Carving Through Data

A different introductory course to Machine Learning

9-13 March 2020, Fiescheralp, Switzerland

 

Carving Through Data

A different introductory course to Machine Learning

9-13 March 2020, Fiescheralp, Valais, Switzerland

Who we are

 

A 2017 ETH Board national Data Science initiative resulted in the creation of a unique joint venture between EPFL and ETH Zurich: the Swiss Data Science Center. The Center’s mission is to accelerate the use of data science and machine learning techniques within academic disciplines of the ETH Domain, the Swiss academic community at large, and the industrial sector. A multi-disciplinary team of senior data scientists and experts in domains such as personalized health and medicine, earth and environmental science, social science and digital humanities, as well as economics enables collaboration on both academic and industrial projects. This unique positioning, at the crossroad of academic excellence and fast-paced business environments agility is key in making the complex data science journey simple.

 

Who we are

 

A 2017 ETH Board national Data Science initiative resulted in the creation of a unique joint venture between EPFL and ETH Zurich: the Swiss Data Science Center. The Center’s mission is to accelerate the use of data science and machine learning techniques within academic disciplines of the ETH Domain, the Swiss academic community at large, and the industrial sector. A multi-disciplinary team of senior data scientists and experts in domains such as personalized health and medicine, earth and environmental science, social science and digital humanities, as well as economics enables collaboration on both academic and industrial projects. This unique positioning, at the crossroad of academic excellence and fast-paced business environments agility is key in making the complex data science journey simple.

 

Our concept

 

In this novel introductory machine learning course we take you skiing. And we are serious about the skiing – so serious that we make it an integral part of the course. When hitting the slopes, you will be collecting data through three lightweight IMU sensors. During extensive hands-on sessions you will perform data exploration, build regression models or identify clusters, all based on your own sensor data. This is not only more motivating than analyzing yet another overly curated example dataset, it is also much closer to the real-world problems that data scientists are facing every day.

Our concept

 

In this novel introductory machine learning course we take you skiing. And we are serious about the skiing – so serious that we make it an integral part of the course. When hitting the slopes, you will be collecting data through three lightweight IMU sensors. During extensive hands-on sessions you will perform data exploration, build regression models or identify clusters, all based on your own sensor data. This is not only more motivating than analyzing yet another overly curated example dataset, it is also much closer to the real-world problems that data scientists are facing every day.

Why?

The importance of data science and artificial intelligence (AI) technologies and their potential impact on economy and society is nowadays widely recognized. To fully grasp the digitalization opportunity, and sometimes purely to survive, companies have to evolve and adapt at a pace they have never experienced before.

In order to navigate today’s digital age, many traditional corporations welcome a new type of experts: data scientists.

These skilled specialists are highly sought-after professionals and access to them is critical to an ever-growing part of the Swiss industry. In order to retain them and allow them to develop, companies need to offer their data experts every opportunity to learn and grow. However, these opportunities are sometimes limited within a corporation that is still at the dawn of its digital transformation journey.

Through participation in the recently established MSc programs of both EPFL and ETH Zürich, the Swiss Data Science Center (SDSC) is playing a vital role in educating tomorrows’ data scientists for Switzerland. Of equal importance to the SDSC is the continued education of professionals in selected topics related to data science. This is why the center is involved in DAS and CAS certificate programs offered by the EPFL and ETH Zürich.

However, not everyone can dedicate several months to continuous learning.

For this reason, we have created “Carving through Data”, a one-week introductory course to Machine Learning for professionals with backgrounds in quantitative disciplines such as mathematics, engineering or natural sciences.

Why?

The importance of data science and artificial intelligence (AI) technologies and their potential impact on economy and society is nowadays widely recognized. To fully grasp the digitalization opportunity, and sometimes purely to survive, companies have to evolve and adapt at a pace they have never experienced before.

In order to navigate today’s digital age, many traditional corporations welcome a new type of experts: data scientists.

These skilled specialists are highly sought-after professionals and access to them is critical to an ever-growing part of the Swiss industry. In order to retain them and allow them to develop, companies need to offer their data experts every opportunity to learn and grow. However, these opportunities are sometimes limited within a corporation that is still at the dawn of its digital transformation journey.

