Geoffrey Hinton’s Unsupervised Capsule Networks Achieve SOTA Results on SVHN
https://medium.com/syncedreview/geoffrey-hintons-unsupervised-capsule-networks-achieve-sota-results-on-svhn-ffe05e871249
https://medium.com/syncedreview/geoffrey-hintons-unsupervised-capsule-networks-achieve-sota-results-on-svhn-ffe05e871249
Medium
Geoffrey Hinton’s Unsupervised Capsule Networks Achieve SOTA Results on SVHN
In 2017 the “Godfather of Deep Learning” Geoffrey Hinton and his students Sara Sabour and Nicholas Frosst proposed the discrimininatively…
35 PhD and Postdoctoral Positions at Lund University, Sweden
https://scholaridea.com/2019/06/26/35-phd-and-postdoctoral-positions-at-lund-university-sweden/
https://scholaridea.com/2019/06/26/35-phd-and-postdoctoral-positions-at-lund-university-sweden/
Scholar Idea
35 PhD and Postdoctoral Positions at Lund University, Sweden
PhD and Postdoctoral Positions at Lund University, one of northern Europe’s oldest universities in Sweden Doctoral Student in Medical Structural Biology MSCA COFUND project PRISMAS (PA2023/508)Med structural biology, Faculty of MedicineDate published: 1 Sep…
Sequential Neural Processes. arxiv.org/abs/1906.10264
Object Tracking Tech from CVPR 2019!
The method, dubbed SiamMask, improves the offline training procedure of popular fully-convolutional Siamese approaches for object tracking by augmenting their loss with a binary segmentation task.
https://www.profillic.com/paper/arxiv:1812.05050
The method, dubbed SiamMask, improves the offline training procedure of popular fully-convolutional Siamese approaches for object tracking by augmenting their loss with a binary segmentation task.
https://www.profillic.com/paper/arxiv:1812.05050
Profillic
Profillic: AI research & source code to supercharge your projects
Explore state-of-the-art in machine learning, AI, and robotics research. Browse papers, source code, models, and more by topics and authors. Connect with researchers and engineers working on related problems in machine learning, deep learning, natural language…
RVOS: End-to-End Recurrent Network for Video Object Segmentation
CVPR 2019
https://imatge-upc.github.io/rvos/
CVPR 2019
https://imatge-upc.github.io/rvos/
imatge-upc.github.io
RVOS: End-to-End Recurrent Network for Video Object Segmentation
Description here
Meta Reinforcement Learning
https://lilianweng.github.io/lil-log/2019/06/23/meta-reinforcement-learning.html
https://lilianweng.github.io/lil-log/2019/06/23/meta-reinforcement-learning.html
Lil'Log
Meta Reinforcement Learning
Meta-RL is meta-learning on reinforcement learning tasks. After trained over a distribution of tasks, the agent is able to solve a new task by developing a new RL algorithm with its internal activity dynamics. This post starts with the origin of meta-RL and…
"How to Understand the Universe When You’re Stuck Inside of It"
By Amanda Gefter: https://www.quantamagazine.org/were-stuck-inside-the-universe-lee-smolin-has-an-idea-for-how-to-study-it-anyway-20190627/
#Physics #Universe
By Amanda Gefter: https://www.quantamagazine.org/were-stuck-inside-the-universe-lee-smolin-has-an-idea-for-how-to-study-it-anyway-20190627/
#Physics #Universe
Quanta Magazine
How to Understand the Universe When You’re Stuck Inside of It
Lee Smolin’s radical idea to reimagine how we view the universe.
