA Certified - What does it take?
20 mins: What is #HaliBA?
20 mins: Living your best #HaliBA Life - How to Network
7:00 - 7:30pm: Networking
We will be discussing how drones or UAVs have become an important tool for engineering purposes. Their versatility has made it possible to be implemented in almost every field. Applications go from construction, inspections, asset management, surveying, energy audits, and more. Everyday there are new uses in this ever expanding field. Join us as we discuss how they can be used and applied to a variety of engineering environments.
Are you short on time to start from scratch to use deep learning to solve complex problems involving topics like neural networks and reinforcement learning? If yes, then this is the workshop to help you.
This workshop is designed to help you to overcome various data science problems by using efficient deep learning models built in TensorFlow.
The aim of this talk is to present research results of the Complex Event Recognition lab (http://cer.iit.demokritos.gr/) of NCSR Demokritos (Athens, Greece), focusing on maritime applications. Maritime monitoring systems support safe shipping, through real-time detection of dangerous, suspicious and illegal vessel activities. We have been developing a complex event recognition system for maritime monitoring in the Event Calculus, allowing both for verification and real-time performance. The basic system is being developed through collaboration with domain experts, constructing effective patterns of maritime activity. In order to refine these patterns, we have developed online, relational learning techniques and applied them on AIS data streams. More recently, we have also been developing complex event forecasting techniques, allowing for predictive maritime analytics. In the talk, we will show results of our techniques on real AIS streams, covering large geographical areas.
The Faculty of Arts and Social Sciences, in cooperation with the Faculty of Medicine and the Faculty of Computer Science, is pleased to invite you to a special panel event to introduce the three Donald Hill Family Postdoctoral Fellows.Their panel will discuss their research intentions and generate stimulating dialogue about the impact of emerging technology on society.
Please join us for this panel event. All are welcome!
Reception to follow.
Our 20th Anniversary Event will be held this September 26th, 2018 at Pier 21. To kick things off, we will begin the evening with our AGM – with a new twist this year.
As part of its 20th Anniversary Year, PMI NS is launching online voting for Board Elections. Call for Nominations was issued July 23rd and is open for nominations until August 24th. Plan to attend our AGM for the firsthand announcement of our new 2019/2020 Board of Directors. In 2019 we are expanding our Board compliment to include two new Director positions: Director of Technology, and Director of Innovation.
PMI Nova Scotia Chapter’s Anniversary Party will feature a walking menu of Nova Scotia favorites old and new, cash bar, local live musicians, views of Halifax Harbour and even access to one of the interactive Pier 21 exhibits. There will be multiple door prizes to be won, gifts for attendees, and for one lucky winner a GRAND PRIZE! Stay tuned for details in the weeks ahead.
A panel of 3 local recruiters will meet and discuss recruitment in the tech sector.
They will discuss trends (local/national), give advice and answer questions.
As always refreshments will be included
We present a collection of lessons learned from the deployment of real world wireless sensor network applications and the corresponding research issues. Our experiences are based on our work with the SmartCondo environment and from residential building sensing. Namely, we will consider: sensor placement, seen as a variety of the "coverage" problem; sensor autonomy, seen from the niche of thermoelectric energy harvesting; and sensor "fusion" from combinations of anonymous and eponymous sensing. We will also address to some extent the problem of sensor data outages -- a phenomenon not uncommon in real environments -- caused by energy depletion or by the unpredictable nature of the wireless propagation environment. We will highlight previous as well as ongoing work to characterize and mitigate such outages.
[Webinar] National industry engagement technical workshops for the Information and Communications Technology (ICT) sector
The CRA is launching a first: a series of national workshops aimed at directly supporting businesses conducting R&D in the ICT sector. During these workshops, businesses will gain valuable insights into how to best relate their innovation to SR&ED. Useful tips and best practices will also be provided which will help in ensuring businesses receive their full entitlements under the SR&ED program. Also this is an opportunity for businesses to engage directly with the CRA.
The 18th ACM Symposium on Document Engineering (DocEng 2018) seeks original research papers that focus on the design, implementation, development, management, use and evaluation of advanced systems where document and document collections play a key role. DocEng emphasizes innovative approaches to document engineering technology, use of documents and document collections in real-world applications, novel principles, tools and processes that improve our ability to create, manage, maintain, share, and productively use these. In particular, DocEng 2018 seeks works involving large-scale document engineering applications of industrial relevance.
We consider trophic networks, a kind of networks used in ecology to represent feeding interactions (what-eats-what) in an ecosystem. Starting from the observation that trophic networks can be naturally modelled as Petri nets, we explore the possibility of using Petri nets for the analysis and simulation of trophic networks. We define and discuss different continuous Petri net models, whose level of accuracy depends on the information available for the modelled trophic network. The simplest Petri net model we construct just relies on the topology of the network.
Trajectory mining has applications including but not limited to transportation mode detection, tourism, traffic congestion, smart cities management, animal behaviour analysis, environmental preservation, and traffic dynamics are some of the trajectory mining applications. Transportation modes prediction as one of the tasks in human mobility and vehicle mobility applications plays an important role in resource allocation, traffic management systems, tourism planning and accident detection. In this work, the proposed framework in Etemad et al. is extended to consider other aspects in the task of transportation modes prediction.