The 7th Machine Learning Clinic Expert Group meeting was held at the new ELCA Innovation Lab on the EPFL campus on May 3, 2018. We were hosted by Silvia Quarteroni, Senior Manager at ELCA in Lausanne (thanks a lot!).
In her talk, Silvia discussed the main “hard nuts to crack” in applied machine learning in the context of ELCA’s customer projects and R&D: “Since we have a strong focus on natural language processing, many of our issues and opportunities concern subjects such as language modelling for natural language understanding, dialogue management in virtual assistants, and dealing with the complex linguistic landscape of Switzerland.” However, the talk also presented challenges in image analysis, speech processing and statistical prediction.
Martin Jaggi, Assistant Professor at EPFL, gave a presentation about his current research work on Unsupervised Machine Learning for Text, which is performed in the EPFL Machine Learning and Optimization Laboratory. Martin showed methods to analyse large scale data, including distributed training algorithms to optimize the training time. Results were given for several industrial systems (from IBM, Google and other providers) which implement these algorithms. We had a discussion on benchmarking the performance of different systems for such tasks.
The next ML Clinic meeting will be held in September. Stay tuned!
On April 19, 2018, the Innovation Forum FORWARD was held at EPFL. About 900 participants from SMBs, large companies, academia and government organizations attended a rich set of keynote presentations, which focused on the on-going digital transformation of SMBs. Several success stories were shared, such as LauraStar, Ecole Moser, Sensile Technologie, JMNC Lutherie and SenseFly. EPFL presented its digital strategy and the new EPFL Extension School. Andre Kudelski and Nicola Thibaudeau presented the Innosuisse strategy and innovation support programmes.
During the afternoon several breakout sessions (Ateliers) were held and the Swiss Alliance for Data Intensive Services in collaboration with Innosuisse hosted the Atelier F2. During the Atelier F2 Gundula Heinatz gave a presentation of the Alliance, and Silvia Quarteroni (ELCA) and Giovanna Di Marzo Serugendo (University of Geneva) showed their activities and contributions to the Alliance. Before the event, Innosuisse and there Alliance had collected proposals prepared by startups and SMBs for an innovation project for data-driven product or services. At the end of the Atelier F2 the 3 best submissions received were presented: by Rob Hamilton Smith from OrchardAI, by Marc Tesch from LeanBI and by Robert von Känel from Panache. These proposals will receive dedicated help and coaching to create an industry-academic partnership and apply for an Innocheck.
On April 12th a joint event with the Data+Service Alliance was conducted in PwC’s Experience Center in Zurich. The Experience Center is a location which has been designed to foster creative thinking and to get a first-hand impression of digital innovations. More than 30 participants from Swiss corporations attended the event to learn how their organisations can generate value from data analytics.
As a first keynote speaker, Ralf-Peter Simon from the International Data Space Association (IDSA) introduced a platform, where companies and institutions can provide and share data in a secure and trustful manner. Each organization can govern and monetize data exported to third parties while staying sovereign of their own data, or can access and use foreign data to improve their own processes. The International Data Spaces Association is has created a reference architecture to implement secure and trustworthy data exchanges. Ralf-Peter presented use cases built using it from the more than 80 European member companies.
In a second keynote, a more company-centric view was presented by Gundula Heinatz Bürki. Gundula described her experience in building up the analytics unit at Mobiliar – her role before becoming General Manager of the Swiss Alliance for Data-Intensive Services. She stressed that not only the support from the executive level is necessary – especially during the initial growth phase – but also quick wins to get visibility and momentum with the business units are essential for long-term success. Finally, an organization’s IT department needs to get actively involved to make sure that new ideas get implemented within a reasonable time horizon.
After discussing questions from the audience, the participants toured a string of experience stations, displaying innovative analytics solutions created by PwC teams and concluded the event at an apéro.
We are very thankful to the participants, the excellent presenters and great discussions.