Generating Diverse High-Fidelity Images with VQ-VAE-2
paper https://arxiv.org/pdf/1906.00446.pdf
paper https://arxiv.org/pdf/1906.00446.pdf
What Does BERT Look At? An Analysis of BERT's Attention
Clark et al.: https://arxiv.org/abs/1906.04341
Code: https://github.com/clarkkev/attention-analysis
#bert #naturallanguage #unsupervisedlearning
Clark et al.: https://arxiv.org/abs/1906.04341
Code: https://github.com/clarkkev/attention-analysis
#bert #naturallanguage #unsupervisedlearning
Generating Diverse High-Fidelity Images with VQ-VAE-2
paper https://arxiv.org/pdf/1906.00446.pdf
blog https://machinethoughts.wordpress.com/2019/06/25/the-inevitability-of-vector-quantization-in-deep-architectures/
paper https://arxiv.org/pdf/1906.00446.pdf
blog https://machinethoughts.wordpress.com/2019/06/25/the-inevitability-of-vector-quantization-in-deep-architectures/
Machine Thoughts
The Inevitability of Vector Quantization in Deep Architectures
I just read VQ-VAE-2 (Vector-Quantized – Variational AutoEncoders – 2) by Razavi et al. This paper gives very nice results on modeling image distributions with vector quantization. It…
Bridging the Domain Gap for Neural Models
https://machinelearning.apple.com/2019/06/15/bridging-the-domain-gap-for-neural-models.html
https://machinelearning.apple.com/2019/06/15/bridging-the-domain-gap-for-neural-models.html
Apple Machine Learning Research
Bridging the Domain Gap for Neural Models
Deep neural networks are a milestone technique in the advancement of modern machine perception systems. However, in spite of the exceptional…
NeurIPS 2019 Call for Post-Conference Workshops
Friday, December 13 and Saturday, December 14, 2019
Vancouver Convention Center, Vancouver, Canada
https://nips.cc/Conferences/2019/CallForWorkshops
Following the NeurIPS 2019 main conference, workshops on a variety of current topics will be held on Friday, December 13 and Saturday, December 14, 2019. We invite researchers interested in chairing one of these workshops to submit proposals. Workshop organizers have several responsibilities, including coordinating workshop participation and content, publicizing and providing the program in a timely manner, and moderating the program throughout the workshop.
With the rapid growth and interest in NeurIPS and its associated workshops, the competition for workshops has intensified. Guidance on workshop goals, evaluation criteria, the assessment process, and the submission format and procedure are available here: https://nips.cc/Conferences/2019/CallForWorkshops
Organizers of workshop proposals should take care to respect this guidance and to provide explicit answers to the questions implied throughout.
Key Dates:
Workshop Application Deadline: June 3, 2019 (6 p.m. Pacific Time)
Workshop Notification: July 23, 2019
Mandatory Cutoff for Workshop Organizers to Notify Participants of Accept/Reject Decisions: October 1, 2019
Workshops: December 13 and 14, 2019
Friday, December 13 and Saturday, December 14, 2019
Vancouver Convention Center, Vancouver, Canada
https://nips.cc/Conferences/2019/CallForWorkshops
Following the NeurIPS 2019 main conference, workshops on a variety of current topics will be held on Friday, December 13 and Saturday, December 14, 2019. We invite researchers interested in chairing one of these workshops to submit proposals. Workshop organizers have several responsibilities, including coordinating workshop participation and content, publicizing and providing the program in a timely manner, and moderating the program throughout the workshop.
With the rapid growth and interest in NeurIPS and its associated workshops, the competition for workshops has intensified. Guidance on workshop goals, evaluation criteria, the assessment process, and the submission format and procedure are available here: https://nips.cc/Conferences/2019/CallForWorkshops
Organizers of workshop proposals should take care to respect this guidance and to provide explicit answers to the questions implied throughout.
Key Dates:
Workshop Application Deadline: June 3, 2019 (6 p.m. Pacific Time)
Workshop Notification: July 23, 2019
Mandatory Cutoff for Workshop Organizers to Notify Participants of Accept/Reject Decisions: October 1, 2019
Workshops: December 13 and 14, 2019
nips.cc
NeurIPS Call for Workshops
NeurIPS Website
Research Fellow in Artificial Intelligence in Medical Imaging
Are you an early-career researcher who enjoys developing fundamental methods with impact in challenging problems in medical image computing? Do you have a strong background in computer science, statistics, mathematics or physics and want to apply it to medical image computing? Would you like to work with cardiologists, oncologists and endocrinologists and have access to massive clinical image databases? Do you have a passion for combining computational algorithms, modelling and simulation to address key problems in medicine? Are you ready to think out-of-the-box, innovate and find solutions to challenging problems?
https://jobs.leeds.ac.uk/Vacancy.aspx?id=15595&forced=2
Further details contact Prof A F Frangi [email protected], www.cistib.org/afrangi
Are you an early-career researcher who enjoys developing fundamental methods with impact in challenging problems in medical image computing? Do you have a strong background in computer science, statistics, mathematics or physics and want to apply it to medical image computing? Would you like to work with cardiologists, oncologists and endocrinologists and have access to massive clinical image databases? Do you have a passion for combining computational algorithms, modelling and simulation to address key problems in medicine? Are you ready to think out-of-the-box, innovate and find solutions to challenging problems?
https://jobs.leeds.ac.uk/Vacancy.aspx?id=15595&forced=2
Further details contact Prof A F Frangi [email protected], www.cistib.org/afrangi
Jobs at the University of Leeds
Job Opportunity at University of Leeds: Research Fellow in Artificial Intelligence in Medical Imaging
Are you an early-career researcher who enjoys developing fundamental methods with impact in challenging problems in medical image computing? Do you have a strong background in computer science, statistics, mathematics or physics and want to apply it...