Through participation in the recently established MSc programs of both EPFL and ETH Zürich, the Swiss Data Science Center (SDSC) is playing a vital role in educating tomorrows’ data scientists for Switzerland. Of equal importance to the SDSC is the continued education of professionals in selected topics related to data science. This is why the center is involved in DAS and CAS certificate programs offered by the EPFL and ETH Zürich.

However, not everyone can dedicate several months to continuous learning.

For this reason, we have created “Carving through Data”, a one-week introductory course to Machine Learning for professionals with backgrounds in quantitative disciplines such as mathematics, engineering or natural sciences.

Testimonials

A great experience

"The workshop was a very fulfilling experience. Besides the technical knowledge throughout the workshop, it was set in a great location which offered relaxing time between sessions for one to one discussions with the tutors."

Achilleas Xydis
PhD Researcher, Gramazio Kohler Research at ETH Zurich

Fun & Intense

"Carving Through Data was a lot of fun but also quite intense. The tutors did a great job explaining the fundamental concepts behind Machine Learning rather than just showing us how to apply some packages as black-boxes."

Alem Filli
Data Engineer at Julius Bär

In summary

Topic Introductory course to machine learning
Date March 9 – 13
(5 course days)
Participants The number of participants is limited to 20.

In summary

Topic Introductory course to machine learning
Date March 9 – 13
(5 course days)
Participants The number of participants is limited to 20.

 

Keynote Talk

Alhussein Fawzi

Research scientist at Google DeepMind, London (UK)

“Recent Advances in AI”

Alhussein Fawzi is a research scientist at Google DeepMind in London, working on making machine learning systems more robust. He received his M.Sc. and PhD degrees from the Swiss Federal Institute of Technology (EPFL), Switzerland, and spent one year as a postdoctoral scholar in the Computer Science Department at UCLA. He received twice the IBM PhD fellowship. More information can be found in his website: http://www.alhusseinfawzi.info/

Keynote Talk

Alhussein Fawzi

Research scientist at Google DeepMind, London (UK)

“Recent Advances in AI”

Alhussein Fawzi is a research scientist at Google DeepMind in London, working on making machine learning systems more robust. He received his M.Sc. and PhD degrees from the Swiss Federal Institute of Technology (EPFL), Switzerland, and spent one year as a postdoctoral scholar in the Computer Science Department at UCLA. He received twice the IBM PhD fellowship. More information can be found in his website: http://www.alhusseinfawzi.info/

Tuition fees

early registration*

regular registration

1800.- CHF 2000.- CHF

Tuition fees

early registration*

regular registration

1800.- CHF 2000.- CHF

Additional costs :

Hotel costs

Price CHF

4 days

5 days (INCLUDING SUNDAY NIGHT)

Room

520.- 650.-

Half board

140.- 175.-

Touristtax

10.- 12.50

Total 

670.- 837.50

With 10 % early registration discount*

604.- 755.-

Early bird rate applies until October 31, 2019

Hotel costs

Price CHF

4 days

5 days (INCLUDING SUNDAY NIGHT)

Room

520.- 650.-

Half board

140.- 175.-

Touristtax

10.- 12.50

Total 

670.- 837.50

With 10 % early registration discount*

604.- 755.-

Early bird rate applies until October 31, 2019

Total cost

early registration*

regular registration

4 Days 2404.- CHF 2670.- CHF
5 Days 2555.- CHF 2837.50 CHF
+ Skipass (optional) +213.- CHF +213.- CHF

Total Cost

early registration*

regular registration

4 days

2404.- CHF 2670.- CHF

5 days

2555.- CHF 2837.50 CHF

+ Skipass (optional)

+213.- CHF +213.- CHF

Early bird rate applies until October 31, 2019

Prerequisites

 –     Basic experience in Python programming.
  • Data types: lists, arrays, dictionaries
  • Control structures: if/else statements, for/while loops
  • Defining and calling functions
While some Python knowledge is recommended, people who are very experienced in other languages are usually able to learn the necessary Python skills on the go.
–     Mathematic
  • Linear algebra: matrix-vector multiplication, matrix eigen-decomposition, vector norms
  • Calculus/analysis : derivative/gradient of a function, minimum, macimum, inflexion point.
  • Statistics and probability: Bayes rule, probability function, density function.
Remark: While we will try our best to help people with individual gaps in some of the above topics, being unfamiliar with the majority of them will make it very hard to follow the course.
–     Skiing is not required. Data will be provided for non-skiers.

