An interesting new algorithm from DeepMind that aligns agent behaviour with a user's objectives in a reinforcement learning setting with unknown dynamics, an unknown reward function and unknown unsafe states.

Paper: https://arxiv.org/abs/1912.05652
Code: https://github.com/rddy/ReQueST

https://medium.com/@xamat/the-year-in-ai-2019-ml-ai-advances-recap-c6cc1d902d5 https://t.iss.one/ArtificialIntelligenceArticles

Generative Teaching Networks: Accelerating Neural Architecture Search by Learning to Generate Synthetic Training Data
Such et al.: https://arxiv.org/abs/1912.07768
#ArtificialIntelligence #MachineLearning #NeuralNetworks

Network of Evolvable Neural Units: Evolving to Learn at a Synaptic Level
Paul Bertens, Seong-Whan Lee : https://arxiv.org/abs/1912.07589
#NeuralComputing #MachineLearning #ArtificialIntelligence

PolSF: PolSAR image dataset on San Francisco
Github: https://github.com/liuxuvip/PolSF
Paper: https://arxiv.org/pdf/1912.07259v1.pdf

A Deep Neural Network's Loss Surface Contains Every Low-dimensional Pattern
Czarnecki et al.: https://arxiv.org/abs/1912.07559
#ArtificialIntelligence #DeepLearning #MachineLearning

PolSF: PolSAR image dataset on San Francisco
Github: https://github.com/liuxuvip/PolSF
Paper: https://arxiv.org/pdf/1912.07259v1.pdf

Love science research? Here's an interesting new service from IBM that produces summaries of Science papers based on information expressed through natural language queries.

White paper: https://arxiv.org/abs/1908.11152
Product: https://dimsum.eu-gb.containers.appdomain.cloud/

Searching for resources to put your deep learning model into production? Check this up: https://github.com/ahkarami/Deep-Learning-in-Production

It includes a comprehensive list of tutorials on :
a) How to convert PyTorch Models in Production
b) How to convert PyTorch Models to C++
c) How to deploy TensorFlow Models in Production
d) How to convert Keras Models in Production
e) How to deploy MXNet Models in Production

With additional tutorial list on:
a) Model Conversion between Deep Learning Frameworks
b) Caffe2
c) Resources for Designing UI (Front-End Development)
d) Mobile app Development
e) Back-End Development Part
f) GPU Management Libraries
g) Speed-up & Scalabale Python Codes

Keep Learning!!

Best of arXiv.org for AI, Machine Learning, and Deep Learning – November 2019
https://insidebigdata.com/2019/12/18/best-of-arxiv-org-for-ai-machine-learning-and-deep-learning-november-2019/

The Neuroscience of Consciousness - Interview with Christof Koch, Chief Scientist of The Allen Institute for Brain Science
https://www.youtube.com/watch?v=NsvEKyzRXdY&feature=youtu.be

From deep learning to mechanistic understanding in neuroscience: the structure of retinal prediction

https://papers.nips.cc/paper/9060-from-deep-learning-to-mechanistic-understanding-in-neuroscience-the-structure-of-retinal-prediction


#DeepLearning #Neuroscience #NeurIPS2019

Interrogating theoretical models of neural computation with deep inference
Bittner et al.: https://www.biorxiv.org/content/10.1101/837567v2
#Neuroscience

What is adversarial machine learning, and how is it used today?

-Generative modeling, security, model-based optimization, neuroscience, fairness, and more!

Here's a fantastic video overview by Ian Goodfellow.
https://videos.re-work.co/videos/1351-ian-goodfellow
#ML #adversarialML #AI #datascience

A great summary of all the invited talks, presentations, posters and workshops at this years Neural Information Processing Systems Conference (NeurIPS2019) happened in Vancouver last week. Thank you Chip Huyen https://huyenchip.com/2019/12/18/key-trends-neurips-2019.html
https://t.iss.one/ArtificialIntelligenceArticles

Kolter’s Team Wins First Place on Kaggle Competition with Over 2700 Teams
https://www.ml.cmu.edu/news/news-archive/2019/december/kolters-team-wins-machine-learning-kaggle-competition.html

Introduction to Artificial Intelligence
Fall 2019, Prof. Gilles Louppe : https://glouppe.github.io/info8006-introduction-to-ai/pdf/lec-all.pdf
#ArtificialIntelligence #DeepLearning #MachineLearning

Professor Karl Friston - Frontiers publications.
https://loop.frontiersin.org/people/20407/overview

Multiple PhD positions on machine learning with simulation and physics modeling of the world
Frantzeska Lavda [[email protected]]
Sent: 19 December 2019 14:11
To:
lhc-machinelearning-wg (Discussion group for machine learning at the LHC)
We have several PhD openings in machine learning research for exploring methods to combine learning with process-driven modeling and simulations.
The successful candidate will enroll as a PhD student in the Computer Science department of the University of Geneva (under the co-direction of myself and Prof. Stephane Marchand-Maillet) and, at the same time, will become a member of the Data Mining and Machine Learning group (https://dmml.ch) as a research and teaching assistant at HES-SO, Geneva. The positions shall be filled in as soon as possible.
The interaction and cooperation between a simulator and a machine learning model can be exploited in a number of areas where data are expensive or difficult to obtain, and/or where domain knowledge within the process-driven models can back the inductive biases factored into the machine learning models.
In the medical domain, machine learning methods can be combined with neuromechanical simulators to develop models of human locomotion that shall support critical medical decisions related to surgical interventions treating pathological gait patterns. In industrial manufacturing, simulations and physical modeling of realistic or extreme operational conditions can support the learning of rare faulty behaviours in order to trigger early alerts. In chemoinformatics, an external system (e.g. RDKit) can provide relevant constraints for generating valid new molecules with specific required characteristics.
Related literature:
- Battaglia, Peter, et al. "Interaction networks for learning about objects, relations and physics." Advances in neural information processing systems. 2016.
- Lionel Blondé, Alexandros Kalousis "Sample-Efficient Imitation Learning via Generative Adversarial Nets." AISTATS 2019: 3138-3148
- Narayanaswamy, Siddharth, et al. "Learning disentangled representations with semi-supervised deep generative models." Advances in Neural Information Processing Systems. 2017.
We seek strongly motivated candidates prepared to dedicate to high quality research in the above domains for a number of years (the expected time to PhD graduation is 4-5 years). The candidate should have (or be close to obtaining) a Master's degree or equivalent in computer science, statistics, applied mathematics, electrical engineering or other related field with strong background in as many as possible (but at least some) of these: machine learning, probability and statistical modeling, mathematical optimization, programming and software development (preferably Pytorch and/or Tensorflow).

If interested, please send the following to [email protected]
- academic CV (max 2 pages)
- academic transcript of the study results
- one page motivation letter explaining why the candidate is suitable for the position
- 500 word research proposal on one of the topics described above
- contact details of three referees (do not send reference letters)
The applications will be processed as they come as of now until the positions are filled. The status of the openings will be update here: https://dmml.ch/recruitment/

In case of any further questions, please contact [email protected].

Generative Teaching Networks: Accelerating Neural Architecture Search by Learning to Generate Synthetic Training Data
https://eng.uber.com/generative-teaching-networks/