Post-Doctoral Fellowship on Ethically Aligned Artificial Intelligence
Mila is looking for a postdoctoral fellow who would work on ethically aligned learning machines, or how an AI can acquire moral competence. The research topic is important in the long-term for AI safety, towards building machines which can achieve specific goals while acting in a way consistent with human values and social norms. The candidate should be able to perform experiments with data (e.g. to acquire or collect relevant data from humans) and train deep learning systems towards these goals. She or he must also be able to engage with some of the philosophical issues raised by artificial moral agents and the idea of moral competence for an AI system.
Priority will be given to candidates that have a substantial knowledge in machine learning, as well as in philosophy, moral psychology or the social sciences.
The position is for one year, with possibility of renewal, at Mila. The project will involve professors Bengio (University of Montreal) and Dominic Martin (UQAM) as well as other potential collaborators in machine learning and philosophy.
To apply, please submit requested documents at https://mila.quebec/en/admission/, and contact Professors Bengio (via his assistant at [email protected]) and Martin ([email protected]) by e-mail.
We will start processing applications as of September 16th 2019, and until the position is filled. Candidates are expected to start their fellowship during the Fall semester.
Mila is looking for a postdoctoral fellow who would work on ethically aligned learning machines, or how an AI can acquire moral competence. The research topic is important in the long-term for AI safety, towards building machines which can achieve specific goals while acting in a way consistent with human values and social norms. The candidate should be able to perform experiments with data (e.g. to acquire or collect relevant data from humans) and train deep learning systems towards these goals. She or he must also be able to engage with some of the philosophical issues raised by artificial moral agents and the idea of moral competence for an AI system.
Priority will be given to candidates that have a substantial knowledge in machine learning, as well as in philosophy, moral psychology or the social sciences.
The position is for one year, with possibility of renewal, at Mila. The project will involve professors Bengio (University of Montreal) and Dominic Martin (UQAM) as well as other potential collaborators in machine learning and philosophy.
To apply, please submit requested documents at https://mila.quebec/en/admission/, and contact Professors Bengio (via his assistant at [email protected]) and Martin ([email protected]) by e-mail.
We will start processing applications as of September 16th 2019, and until the position is filled. Candidates are expected to start their fellowship during the Fall semester.
Postdoctoral Position at Washington University (AI/ML for telemedicine)
We are seeking a full time Postdoctoral Research Associate / Scholar
to work on exciting projects at the intersection of machine learning
(ML), AI, and clinical decision support at Washington University in St
Louis. The candidate will work with massive real-time medical data
collected in a leading hospital, and will solve challenging problems
with significant impacts on next-general smart healthcare.
The Project
The goal of the TECTONICS (Telemedicine Control Tower for the OR:
Navigating Information, Care and Safety) project is to harness large
scale clinical data to improve the care of perioperative patients by
providing timely, explainable, and actionable ML predictions of
patient trajectories. The TECTONICS team is externally funded for a
randomized trial of integrated telemedicine and real-time risk
prediction for anesthesia providers.
TECTONICS is deeply interdisciplinary, including experts in ML, optimization, planning, big data mining, human-computer interaction, biostatistics, clinical informatics, and clinical trials. Currently the team consists of 6 full-time faculty members, 2 postdoctoral researchers, 3 PhD students, 1 informatics technicians, 1 clinical experts, and 3 staff members and we are seeking to hire 1-2 additional post-docs and research staff.
The candidate will take a leading role in refinement of existing methods
and development of novel techniques in ML for clinical prediction. The
candidate will be expected to pursue independent lines of research,
and should anticipate using one of the multiple clinical datasets
available to the TECTONICS team. The datasets contain detailed
electronic records and real-time operation data for hundreds of
thousands of patients. The candidate will co-supervise graduate
students and undergraduate research assistants, and be expected to
work closely with TECTONICS team members from other disciplines.