Note that skiing participants are required to bring their own equipment or rent at a local sport shop. While the course is built on data collected while skiing, non-skiers can also join the course (data will be provided to them). No ski tuition will be provided as part of the course. Please contact us should you have any question.

Prerequisites

 

 –     Basic experience in Python programming.

      • Data types: lists, arrays, dictionaries
      • Control structures: if/else statements, for/while loops
      • Defining and calling functions

While some Python knowledge is recommended, people who are very experienced in other languages are usually able to learn the necessary Python skills on the go.

 

–     Mathematics

      • Linear algebra: matrix-vector multiplication, matrix eigen-decomposition, vector norms
      • Calculus/analysis : derivative/gradient of a function, minimum, maximum, inflexion point.
      • Statistics and probability: Bayes rule, probability function, density function.

While we will try our best to help people with individual gaps in some of the above topics, being unfamiliar with the majority of them will make it very hard to follow the course.

 

–     Skiing is not required. Data will be provided for non-skiers.

Topics covered by the course

 Here is an indicative list of the topic covered by the course. Some of them will only be broadly discussed:

  • Supervised learning: regression, classification, overfitting, cross-validation
  • Basic of optimization: gradient descent, backpropagation, stochastic gradient descent
  • Unsupervised learning: dimensionality reduction clustering
  • Algorithms: SVM, Random Forest. K-MEANS, PCA
  • Deep learning: Neural Network (NN), Fully connected NN, Convolutional NN, traditional NN tricks

Venue

The course will be held in Fiescheralp in the canton of Valais. Course participants reside in the Alpina Hotel.

images credits:
Alpina Hotel

Venue

The course will be held in Laax in the canton of Graubünden. Course participants reside in the Signina Hotel where a conference room for evening hacking sessions is available.

images credits:
www.signinahotel.com

Program

Monday

09.30:

Beginning of the program

Course A

Introduction do basic data science software tools. (Jupyter, Numpy, Pandas, Sckit-learn, Git)

Course B

Data exploration and plotting

Practical work:

understanding and exploring a data pipeline, data extraction and plotting


SDSC – speaker:

Dr. Andreas Bleuler, Sr. Computer Scientist

Tuesday

Course C

Supervised learning 1. (Linear regression, basic optimization)

Course D

Dataset preparation. (Overfitting, dataset splitting)

Hands-on session

Implementation of a regression algorithm to predict the skier speed from other variables


SDSC – Tutor:

Dr. Luis Salamanca, Sr. Data Scientist

Wednesday

Course E

Supervised learning 2. (Classification)

Course F

Un-supervised learning. (Clustering, k-means, basic embedding, PCA)

Hands-on session

Group skier with similar ski styles using unsupervised learning (k-means). Create an embedding of the participants ski data to extract their differences in ski styles


SDSC – Tutor:

Dr. Michele Volpi, Sr. Data Scientist

Thursday

Course G

Introduction to deep learning. (What is deep learning? Why does it work? When should it be used? What are its limitations?)

Course H

Convolutional neural networks

Hands-on session

Learn features and improve the previous regression/classification tasks using convolutional neural networks


SDSC – Speaker:

Dr. Nathanaël Perraudin, Sr. Data Scientist

Friday

Backup course session or one advanced ML topic such as (Generative Adversarial Network, Reinforcement Learning, Graph neural network, Natural language processing).

Hands-on session

backup session or challenge

Keynote speaker:

An overview of current research directions in ML.

Sunday (optional)

Optional :

Join us one day earlier to collect a large bunch of your own data using the snow-cookie sensors on the slopes of Fiescheralp before the actual course starts.