The Expert group seminary started with meeting on the demonstrator project. The initial idea of using the OTW-Headstock from BW-Tech was abandoned, however several ideas of new systems to use was discussed. Several requirement for a demonstrator such as possibility do simulate multiple failure/stress modes, being able to run it in building without specific requirement, being able to cover both mechanical and electrical failures. To put some momentum behind this project, a next meeting will take directly after Easter. This is great opportunity to show end to end solution which can foster innovation in the different disciplines involved.
The seminary itself included 6 different presentations, spanning from sensor from conditions monitoring railway applications, sensing of stability level of gas turbine combustion flames, chair lift clamping force, energy harvesting and smart sensor for motors and pumps.
It was very interesting presentation, including both technical depth as well high level discussion about the sensors role in IOT and predictive maintenance.
I workshop to discuss and document the requirement form the participants was planned but due to time reasons it was moved to the next meeting planned in June. A longer workshop will be planned where we can document and catch all opinions on sensor development to support predictice maintence, and this is a big topic. The variety of participants provides a great to catch all different requirement and ideas.
The meeting ended with an Apero and some food.
We are very thankful to all the people attending the seminary, the excellent presenters and the input during the discussion.
Recently, the Swiss Alliance for Data-Intensive Services co-organised the “Digitization and Innovation” workshop at the national “Konferenz Digitale Schweiz”. At the heart if the conference, overall seven workshops took place to revise Switzerland’s future digital strategy by including the ideas of more than 500 invited stakeholders. It has been an immense honour for us to be asked together with CTI to contribute to what happened to be the best attended workshop of the conference (nearly 180 participants). It also underlines the recognized position of our Alliance as the hub for digital innovation in Switzerland.
The workshop started by looking at some of the success stories from our invited panelists. They demonstrated essentially two different types of digitization-driven innovations:
On one side of the spectrum are innovations that require intensive research and development, which results in a new product. Such innovations are often implemented through spin-offs and start-ups. It is important for such companies to have access to well-trained employees and to be able to grow fast enough so that continuing business from and in Switzerland is viable.
On the other side of the spectrum are innovations that are not based on a new hi-tech product or algorithm, but rather on an innovative combination of existing digital technologies with new process and business models to better meet customer needs. For the success of such innovations, the speed from idea to market is crucial. Correspondingly, such innovations are agile and closely dependent on the market.
There was a great consensus in the workshop that networks and co-operations are becoming increasingly important for all types of innovation activities. It is crucial that the right players work closely together for success. Such networks are more and more interdisciplinary and are formed by actors from different sectors.
Start-ups especially are perceived as drivers of innovation. However, also existing small and medium-sized enterprises can profit from digitization, even if they today are still apprehensive to digital transformation. Here there is a need for further awareness and establishing a corresponding culture of innovation. It is not only about reducing fears, but also about helping to recognize and realize the opportunities of digitization.
This is what we, at the Data+Service Alliance are striving to achieve at our core: fostering data-driven innovation in Switzerland through industrial, academic and individual cooperation. Together we move faster.
Activity Data-based process optimization – A workshop at TRUMPF AG in Grüsch
The Swiss Alliance for Data-Intensive Services provides a significant contribution to make Switzerland an internationally recognized hub for data-driven value creation. To transfer this mission into industrial practice and measurable use LeanBi is, besides other enterprises, an early and active member of this coalition: LeanBI offers data analysis and data modelling workshops to support members in their concrete and daily data based operations and issues. As a member of the Data+Service alliance, some days ago TRUMPF Laser Marking Systems AG located in Grüsch (GR) conducted such a data workshop together with LeanBI.
TRUMPF Laser Marking Systems AG
TRUMPF is a high-tech company offering manufacturing solutions in the fields of machine tools, laser technology and electronics. TRUMPF is driving the digital networking of the manufacturing industry through consultancy, platform products and software products. With more than 70 subsidiaries TRUMPF is represented worldwide in all major markets.