The Candidate
Successful candidates will hold an MD or PhD and have a strong
research record in machine learning and a strong interest in
developing practical systems for healthcare and medicine. Expertise
with time-series learning, interpretable and actionable learning,
online learning, dealing with imbalanced, noisy, and incomplete data,
knowledge embedding, and clinical data mining will be a plus. The
duties and responsibilities of this position include conducting
individual and collaborative research on the themes of the project,
developing novel algorithms and systems based on the needs of the
project, and producing high-quality outputs for publication in
high-profile journals or conference proceedings.
The position is initially for 2 years, with possible extension up to 5
years total. We offer a highly competitive benefits package including
training and development opportunities. Washington University in St.
Louis is one of the best places for this type of interdisciplinary
research on ML and AI in telemedicine and clinical decision-making.
The position is based in St. Louis, Missouri, USA where the university
is situated in a beautiful, leafy campus near the Forest Park and many
metropolitan activities. This is a full-time position, available from
July 2019 or as soon as possible thereafter.
To Apply
To apply, please send your CV to Maureen Arends at [email protected] .
Please mention the TECTONICS project in your email.
For any questions regarding this opportunity, please feel free to contact:
- Prof. Yixin Chen ([email protected]), Department of
Computer Science and Engineering
- Prof. Michael Avidan ([email protected]) , Department
of Anesthesiology
We are seeking a full time Postdoctoral Research Associate / Scholar
to work on exciting projects at the intersection of machine learning
(ML), AI, and clinical decision support at Washington University in St
Louis. The candidate will work with massive real-time medical data
collected in a leading hospital, and will solve challenging problems
with significant impacts on next-general smart healthcare.
The Project
The goal of the TECTONICS (Telemedicine Control Tower for the OR:
Navigating Information, Care and Safety) project is to harness large
scale clinical data to improve the care of perioperative patients by
providing timely, explainable, and actionable ML predictions of
patient trajectories. The TECTONICS team is externally funded for a
randomized trial of integrated telemedicine and real-time risk
prediction for anesthesia providers.
TECTONICS is deeply interdisciplinary, including experts in ML, optimization, planning, big data mining, human-computer interaction, biostatistics, clinical informatics, and clinical trials. Currently the team consists of 6 full-time faculty members, 2 postdoctoral researchers, 3 PhD students, 1 informatics technicians, 1 clinical experts, and 3 staff members and we are seeking to hire 1-2 additional post-docs and research staff.
The candidate will take a leading role in refinement of existing methods
and development of novel techniques in ML for clinical prediction. The
candidate will be expected to pursue independent lines of research,
and should anticipate using one of the multiple clinical datasets
available to the TECTONICS team. The datasets contain detailed
electronic records and real-time operation data for hundreds of
thousands of patients. The candidate will co-supervise graduate
students and undergraduate research assistants, and be expected to
work closely with TECTONICS team members from other disciplines.
The Candidate
Successful candidates will hold an MD or PhD and have a strong
research record in machine learning and a strong interest in
developing practical systems for healthcare and medicine. Expertise
with time-series learning, interpretable and actionable learning,
online learning, dealing with imbalanced, noisy, and incomplete data,
knowledge embedding, and clinical data mining will be a plus. The
duties and responsibilities of this position include conducting
individual and collaborative research on the themes of the project,
developing novel algorithms and systems based on the needs of the
project, and producing high-quality outputs for publication in
high-profile journals or conference proceedings.
The position is initially for 2 years, with possible extension up to 5
years total. We offer a highly competitive benefits package including
training and development opportunities. Washington University in St.
Louis is one of the best places for this type of interdisciplinary
research on ML and AI in telemedicine and clinical decision-making.
The position is based in St. Louis, Missouri, USA where the university
is situated in a beautiful, leafy campus near the Forest Park and many
metropolitan activities. This is a full-time position, available from
July 2019 or as soon as possible thereafter.
To Apply
To apply, please send your CV to Maureen Arends at [email protected] .
Please mention the TECTONICS project in your email.
For any questions regarding this opportunity, please feel free to contact:
- Prof. Yixin Chen ([email protected]), Department of
Computer Science and Engineering
- Prof. Michael Avidan ([email protected]) , Department
of Anesthesiology
PhD positions in machine learning at Max Planck Institute Tübingen
I am looking for prospective Ph.D. students to work with me at the Max Planck Institute for Intelligent Systems in Tübingen, Germany. In particular, I am looking for students who have a strong mathematical background. The programming skill is a plus but is not mandatory. Our research projects will be at the intersection of kernel methods, counterfactual inference, causal inference, and learning theory (see https://krikamol.org/ for more details on my research interest). Successful candidates will also have a chance to engage in international collaborative projects. Last but not least, the successful candidates should be able to demonstrate that their motivation for the Ph.D. application is driven by a desire to solve difficult real-world problems and to push the boundary of science.