09.30:

Beginning of the program

Course A

Introduction do basic data science software tools. (Jupyter, Numpy, Pandas, Sckit-learn, Git)

Course B

Data exploration and plotting

Practical work:

understanding and exploring a data pipeline, data extraction and plotting


SDSC – speaker:

Dr. Andreas Bleuler, Sr. Computer Scientist

Course C

Supervised learning 1. (Linear regression, basic optimization)

Course D

Dataset preparation. (Overfitting, dataset splitting)

Hands-on session

Implementation of a regression algorithm to predict the skier speed from other variables


SDSC – Tutor:

Dr. Luis Salamanca, Sr. Data Scientist

Course E

Supervised learning 2. (Classification)

Course F

Un-supervised learning. (Clustering, k-means, basic embedding, PCA)

Hands-on session

Group skier with similar ski styles using unsupervised learning (k-means). Create an embedding of the participants ski data to extract their differences in ski styles


SDSC – Tutor:

Dr. Michele Volpi, Sr. Data Scientist

Course G

Introduction to deep learning. (What is deep learning? Why does it work? When should it be used? What are its limitations?)

Course H

Convolutional neural networks

Hands-on session

Learn features and improve the previous regression/classification tasks using convolutional neural networks


SDSC – Speaker:

Dr. Nathanaël Perraudin, Sr. Data Scientist

Backup course session or one advanced ML topic such as (Generative Adversarial Network, Reinforcement Learning, Graph neural network, Natural language processing).

Hands-on session

backup session or challenge

Keynote speaker:

An overview of current research directions in ML.

Typical schedule

(7.5 + hours of tuition per day)

08:00 -09:45 Course A
09:45 – 10:00 Coffee Break
10:00 – 11:45 Course B
11:45 – 15:00 Lunch break (on the resort)
15:00 – 19:00 Hands-on session and data acquisition (ski)
19:30 – 20:30 Dinner
21:00 Open end: Hacking/questions in the lobby

Program

 

09.30:

Beginning of the program

Course A

Introduction do basic data science software tools. (Jupyter, Numpy, Pandas, Sckit-learn, Git)

Course B

Data exploration and plotting.

Practical work:

understanding and exploring a data pipeline, data extraction and plotting


SDSC – speaker:

Dr. Andreas Bleuler, Sr. Computer Scientist

Course C

Supervised learning 1. (Linear regression, basic optimization)

Course D

Dataset preparation. (Overfitting, dataset splitting)

Hands-on session

Implementation of a regression algorithm to predict the skier speed from other variables


SDSC – Tutor:

Dr. Luis Salamanca, Sr. Data Scientist

Course E

Supervised learning 2. (Classification)

Course F

Un-supervised learning. (Clustering, k-means, basic embedding, PCA)

Hands-on session

Group skier with similar ski styles using unsupervised learning (k-means). Create an embedding of the participants ski data to extract their differences in ski styles


SDSC – Tutor:

Dr. Michele Volpi, Sr. Data Scientist

Course G

Introduction to deep learning. (What is deep learning? Why does it work? When should it be used? What are its limitations?)

Course H

Convolutional neural networks

Hands-on session

Learn features and improve the previous regression/classification tasks using convolutional neural networks


SDSC – Tutor:

Dr. Nathanaël Perraudin, Sr. Data Scientist

Backup course session or one advanced ML topic such as (Generative Adversarial Network, Reinforcement Learning, Graph neural network, Natural language processing).

Hands-on session

backup session or challenge

Keynote speaker:

An overview of current research directions in ML.

Optional

Join us one day earlier to collect a large bunch of your own data using the snow-cookie sensors on the slopes of Fiescheralp before the actual course starts.

Typical schedule

(7.5 + hours of tuition per day)

08:00 -09:45 Course A
09:45 – 10:00 Coffee Break
10:00 – 11:45 Course B
11:45 – 15:00 Lunch break (on the resort)
15:00 – 19:00 Hands-on session and data acquisition (ski)
19:30 – 20:30 Dinner
21:00 Open end: Hacking/questions in the lobby

Sensors

Sensors

The sensors used (named Snowcookies) are produced in Switzerland by a Swiss-Polish start-up (http://snowcookiesports.com). Each sensor measures acceleration, angular velocity and orientation. The sensors are connected to a mobile phone which records the data. Three sensors per skier (left ski, right ski, chest) together with GPS and air pressure data measured by the phone allow for a detailed analysis of the skier’s environment and technique.

Contact us :


About us

We accompany the academic community and the industrial sector in their data science journey, putting to work AI and ML and facilitating the multidisciplinary exchange of data and knowledge…

Visit the Swiss Data Science Center website