One of these subsidiaries is TRUMPF Laser Marking Systems AG, located in Grüsch (GR) with development and production of Laser Marking Systems. This subsidiary offers a wide variety of marking lasers in many different power classes and with all standard wavelengths (infrared, green, ultraviolet). These lasers are ideal for marking processes such as engraving, ablation, annealing, discoloration, and foaming. In addition to metals, marking lasers from TRUMPF also process many other materials such as plastics, glass, silicon, ceramics and organic substances.
The workshop: Data analytics and the future use of data
Evaluating the best laser marking settings can be very time consuming, as the optimum marking parameters first needs to be determined. And these parameters are highly material dependent. As a first step to reduce the processing time in application centres significantly, TRUMPF together with LeanBI set up a first common workshop. The primary objective of this workshop was to get some early insights into what is possible to achieve with analytic algorithms on already available historic marking process data. In addition to this analysis TRUMPF wanted to identify the main tasks needed to implement structured data collection as a basis. And by this, finally, to reduce costs and the throughput time of application tests for external customers.
First insights of LeanBI’s data analysis
LeanBI created a demo on what is possible with existing data analytics algorithms on existing data with the following results:
Data quality is one of the most important issues, i.e. the know-how is necessary, which data are really important and in which granularity those data have to be collected -> one needs to know what has to be to achieved via data analytics and then a standardized way to collect the necessary data into a structured data base has to be implemented.
Most of TRUMPFs speculations about the influence of all the different laser parameters, which affect the result of the marking quality, were confirmed by the first analysis of LeanBI. On the other hand, there were surprising additional aspects as well.
The overall most time consuming activity in the laser application centers worldwide are application tests for standard objects. The reduction of the processing time for such standard objects is very interesting from an economic point of view.
One of the big questions is: Is it possible to automate the process for standard laser applications partly or even completely? This would help to save expensive development resources, which then could be used for other, more complex and highly specific projects.
A first Pilot Project
During the second part of the workshop also the objective of a first pilot project was defined to use intelligent data analytics algorithms for the prediction of laser parameters based on existing future data. Considering in this pilot project only a very limited field of applications could help to realize a rather quick MVP for valuable results. The main work packages of this pilot project were discussed and responsibilities were defined in a coarse project plan.
Next steps?
This first pilot project will work as a proof of concept (PoC) and will be the basis for a later complete project roadmap to be set up after successful completion of the PoC. After evaluating the pilot, this roadmap will be generated mainly under economical aspects, i.e. based on the respective business cases.
The “Voice“ of TRUMPF Laser Marking Systems AG
How does Mr. Matthias Werner, project manager digitalization at TRUMPF Laser Marking System AG, comment the workshop results? “The digital transformation helps us to improve productivity along the whole order-to-cash process. Particularly, in non-value-adding areas we have the biggest technical and economical potentials. In general we work on questions like: How can data help us to make processes more efficient and which data do we need in the future to enable the effective use of intelligent data algorithms? This is one of the projects, which helps us to improve the productivity of our order-to-cash process.”
LeanBI and Swiss Alliance for Data-intensive Services
LeanBI offers for all members of Swiss Alliance for Data-Intensive Services workshops in the areas of Business Intelligence (BI), Big Data and Industry 4.0.
By this, the mental hurdle to implement use-oriented and data-driven analysis and algorithmic methods shall be lowered. But the purpose is also to support interested companies in their daily operative challenges based on data analytics. In order to generate concrete and measurable use. A list of all services offered by the Data+Service alliance members are available here, and also the service provided by LeanBI particularly: “Innovation Bootcamp on Predictive Analytics”.
Conclusion
A first promising step is made. The data available at TRUMPF Laser Marking Systems AG were converted into first valuable results. And a coarse planning of a pilot project is done (start in Nov. 2017). The outcome of this project will serve as an important input for a project roadmap to implement further marking process optimizations.
