MidiMe: Personalizing MusicVAE
Dinculescu et al.: https://magenta.tensorflow.org/midi-me
#ArtificialIntelligence #DeepLearning #Transformer

Sparse Networks from Scratch: Faster Training without Losing Performance

Tim Dettmers and Luke Zettlemoyer : https://arxiv.org/abs/1907.04840

#ArtificialIntelligence #MachineLearning #NeuralComputing

"Deep Image Prior": super-resolution, inpainting, denoising without learning on a dataset and pretrained networks. Comparable results to learned methods.
code https://github.com/DmitryUlyanov/deep-image-prior
project page https://dmitryulyanov.github.io/deep_image_prior

"Hamiltonian Neural Networks"
Greydanus et al.: https://arxiv.org/abs/1906.01563
Blog: https://greydanus.github.io/2019/05/15/hamiltonian-nns/
#Hamiltonian #NeuralNetworks #UnsupervisedLearning

Synthesizing Programs for Images using Reinforced Adversarial Learning
Ganin et al., 2018: https://proceedings.mlr.press/v80/ganin18a.html
Agents and environments : https://github.com/deepmind/spiral
#ArtificialIntelligence #DeepLearning #ReinforcementLearning

NeuPDE: Neural Network Based Ordinary and Partial Differential Equations for Modeling Time-Dependent Data

https://arxiv.org/pdf/1908.03190.pdf

Self-supervised Attention Model for Weakly Labeled Audio Event Classification arxiv.org/abs/1908.02876

Location Field Descriptors: Single Image 3D Model Retrieval in the Wild. arxiv.org/abs/1908.02853

Solar image denoising with convolutional neural networks. arxiv.org/abs/1908.02815

Attend To Count: Crowd Counting with Adaptive Capacity Multi-scale CNNs. arxiv.org/abs/1908.02797

Ten quick tips for effective dimensionality reduction

Lan Huong Nguyen and Susan Holmes : https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1006907

#ArtificialIntelligence #BigData #DataScience

Ten simple rules for writing and sharing computational analyses in Jupyter Notebooks

Rule et al.: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007007

#Jupyter #Notebooks #JupyterNotebooks

Book title: Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again by Eric Topol (Author)


Same author’s third book about the future of medicine

Author is a world-renowned cardiologist, Executive Vice-President of Scripps Research, founder of a new medical school and one of the top ten most cited medical researchers
He pointed out the transformational potential of AI for medicine , .is set to save time, lives and money.

a rough summary:
Dr. Topol tells us we are living in the Fourth Industrial Age, through AI ; AI has been sneaking into our lives;

The promise is to provide composite views of patients medical data; to improve decision support; to avoid error such as misdiagnosis and unnecesary procedures; to help in the ordering and interpretation of appropriate tests; and to recommend treatments;

For medicine, big datasets take the form of whole-genome sequences, high-res images, and continuous outputs from wearable sensors.

3D Medicine: digitizing, and democratizing, and deep ;

*Propose three components of deep medicine: (1) deep phenotyping; (2) deep learning based pattern learning; (3) deep empathy and connection between doctors/health systems and patients. *

current medicine practice is Shallow medicine, indicated by e.g., physicians spending the majority time looking for information and only ~20 percent of time in talking with patients; doctors are overloaded with many duties not about caring patients at all; computer, keyboards screens, scans et al are pushing doctors away from close relationships with patients; Besides, current healthcare “is resulting in extraordinary waste, suboptimal outcomes, and unnecessary harm.”

Dr. Topol reviews the states of the art: AI is pushing progress in medicine on multiple narrow aspects; He also reviewed the DeepMind controversial beginning push in medicine due to the risk of privacy;

the author tried to connect self-driving cars and medicine in Chapter four; e.g., Five levels of self-driving (from no automation to full automation)

chapter6: doctors with patterns, e.g. 1. radiologiests are conducting pattern-centric practices; 2. pattern-heavy elements in other primary care and specialites include such as scans or slides,

clinicians without patterns: most physicians, nurses and clinicians do not have pattern-centric practices; their predominant function is making an assessment and formulating a plan. Here the author reviewed the IBM Waston efforts in medicine; reviewed a few potential areas of AI: eye doctor, cancer doctor, heart doctor, surgeon, and other healthcares like neurologists;

Dr. Topol devoted a whole chapter to discuss Mental health and potential of AI for it, especially the “no judgemental” chatbots.