*** How to Apply ***
The applicants should already hold or are soon expected to obtain a master or equivalent degree in related areas. The application package must be a single PDF file and include
1) Curriculum vitae (CV),
2) academic transcripts (only bachelor and master degrees),
3) up to two pages in A4 of research statement describing your research experience and your research interest,
4) Names, affiliations, and email addresses of at most 3 references who can provide recommendation letters for you. Note that we will contact them directly so please make sure that the email addresses are correct.
The application package must be sent directly to [email protected]. The application will be open until we find the right candidates.
If you are at ICML19 or CVPR19, I am happy to talk in person.
*** Max Planck Institute & Tübingen ***
Max Planck Institute for Intelligent Systems (https://www.is.mpg.de/) is a world-class center for foundational research in machine learning and related areas. It is located in Tübingen, Germany (https://www.tuebingen.de/en/). Tübingen is a scenic medieval university town, cradled in what is simultaneously one of Germany’s most beautiful landscapes and one of Europe’s most economically successful areas. Most locals speak English and knowledge of German is not required to live here. The city is also not far from the Alps which is home to hundreds of high-quality ski resorts.
Best,
Krikamol
I am looking for prospective Ph.D. students to work with me at the Max Planck Institute for Intelligent Systems in Tübingen, Germany. In particular, I am looking for students who have a strong mathematical background. The programming skill is a plus but is not mandatory. Our research projects will be at the intersection of kernel methods, counterfactual inference, causal inference, and learning theory (see https://krikamol.org/ for more details on my research interest). Successful candidates will also have a chance to engage in international collaborative projects. Last but not least, the successful candidates should be able to demonstrate that their motivation for the Ph.D. application is driven by a desire to solve difficult real-world problems and to push the boundary of science.
*** How to Apply ***
The applicants should already hold or are soon expected to obtain a master or equivalent degree in related areas. The application package must be a single PDF file and include
1) Curriculum vitae (CV),
2) academic transcripts (only bachelor and master degrees),
3) up to two pages in A4 of research statement describing your research experience and your research interest,
4) Names, affiliations, and email addresses of at most 3 references who can provide recommendation letters for you. Note that we will contact them directly so please make sure that the email addresses are correct.
The application package must be sent directly to [email protected]. The application will be open until we find the right candidates.
If you are at ICML19 or CVPR19, I am happy to talk in person.
*** Max Planck Institute & Tübingen ***
Max Planck Institute for Intelligent Systems (https://www.is.mpg.de/) is a world-class center for foundational research in machine learning and related areas. It is located in Tübingen, Germany (https://www.tuebingen.de/en/). Tübingen is a scenic medieval university town, cradled in what is simultaneously one of Germany’s most beautiful landscapes and one of Europe’s most economically successful areas. Most locals speak English and knowledge of German is not required to live here. The city is also not far from the Alps which is home to hundreds of high-quality ski resorts.
Best,
Krikamol
Max Planck Institute for Intelligent Systems
Home
Our goal is to understand the principles of <strong>Perception</strong>, <strong>Action</strong> and <strong>Learning</strong> in autonomous systems that successfully interact with complex environments and to use this understanding to design future artificially…
Berkeley using a new deep learning program to assess risk of suicide amongst veterans
Identifying patterns of risk within patients often involves a massive amount of data interpretation and algorithmic examination. New computer resources through Berkeley are today being dedicated to producing tailored algorithms for dynamic risk scores for VA patients and caregivers.
https://www.marktechpost.com/2019/04/19/berkeley-using-a-new-deep-learning-program-to-assess-risk-of-suicide-amongst-veterans/
Identifying patterns of risk within patients often involves a massive amount of data interpretation and algorithmic examination. New computer resources through Berkeley are today being dedicated to producing tailored algorithms for dynamic risk scores for VA patients and caregivers.
https://www.marktechpost.com/2019/04/19/berkeley-using-a-new-deep-learning-program-to-assess-risk-of-suicide-amongst-veterans/
MarkTechPost
Berkeley using a new deep learning program to assess risk of suicide amongst veterans
Berkeley using a new deep learning program to assess risk of suicide amongst veterans. massive amount of data interpretation and algorithmic examination.