Data+Service alliance is happy to be part of the “Industry 4.0 Conference”, organised by the Institute for Data Analysis and Process Design (IDP), ZHAW and Industrie2025. The conference will take place on the 6th of Sept. 2017 at the ZHAW campus at Winterthur.
One of the focus areas of the Data+Service alliance is Industry 4.0 and it strives to bring the Swiss hardware industries on the same platform with the data and services industries so as to realise the vast innovation possibilities brought out by automation and data exchange.
The Industry 4.0 conference presents concrete perspectives for Swiss SMEs to meet the new challenges in the age of industry 4.0. It is intended to help companies take the next step in the development of digitalisation. The focus of the conference is on the economic benefits that can be achieved through the use of industry 4.0 technologies. Representatives of companies and universities will provide concrete examples of what they have already done and report on their experiences as well.
At the conference, will also be representatives from various Swiss organisations that offer support research projects. They will advise companies about the various funding opportunities that are available to initiate an industry 4.0 innovation research project.
The conference still has early bird prices ongoing, but only for a short time now! The members of the Data+Service alliance benefit from the 30% discount that is being offered by the conference organisers. The industrial members of the alliance can also pay via their Flex Budget*. There are discounted price for students as well, information on which is found on the website.
We look forward to your participation at the conference.
* Flex Budget: Up to 60% of the cash contribution of an Industrial member of the Data+Service alliance is considered as Flex Budget. This amount can be used to buy services from the alliance. Please contact us if you need more information on it.
The Swiss Alliance for Data-Intensive Services just completed its 1st quarter of operations. During this period, we are happy to say we accomplished a number of tasks, some with expected results, and in some with extraordinary results.
This post will highlight some of those achievements and activities that our members would find interesting to note.
Conferences
The academic members of the alliance contribute with bringing in conferences that they have been organising, into the network. The benefit for the organiser is that they receive support from the alliance with program elements, organisers have the possibility to gain sponsors from the alliance’s members, there is a wider circle of dissemination and last but not the least, a possibility of financial contribution from the alliance.
The alliance benefits from this model by having a rich conference/events program throughout the year which its members can avail as well as providing discounted registration to its members.
Current & future events committed and scheduled:
The main conference of the alliance is the Swiss Data Science conference, which is going to take place on the 16th of June at Kursaal Bern. The program also includes a pre-conference event on 15th of June at Kursaal Bern, the “Data Science Bootcamp”. Please find more information on these two events on the website www.sds2017.ch.
We have the pleasure to claim that there are already 7 active expert groups run by the alliance. The topics of these expert groups encompass the wide area of data & service science and each have the goal to foster R&D activities and ultimately enable transfer of know-how from academia to industry and vice versa.
The following figure depicts a clear picture of the active expert groups and the statistics behind each group. More information on each of these expert groups can be found at www.data-service-alliance.ch/activities.
Education
There are 3 pillars of Data+Service Alliance: Innovation, Education and Inspiration. Therefore education occupies an important position in the activities and offerings of the alliance. The academic members have the obligation of including educational courses (CAS/MAS/DAS) offered by their university in the NTN program. These educational courses can be availed by the industry members of the alliance and paid through the flex budget model.
Currently there are 18 courses that are offered by the Data+Service alliance and the details on these courses can be found at www.data-service-alliance.ch/education.
In addition to these achievements and offerings of the alliance, we are also closely working to foster project activities. The alliance offers support to industries as well as academic institutions to find ideal project partners from its pool of members and offers support and guidance in terms of providing vital information regarding funding opportunities through the various funding agencies like CTI & EUResearch.
We encourage you to come forward if you are looking for answers concerned to any of the activities and offerings of the alliance. The broad network of experts ensure solution to a plethora of problems, if not all!
The recently formed expert group “Data Ethics” is going to have its kick-off meeting on the 4th of April 2017 at Katholische Hochschulgemeinde aki, at Zürich.