Dr. Topol did a wonderful summary of cutting edge Deep learning efforts in life science and drug discovery

Dr. Topol used a whole chapter to survey Diet, nutrition and potential of AI in this important and messy field

AI for clinical output prediction and Virtual medical assitant through better input channels like Alexa…

Deep empathy: the last chapter deeplyl discuss the potential of using AI to save time in medicine, which will result in deep bonding between patients and doctors.


https://www.amazon.com/Deep-Medicine-Artificial-Intelligence-Healthcare/dp/1541644638
https://t.iss.one/ArtificialIntelligenceArticles

EdgeNet: Semantic Scene Completion from RGB-D images. arxiv.org/abs/1908.02893

3 open faculty positions in cognitive neuroscience at NYU
Dear colleagues,

The Department of Psychology in the Faculty of Arts and Science at New York University is searching to hire up to three new faculty members in Cognitive Neuroscience. One search will be open rank and up to two will be at the tenure-track assistant professor level.

We seek applicants with an outstanding record of research in human cognitive neuroscience, broadly construed. Preference will be given to applicants whose program of research includes advanced measurement techniques and computational approaches to understanding brain and behavior. Applicants that bridge between or complement the existing research strengths of the Psychology Department are encouraged.

The Faculty of Arts and Science at NYU is at the heart of a leading research university that spans the globe. We seek scholars of the highest caliber, who embody the diversity of the United States as well as the global society in which we live. Because broad diversity is essential for creating an inclusive climate, we are committed to the fair treatment of and equal access to opportunity and advancement for all, and will assess the many qualifications of all applicants. We strongly encourage applications from women, racial and ethnic minorities, and other individuals who are under-represented in the profession, across color, creed, race, ethnic and national origin, physical ability, gender and sexual identity, or any other legally protected basis. NYU affirms the value of differing perspectives on the world as we strive to build the strongest possible university with the widest reach. To learn more about the FAS commitment to diversity, equality and inclusion, please read here: https://as.nyu.edu/facultydiversity.html

Review of applications will begin October 1, 2019.

The electronic application should include a CV, statements of research (no more than three pages) and teaching interests (no more than two pages), copies of at least three representative publications, and at least three reference letters.

Diversity and inclusion are important parts of the NYU mission. Applicants should include a statement describing how your (1) scholarship, (2) teaching and mentoring, and/or (3) service and engagement demonstrate your commitment to diversity, equity, and inclusion. We are particularly interested in hearing about (1) concrete steps you have taken (or are planning to take) to foster an inclusive intellectual environment in your lab, in the classroom, in the department and on campus, and/or in your field more generally, and (2) how these steps connect with your broader views on the topics of diversity, equity, and inclusion.

Please address any questions to the Assistant to the Department Chair, Paulo Campos at [email protected].

"A machine learning algorithm claims to predict heart attacks and death from heart disease with a degree of accuracy beating human practitioners."

https://www.artificialintelligence-news.com/2019/05/14/ml-algorithm-predicts-heart-attacks/

What kind of medical innovations can we see in 2019? Also, is Google and Microsoft taking the lead?

https://t.ly/Zj3m9

Blog Post Version: https://bit.ly/31tFbRD

#Medical #ArtificialIntelligence #healthcare #HealthTech #opioids #opioidcrisis #Google #Microsoft #AI #money

Generation of dopaminergic or cortical neurons from neuronal progenitors

Chen et al.: https://zenodo.org/record/3361000#.XU9Lny0ZPUK

"Yes its tough to make neurons, but its doable (...)"
- Thomas Durcan, in: https://openlabnotebooks.org/years-and-years-of-dopaminergic-neurons/

#Neurons #CorticalNeurons #DopaminergicNeurons

Facebook launches online Global Pytorch Hackathon. $61,000 in prizes. Submissions due Sept 16th.


https://pytorch.devpost.com/

I had the pleasure of attending their in person hackathon at Menlo Park yesterday. If you want some inspiration for potential projects, checkout their submissions page here, they were really good.

https://pytorchmpk.devpost.com/submissions

Pytorch rolled a bunch of new features out a few days ago. They seem to be really stepping up in response to TF 2.0.

If you're looking for teammates, signup on the page, then you can look at other profiles of those looking for teammates

https://pytorch.devpost.com/participants

Wide Neural Networks of Any Depth Evolve as Linear Models Under Gradient Descent
Lee et al.: https://arxiv.org/abs/1902.06720
#DeepLearning #MachineLearning #NeuralNetworks