|
Project Data Analytics Community Newsletter
February 2020
We produce this newsletter to keep you plugged into the very latest developments, building capacity and understanding within the project delivery community. As part of this quest we are also mobilising our brand new (Project) Data Analyst apprenticeship. It’s equivalent to a foundation degree and will develop a brand new cadre of professionals who straddle the divide between project delivery and data analytics. If you are interested in being one of the very first then get in touch soon; more details can be found here.
|
|
|
|
Project, Programme and Portfolio Management
|
|
|
Why Projects Aren’t Complex
31/01/2020 – By Martin Paver
Our CEO was on holiday last week and couldn’t resist a controversial post on how we manage project complexity. It attracted a lot of discussion that will help us to challenge how we apply data analytics in a complex project environment. Read more ...
|
|
|
AI in Project Management
30/07/2019 – By Ron Schmelzer
Another article on how AI can help to improve how we can deliver projects. There are a lot of these about. We prefer to use ‘project delivery’ rather than ‘project management’ because it encompasses all the aspects of projects from logistics through to compensation events. Helpful article nevertheless for those new to the subject. Read more ...
|
|
|
Skanska: curbing equipment emissions through AI
14/01/2020 – By Construction Global – The Global Construction Platform
Climate change is an important challenge currently facing the world, and companies in all industries are under pressure to reduce their environmental impact. Skanska have worked with partners on a project part funded by the Norwegian government to use equipment operating data to better manage vehicles and machines. This provides multiple benefits in both reducing carbon emissions and also saving costs. Read more ...
|
|
|
Predictive climate and weather analytics & AI in construction
10/01/2020 – By Planning, BIM & Construction Today
A really interesting ebook from Metswift looking at how AI can be harnessed to improve weather prediction in construction. Current weather forecasts use models based on atmospheric physics, but only have a high accuracy level in the short to medium term before dropping off. Due to schedule changes and project planning, the construction industry has a need for more accurate long term forecasts. Metswift have made progress in long term predictions by using a machine learning model built off billions of historical weather data points. Read more ...
|
|
|
DfT completes AI health check for local roads
17/01/2020 – By Highways Magazine
The Department for Transport has completed a local road ‘health check’ using AI to check road and footway conditions and road markings. They say they will now analyse this data to target road maintenance in problematic areas. The health check mostly used data from an existing bank of images analysed by human inspectors, and built a model off these images. Through the model it is then possible to quickly inspect any section of road you have image data for, allowing identification of problem areas and targeted maintenance. Read more ...
|
|
|
AI Startup Digs Up Business Opportunity in Aging Water Pipes
29/01/2020 – By Yahoo Finance
Fracta is a startup that has developed a technology that allows utilities companies to identify the decay of water pipes before digging. The technology uses a machine learning model built from data including soil quality and the population density of an area. Read more ...
|
|
|
Revolutionary Artificial Intelligence warship contracts announced
14/01/2020 – By GOV.UK
The Ministry of Defence has announced funding to develop solutions to overcome the ‘information overload’ faced by crews of warships. The different projects will focus on automation and artificial intelligence, and will look to address the evolving threats to our national security. Read more ...
|
|
|
Open AI just chose PyTorch over Tensorflow
03/02/2020 – By Arun C Thomas
Open AI, an artificial intelligence research organization, have announced their primary framework for development will be PyTorch, an open source ML library. This will be a disappointment to Tensorflow, who recently released Tensorflow 2.0, however Open AI do state they will still use other machine learning frameworks as required. Read more ...
|
|
|
Towards a Conversational Agent that Can Chat About… Anything
28/01/2020 – By Google AI
Google AI have developed a chatbot, Meena, that is designed to conduct conversations that are more ‘sensible and specific’ than existing chatbots. They have also created a ‘Sensibleness and Specificity Average (SSA)’ to capture important attributes for human conversation. Read more ...
|
|
|
Can You Trust Your Model’s Uncertainty?
15/01/2020 – By Google AI
This blog post by Google details the paper they presented at NeurIPS 2019 on benchmarking state-of-the-art deep learning models. The idea behind the research is that often the data which a model is trained on may have a different distribution to the data we want to test or apply our model with. This work is important as it highlights the need to consider the uncertainty under dataset shift when training deep learning models. Read more ...
|
|
|
Dopamine and temporal difference learning: A fruitful relationship between neuroscience and AI
15/01/2020 – By Deepmind
A lot of recent progress in machine learning and deep learning has been inspired by neuroscience, the study of the brain. In the 1990’s a group of researchers versed in both neuroscience and AI noticed that particular neurons in the brain behaved like the TD learning algorithm. Recently there has been progress in distributional reinforcement learning combined with deep neural networks. This work posed the question whether this type of learning is used in the brain. A fascinating article to read for those with an interest in deep learning or neuroscience. Read more ...
|
|
|
Introduction to Streaming Algorithms
29/01/2020 – By Dr. Robert Kubler
An interesting article explaining the intuition and importance behind memory-efficient algorithms. Different algorithms are presented to approach the memory constraint problem. Read more ...
|
|
|
Is it easy to explain? Global explainability in Boosted Trees
23/01/2020 – By Dina Berenbaum
A series of two blog posts surveying existing methods for model explainability. Research in the area of explainability in AI is vital to ensure that complex ‘black box’ models are accountable, and there is an informed understanding of how a model is producing certain outputs. Read more ...
|
|
|
HiPlot: High-dimensional interactive plots made easy
31/01/2020 – By Facebook AI
This is a cool interactive visualization tool developed by Facebook to help discover correlations and patterns in high-dimensional data. It allows researchers to more easily evaluate the influence of hyperparameters and architecture of models. Visuals tools like this are great in helping researchers understand how changes they are making will affect the final model. Read more ...
|
|
|
|
Robotic Process Automation
|
|
|
Applica and UiPath Enter Into Technology Alliance to Drive Intelligent Automation
31/01/2020 – By AiThority
UiPath have partnered with Applica, a robotic text automation company, to improve and expand solutions designed to automate text-intensive processes. This will give users the ability to extract meaning from semi-structured and unstructured documents to transform them into actionable data. This partnership will hopefully widen the scope of processes that can be automated to include those containing more unstructured documents. Read more ...
|
|
|
|
|