Issues such as data protection, workplace surveillance, or the potential consequences of future technologies require constant ethical reflection and stakeholder dialogue. This expert group will offer a forum to discuss ethical issues that matter to its members in their respective contexts, enabling them to develop concrete solutions for ethical challenges. Such issues may range from professional ethics (legitimate conduct as a professional in various work-related scenarios) to organisational aspects (data collection, usage, and processing within companies) to legal and political issues (data protection, industry agreements, codes of conducts, standards and labels, and public decision-making processes). The main goal of the group is to develop and foster competence regarding ethical issues in data usage, and to make sure that state-of-the-art approaches are shared so they can be applied by group members in their respective fields.
Potential activities of the expert group may include:
discussing specific industry cases and options for ethically sound implementation of data usage
organizing events with nationally and internationally renowned experts on ethics of data usage in order to exchange ideas and learn about existing standards in Switzerland and abroad
developing ethical standards for companies and organisations
discussing, preparing, and presenting input to political decision makers
Benefits of attending this group are not limited to networking but also include self-reflection, professionalization in the field of data ethics, and laying the groundwork for tangible long-term outcomes, such as codes of conduct, standards or labels regarding the legitimate use of data in Switzerland.
The first meeting will have a round of introduction and topics of interest from each participant. On the basis of these, activities and goals of the group will be decided and ways to actively shape the group will be discussed.
The group is lead by Prof. Dr. Christoph Heitz from ZHAW and Dr. Thorsten Busch from Lassalle-Institute as the academic leads and Dr. Gundula Heinatz from Swiss Mobiliar as the industry lead. The participation in the group meeting is limited to the members of the Data+Service alliance, therefore if you are already a member of the alliance and would like to participate in the first meeting, please directly get in touch with the group leaders. However if you are not a member of the alliance yet, please get in touch with us to discuss the various member benefits and the different membership plans.
We look forward to hearing from you and have your participation in the expert group meeting!
Data+Service is proud of its concept to reliably come up with innovations. We call it the “innovation funnel”: It is based on the belief that innovation happens when the right people with the right needs and ideas (and, yes, chemistry) meet and discuss; so we built a funnel of successively more concrete opportunities to meet, from broad and large conferences to smaller and more focused workshops – and finally expert groups. We assume that bringing together experts to discuss lessons learned as well as current challenges in an intimate setting leads sooner or later to innovative ideas, which can be fully developed outside of the group in a collaborative project.
The alliance is extremely happy that this happened successfully already at one of the first meetings of the expert group “Machine Learning Clinic”: While discussing use cases as well as personal interests, it became clear that several partner shared a common interest in the challenge of applying machine learning to itself – that is, in automating certain aspects of the process of build models of data by learning algorithms.
Originating in spirit from our member PwC, the idea was caught up by researchers from ZHAW and EPFL, and the trio had been busy during Nov-Jan to produce a proposal for funding to CTI. With great pleasure we can announce today that our proposal for “Ada – the artificial data analyst” has been accepted, and development will start soon to raise the productivity of data science endeavours by applying data science to itself (applying empirical optimisation also to algorithm and feature selection). Recent developments, e.g. from the MIT, are thus made available as a data product for Swiss industry.
Another CTI funded R&D project that came out of the “Machine Learning Clinic” is the QualitAI project between the ZHAW Datalab and the industrial member BW-TEC AG. This project will innovate in the area of deep learning on small data sets for industrial quality control on images.
The above 2 projects are examples of the first successful joint project applications between network members.
The Machine Learning Clinic expert group is going to meet again on the 6th of April 2017. The expert group is lead by Dr. Thilo Stadelmann from ZHAW as the academic lead and Mr. Michel Benard from Google Zurich as the industry lead. Participation is limited to the members of the Data+Service alliance, therefore if you are already a member and would like to participate in the meeting, directly get in touch with the leads. If however you are not a member of the alliance and would like to know about the various member benefits and the several membership plans, please get in touch with us.
We look forward to hearing from you and your participation in the group!
