Deep networks work by learning complex, often hierarchical internal representations of input data. These form a kind of functional language the network uses to describe the data.

Language can emerge from tasks like object recognition: has pointy ears, whiskers, tail => cat.

This relates to Wittgenstein’s "language-game" in Philosophical Investigations, where a functional language emerge from simple tasks before defining a vocabulary.

The visual vocabulary of a convolutional neural network seems to emerge from low level features such as edges and orientations, and builds up textures, patterns and composites, … and builds up even further into complete objects: houses, dogs, etc.

Source: NeurIPS 2018“Unsupervised Deep Learning” Tutorial – Part 1 by Alex Graves - https://lnkd.in/eXjxA2n

#artificialintelligence #deeplearning #language #machinelearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

If you ever need to find the shortest distance from a bunch of things to a bunch of other things, say when planning a distribution or communication network, I wrote a thing that might help.
It's a many-to-many variant of Dijkstra's shortest path algorithm.
It's in python and relies on numba to speed up the computationally intensive bits.
https://github.com/facebookresearch/many-to-many-dijkstra

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Very nice fast.ai just launched their 2019 edition of Practical Deep Learning for Coders. With 100% new materials, new design and applications that have never been covered by an introductory deep learning course before. Key applications covered are: Computer vision, NLP, Tabular data (e.g. sales prediction) and collaborative filtering. There are seven lessons in total (2hrs long) with assignments. So if you want to get started with deep learning today then here's your big chance and definitely check this course out! #deeplearning #machinelearning

Article: https://lnkd.in/d6y-wjm
Course: https://course.fast.ai/

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Every decade, in other words, has essentially seen the reign of a different technique: neural networks in the late ’50s and ’60s, various symbolic approaches in the ’70s, knowledge-based systems in the ’80s, Bayesian networks in the ’90s, support vector machines in the ’00s, and neural networks again in the ’10s.
The 2020s should be no different, says Domingos, meaning the era of deep learning may soon come to an end. But characteristically, the research community has competing ideas about what will come next—whether an older technique will regain favor or whether the field will create an entirely new paradigm.
“If you answer that question,” Domingos says, “I want to patent the answer.”"
https://lnkd.in/gvn9beG

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

The better you understand yourself, the more empowered you'll be in your life.

Here are a few resources to help you understand yourself better:

👉 Myers-Briggs personality test - https://lnkd.in/gJk_h5n
👉 HEXACO personality test - https://lnkd.in/gZaQ7TA
👉 StrengthsFinder 2.0 - https://lnkd.in/gpCxStq
👉 Managing Oneself - https://lnkd.in/gwkjftP

Start by taking the first personality test and then reading about your personality type.

Trust me, this can change your perspective on everything in your life.

#datascience #knowthyself

Edit: for those that were wondering, I’m currently an ENTJ-T :)


The better you understand yourself, the more empowered you'll be in your life.

Here are a few resources to help you understand yourself better:

👉 Myers-Briggs personality test - https://lnkd.in/gJk_h5n
👉 HEXACO personality test - https://lnkd.in/gZaQ7TA
👉 StrengthsFinder 2.0 - https://lnkd.in/gpCxStq
👉 Managing Oneself - https://lnkd.in/gwkjftP

Start by taking the first personality test and then reading about your personality type.

Trust me, this can change your perspective on everything in your life.

#datascience #knowthyself

Edit: for those that were wondering, I’m currently an ENTJ-T :)

The better you understand yourself, the more empowered you'll be in your life.

Here are a few resources to help you understand yourself better:

👉 Myers-Briggs personality test - https://lnkd.in/gJk_h5n
👉 HEXACO personality test - https://lnkd.in/gZaQ7TA
👉 StrengthsFinder 2.0 - https://lnkd.in/gpCxStq
👉 Managing Oneself - https://lnkd.in/gwkjftP

Start by taking the first personality test and then reading about your personality type.

Trust me, this can change your perspective on everything in your life.

#datascience #knowthyself

Edit: for those that were wondering, I’m currently an ENTJ-T :)

The better you understand yourself, the more empowered you'll be in your life.

Here are a few resources to help you understand yourself better:

👉 Myers-Briggs personality test - https://lnkd.in/gJk_h5n
👉 HEXACO personality test - https://lnkd.in/gZaQ7TA
👉 StrengthsFinder 2.0 - https://lnkd.in/gpCxStq
👉 Managing Oneself - https://lnkd.in/gwkjftP

Start by taking the first personality test and then reading about your personality type.

Trust me, this can change your perspective on everything in your life.

#datascience #knowthyself

Edit: for those that were wondering, I’m currently an ENT

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

The data journalism site Fivethirtyeight has made numerous datasets available from their articles:
https://lnkd.in/dHZugbc

#datascience #analytics #machinelearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Berkeley STAT-157 (Introduction to Deep Learning)

Lecture videos for the first week, by Alex Smola: https://lnkd.in/eW6MfzM

#artificialneuralnetworks #deeplearning #neuralnetworks

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

It's not easy for a College Student to graduate with Business Domain knowledge. Reading Whitepapers and Case Studies by Analytics Consulting/MR Companies (like Deloitte, Accenture, Garnter) could help. But entering their large site, sometimes you'd be a lost child.

Here's how you can use Advanced Search Operator in Google to extract only PDF output from a particular website along with a keyword like #Analytics or #Datascience thus get relevant case studies or whitepapers.

