Telanova Blogs


Does Google really give you the answers you want ?

Google Search

Many of us have been there, you want to book a hotel, you have a rough idea of where, when, how much to spend etc. You enter a search, and 4 hours later you are still trying to find a booking that matches what you need. Every additional search shows more and more possibilities without giving you the ideal answer.

A recent study shows that the current methods of search and discovery are ending up meeting only around 51% of our functional properties, and when it comes to quality and price it can be as low as 16% satisfaction.

The recent paper by Messaoud WB at the Universite de la Manouba, ran a study using traditional and a version of Behavioural Web Service Discovery (B-WSD) (B-WSD Approach and Validation: Use Satisfaction Survey Nov 2017). Looking at the different areas of needs of a consumer and surveying how satisfied the consumer was after the selection was made in each area, including Functional Properties, Non-Functional, Execution cost, Quality/Price, Waiting time.

In conclusion the way searching is performed now, with the explosion of data on the internet, is not always meeting our needs. Using B-WSD, a sequence of operations of a web service, leads to a translation of the the consumers needs, and once the problem is analysed B-WSD approach proposes the modelling of a discovery system allowing a search to be developed according to customer data.


Deep Reinforcement Learning may bring more efficient Deliveries

Truck doing a delivery

In the beginning it was sufficient to get the parcel to the destination. Highwaymen, think Dick Turpin, would regularly empty carriages as they travelled between destinations. Next came the next day service, ensuring parcels were collected and delivered successfully. Adding in different services brought mixture and variety. Then came the realisation that costs needed to be reduced to stay competitive, save fuel, and also protect the environment.

The Travelling Salesman Problem and Vehicle Routing Problems, are long standards in problems that need solutions. Each new solution has brought it’s rewards. The team of Mohammadreza Nazari, Afshin Oroojlooy, Lawrence V. Snyder, Martin Takáč have released their paper on their solution to the Vehicle Routing Problem using Deep Reinforcement Learning.

The solution is not better than the recent solution to TSP, but comes at a faster time for near identical solution. Their approach outperforming classical solutions in quality and computational time. All that it requires is some training. With an added advantage that it is scalable and doesn’t require a distance matrix calculation, and copes with vehicles of different sizes, and multiple loading back at the depot, and differing demands by the delivery points.

Next time you’re waiting in for that delivery, it may just be that the vehicle has been re-routed.


Download the full paper from Cornell University

Your pavements may be changing. Neural networking in the real world

Pavements changing
New Convolutional Neural Network research has opened up a crack of light on to the problem of broken pavements. A team of researchers at the College of Engineering in Shantou University have been analysing images of pavements to better understand them. It may be odd to think that pavement fixing was just “down to the council”. However, as technology has progressed, reporting has increased, and some may say fixing has decreased. However, understanding those cracks and more importantly being able to identify where the cracks are now, and where new cracks are appearing is essential to efficient council works.

So it has been discussed for decades on how best to photograph and catalogue, report, repair pavements which are an expensive overhead in these austere times. The team Zhun Fan, Yuming Wu and Jiewei Lu have just published their paper, (Automatic Pavement Crack Detection Based on Structured Prediction with the Convolutional Neural Network Feb 2018) describing their amazing technology and findings.

In their study they used images taken from an iphone5 of pavements in Beijing and another set of images from from French pavements. They ran a number of training and testing routines, both using current methods and their proposed method. The conclusion being that the proposed method shows a better performance of dealing with pavement texture, and predicts the crack structure much closer to manual methods that current automated methods. CNN has a good ability to learn about images, and the network architecture can be used for other image methods, not only photographic. Being close to the manual method means that soon continuous rapid photographs of pavements could be taken and cracks identified and labelled without human interaction.

Next time you see a broken paving slab, think that it may soon become a thing of the past as a drive by camera will be identifying and reporting the flaw sooner rather than later.

Telanova : Managed IT

As technology uses data smarter, sharing economy is growing


Shanghai, Friday, rain expected. Another rainy day in the city, leaving for work by public transport is quite popular, and carrying an umbrella for the regular days of rain the norm. However, all this might be about to change, and before you begin thinking of a conspiracy to change the weather, it’s the smaller item that is changing.

With advances in technology, and decrease in cost of components, there are now 10,000 umbrellas on the streets of Shanghai, each can be rented instantly. Some may remember when this happened before at the beginning of 2017, that project didn’t get off to such a great start, losing 300,000 umbrellas. Justin Jia, Zhejiang Tianwei Umbrellas, has taken up the challenge, the Chinese entrepreneur has created a pioneering sharing app for umbrellas, and currently you can rent one of the 10,000 umbrella’s in the city.

Using both a deposit, and then scoring on the person’s behaviour, eg. reporting , damaging, returning on time, etc. the data is key to the profitability. Rather than monitoring the umbrella, it is more the person that is being monitored. Much like you may improve your credit rating, soon you’ll be able to improve your share-ability rating.

So next time it is a rainy day, you soon may be able to leave your umbrella at home.


Self driving… the small things that will make a big difference.

BBC micro

Apologies to the younger people in the audience, but the image for today’s article might be classed as what your iPad is descended from, it’s a BBC Micro (Short for Microcomputer) if anyone asks. The relevance being that what is under the bonnet of a self driving car today, will in a few years time seem like the BBC Micro’s of the self driving world. Currently developers and technology architects are struggling with the cost, reliability, and cooling of the chips and computational power that they are having to put under the bonnet to create relatively safe and secure environments.

Over at Huawei Research Canada, the team have just released what could be a significant step forward in image prediction. There have been software models available (CDNA) that by using over 12 million points, have been able to give a good representation of what the next image would look like. In a sense this is predicting the future image on your dashcam. The processing power to deal with that amount of data as you can imagine is costly, and runs hot.

Thanks to work by the team at HRC ( Donglai Zhu∗, Hao Chen∗, Hengshuai Yao, Masoud Nosrati, Peyman Yadmellat, Yunfei Zhang ) (Practical Issues of Action-conditioned Next Image Prediction 9-Feb-2018) using a smaller tiling set of under 1 million points with a single-decoder feedforward architecture manages to give a comparable performance on image prediction.

Predicting images, ie. how the world around a car will look in the (very near) future is essential for matching how humans predict situations when driving. Eg. if you notice a jogger running towards the road on a trajectory that will cross the road, it allows you time to break should the jogger not stop. With automated car driving having predictive imaging, this same action of slowing can take place, ensuring other human drivers around the automated car will not be disturbed by sudden breaking when the jogger ends up running across the road.

The research by the HRC team is excellent, bringing together of 2 technologies to give lighter faster design, which will enable cheaper, more energy efficient vehicles in the future.

Email Facebook Google LinkedIn Twitter

We use cookies to provide you with the best possible experience in your interactions on our website

You agree to our use of cookies on your device by continuing to use our website

I understand