Plenary lectures

Accepted papers

Following papers have been accepted for Journal Track of our conference:

Latebreaking track talks

Following talks have been accepted for Latebreaking track:

About

Conference on Theoretical Foundations of Machine Learning (TFML 2017) will take place in Kraków, Poland, on February 13-17, 2017. We invite submissions of papers addressing theoretical aspects of machine learning and related topics. We strongly support a broad definition of learning theory, including, but not limited to:

• Theoretical foundations of
  • supervised learning
  • unsupervised learning
  • semi-supervised learning
  • deep learning
  • active learning
  • concept drift
  • natural language processing
  • neuroinformatics
• cheminformatics
  • drug design
  • activity prediction
  • representation learning
• generalization bounds
• optimization methods
• computational complexity of learning
• information theoretic learning

Conference is organized by Department of Machine Learning, Institute of Computer Science and Computational Mathematics, Faculty of Mathematics and Computer Science, Jagiellonian University.

Conference materials

Proceedings

Place

The conference will take place at Jagiellonian University, at the Faculty of Mathematics and Computer Science, Cracow, Łojasiewicza 6, room 1016.

Directions

• from Kraków Airport
• from train station
• to gala dinner

Key dates

• paper submission - 30 Nov 11 Dec 2016
• decision notification - 5 Jan 2017
• camera ready version - 11 17 Jan 2017
• registration 5-19 Jan 2017
• conference - 13-17 Feb 2017

Registration

In order to register for TFML conference you have to:

• Fill in the registration form
• Pay the registration fee

The registration fee of 600 PLN (150 euro) should be transfered to: As a title of the transfer use the following template.

Fees

Registration fee of 600 PLN (150 euro) covers:

• presentation slot
• publication of one paper
• dinners
• coffee breaks
• gala dinner
• proceedings

There is possibility of discounts for phd students.

Call for papers

Original, unpublished research papers can be submitted through easychair system. They will be peer reviewed and after acceptance and presentation during the TFML conference published in Schedae Informaticae Journal.

Submission format

Papers should be prepared in English, according to the Schedae Informaticae template. Page limit (including references, and other additional materials) is 8 pages (soft limit) and a hard limit of 10 pages. Papers longer than 10 pages will be rejected without reviewing. The final paper accepted for the conference should be no longer than 20 pages.

Double-blind review

Papers will be selected via a double-blind peer-review process (the reviewers will not know the identities of the authors, and vice versa). It will be up to the authors to ensure the proper anonymization of their paper. In particular, the first page cannot include names or affiliations of the authors. When referring to one's own work, use the third person, rather than the first person. Violation of the above rules may lead to rejection without review.

Submit your paper

EasyChair submission system is now open. submit your paper

Publication

All papers will be peer-reviewed. All accepted papers will be published in Schedae Informaticae Journal¹. At least one author of each publication has to attend the conference and present it during oral/poster session.

¹For Polish scientists, this journal is located on the B list of the Polish Ministry of Science and Higher Education, with 11pts.
www.ejournals.eu/Schedae-Informaticae/

Schedule

Each plenary lecture has 60 min + 10 min discussion time slot.

Plenary lectures

1. K. Dembczyński "Extreme Zero-Shot Learning"
2. G. Chechik "Large-scale machine learning for understanding images"
3. K. Geras "Recent advances in model compression"
4. B. Cyganek "Multi-dimensional Data Classification – Theory and Applications"
5. A. Storkey "Rise of the Adversary"
6. E. Hüllermeier "Superset learning and data imprecisiation"
7. M. Kloft "Extreme Classification: Tighter Bounds, Distributed Training, and new Algorithms"
8. P. Kolb "In silico searches for small-molecule ligands in chemical space"

Each presentation at oral session has a 25 min + 5 min discussion time slot.

Session 1

1. Split-and-merge Tweak in Cross Entropy Clustering
2. Data Selection For Neural Networks
3. Portfolio inputs selection from imprecise training data

Session 2

1. Towards Learning Word Representation
2. Impact of clustering parameters on the efficiency of knowledge mining process in rule-based knowledge bases
3. Word embeddings for morphologically complex languages
4. Online supervised learning approach for machine scheduling

Session 3

1. Misclassification-Driven Sample Relabeling for Supervised Kernel Principal Component Analysis
2. Pairwise versus Pointwise Ranking: A Case Study
3. Regression SVM for incomplete data

Session 4

1. Adverse Event Detection in Medical Safety Reports at Roche
2. Optimization of l^p-regularized Linear Models via Coordinate Descent
3. Logistic regression classifier with pairwise constraints

Session 5

1. On Loss Functions for Deep Networks in Classification
2. On certain limitations of Recursive Representation model
3. Extreme classification under limited space and time budget

Session 6

1. Models for personalized recommendations in Brainly

Session 7

1. Uniform Cross-entropy clustering
2. Molecular shape descriptors
3. Search for resolution invariant wavelet features of melanoma learned by a limited ANN classifier
4. A Translation Evaluation Function based on Neural Network

Co-organisers

Sponsorship