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

BEST DEEP LEARNING JOKES SO FAR

I can't stop laugh when watching this, an over-dramatization film on Deep Learning deployment. Funny, because its really often happen in Data Science live.

#deeplearning #python #anaconda #pyception

Disclaimer: This trailer is made by Anaconda, Inc.. This is made for AnacondaCON 2018 you can see full video in their youtube channel https://lnkd.in/fe5CW9N For 2019 in April you can register this https://anacondacon.io/

#technology #datascience #DataScientists

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

n Nature Machine Intelligence today: Whetstone, a method for turning general Keras neural networks into spiking neural networks: https://www.nature.com/articles/s42256-018-0015-y.epdf?author_access_token=HIFIT_s3XXRdKKF3DTspd9RgN0jAjWel9jnR3ZoTv0P7sMl50Mvxe5hygHWfkIWjiyJe1kEkFLNBiorlpBWGyE5yRNu7SaSa6rWLAwmUPf1dL47QUigBag24erZ3G6Ue-9ZkZNtWzrZVVkxMrGE8eA%3D%3D

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

I always prefer papers with source code so I can simply use them instead of spending plenty of time on implementing them. I wrote some code tonight to find Arxiv NLP papers with Github link and here is a list of papers from last December. Plan to run through the past year during this weekend. Hope it is useful.

https://kaggle.com/shujian/arxiv-nlp-papers-with-github-link

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

To become a data scientist, is it better to be a generalist or a specialist?

The answer: you need to be both.

There is a very broad set of requirements to work as a data scientist, and you need familiarity with all of them to do the job:

* Data loading
* Data manipulation
* Feature engineering
* Model selection
* Model tuning
* Model evaluation
* Coding
* Visualization
* Report creation
* Presenting
* Business acumen
* Etc

What you also need:

👉 One area of specialization where you bring unique expertise to the team.

Data science is a team sport, and in order to build an effective team, you need complimentary skills that lift the team above the sum of its parts.

Each individual should be able to function on their own, but also contribute a unique skills set to the team.

👉 Agree or disagree?

#datascience #teams #aspiring #DataScientists

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

The Matrix Cookbook: One of the most useful reference books out there...if you enjoy taking a deeper dive into Machine Learning algorithms
#MachineLearning #DataScience
http://bit.ly/2rjrQLU

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Which language should you learn to get into Data Science?

Python.

Ranked no. 1 Machine Learning Language on GitHub.

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Amazon Comprehend Medical – Natural Language Processing for Healthcare Customers | Amazon Web Services https://amzn.to/2QJLS0W #AI #DeepLearning #MachineLearning #DataScience

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

HERE’S SOME LATEST BREAKTHROUGH ON NLP SPACE

BERT
https://lnkd.in/fR6p4Ut
Sequence Classification with Human Attention
https://lnkd.in/fen6xB8
Phrase-Based & Neural Unsupervised Machine Translation
https://lnkd.in/fE4CfVF
Probing sentence embeddings for linguistic properties
https://lnkd.in/fHpE3KP
SWAG
https://lnkd.in/fgPSxTG
Deep contextualized word representations
https://lnkd.in/ftMAz-g
Meta-Learning for Low-Resource Neural Machine Translation
https://lnkd.in/fYF5Hsx
Linguistically-Informed Self-Attention for Semantic Role Labeling
https://lnkd.in/fkz8usu
A Hierarchical Multi-task Approach for Learning Embeddings from Semantic Tasks
https://lnkd.in/fGYsEcD
Unanswerable Questions for SQuAD
https://lnkd.in/fddKepX
An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
https://lnkd.in/fa6a8FJ
Universal Language Model Fine-tuning for Text Classification
https://lnkd.in/fnTzYpw
Improving Language Understanding by Generative Pre-Training
https://lnkd.in/fpA73wA
Dissecting Contextual Word Embeddings: Architecture and Representation
https://lnkd.in/fg6ck7w

Original by TOPBOTS https://lnkd.in/f_8R-8e


❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Nice article on the differences between the symbolic and imperative APIs in TensorFlow 2.0. In particular, it's about the differences between Keras's Sequential, Functional and Subclassing API. If you want to create something fast without much abstraction then you should go with the Keras Sequential and Functional API (like plugging together LEGO bricks). Otherwise you should go with the Subclassing API where you think about your models as object-oriented, a choice that my team and me actually prefer. Very insightful article. Definitely check it out! #deeplearning #machinelearning

Article: https://lnkd.in/drDN-NS

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Excited to announce StanfordNLP, a natural language processing toolkit for 53 languages with easily accessible pretrained models. It allows you to tokenize, tag, lemmatize, and (dependency) parse many languages, and provides a Python interface to CoreNLP.

https://stanfordnlp.github.io/stanfordnlp/

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

You will never be a data scientist without knowing #Calculus, #Probability and #InformationTheory.

www.interviews.ai

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Shuffling large datasets, have you ever tried that?

Here the author presents an algorithm for shuffling large datasets.
Here you learn the following;

0. why Shuffle in the first place?
1. A 2-pass shuffle algorithm is tested
2. How to deal with oversized piles
3. Parallelization & more

Link to article : https://lnkd.in/dZ8-tyJ
Gist on #Github: for a cool visualization of the shuffle https://lnkd.in/d8iK8fd

#algorithms #github #datasets #deeplearning #machinelearning

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN