Nice tips and tricks for training neural networks by Andrej Karpathy. Most important point which I also can agree on based on my experience: "becoming one with the data" which means understanding your dataset (e.g. understanding distributions, looking for patterns etc.) is core to training your neural network as "neural net is effectively a compressed/compiled version" of the dataset. There are many more other interesting points around tuning the model, establishing model baseline etc.. Definitely check it out. It will save your time to make training neural networks right. #deeplearning #machinelearning

🌎 Link: https://lnkd.in/dppUnnT

✴️ @AI_Python_EN

Building a #Conversational #AI #Agent for medical and healthcare services is one of the products in our pipeline in the coming months.

Here is how a typical chatbot recirculation recurrent #pipeline looks like

#CNN #RNN #GAN #DeepLearning #NLP

✴️ @AI_Python_EN

"Neural Networks for Machine Learning by Geoffrey Hinton" (Coursera 2013)

Video Lectures: https://lnkd.in/eSJjXGd

#ArtificialIntelligence #DeepLearning #NeuralNetworks #MachineLearning

✴️ @AI_Python_EN

What type of a presenter are you?

Are you a "diva", a "penguin" or "Mr. Toscanini"?

Presenting your #MachineLearning #AI #research or project is an art which you must master very well to succeed.

In our internal lectures / classes we do our best to teach our team members to develop a great storyline and present like a star.

#presentationskills #AI #soft #skills

✴️ @AI_Python_EN

"Training Neural Nets on Larger Batches: Practical Tips for 1-GPU, Multi-GPU & Distributed setups"

By Thomas Wolf: https://lnkd.in/etyMzjQ

#ArtificialInteligence #DeepLearning #MachineLearning #NeuralNetworks #Research

✴️ @AI_Python_EN

Reinforcement Learning: An Introduction

Book by Andrew Barto and Richard S. Sutton

Link - https://lnkd.in/f6byhDw

#artificialintelligence #deeplearning #reinforcementlearning

✴️ @AI_Python_EN

The best way to learn #DeepLearning is by practicing it. But which framework to use? Here are 5 articles to get you started!

A Comprehensive Introduction to #PyTorch - https://bit.ly/2L8Rj7n

Learn How to Build Quick & Accurate Neural Networks using PyTorch (& 4 Case Studies) - https://bit.ly/2Vts9nY

Get Started with Deep Learning using #Keras and #TensorFlow in #R - https://bit.ly/2Iro2BY

TensorFlow 101: Understanding Tensors and Graphs - https://bit.ly/2GNg195

An Introduction to Implementing #NeuralNetworks using TensorFlow - https://bit.ly/2V17cBs

✴️ @AI_Python_EN

Must Read Articles For Data Science Enthusiat.

1) Every Intro to Data Science Course on the Internet, Ranked.
(https://lnkd.in/fQDMiNX)

2) What would be useful for aspiring data scientists to know?
(https://lnkd.in/fmcFyN7)

3) 8 Essential Tips for People starting a Career in Data Science.
(https://lnkd.in/f5vUg6i)

4) Cheat sheet: How to become a data scientist.
(https://lnkd.in/fMEhi4D)

5) The Art of Learning Data Science.
(https://lnkd.in/fruY2AC)

6) The Periodic Table of Data Science.
(https://lnkd.in/fxReDab)

7) Aspiring Data Scientists! Start to learn Statistics with these 6 books!
(https://lnkd.in/fXSE-us)

8) 8 Skills You Need to Be a Data Scientist.
(https://lnkd.in/f8S3Ygd)

9) Top 10 Essential Books for the Data Enthusiast
(https://lnkd.in/fKugicE)

10) Aspiring data scientist? Master these fundamentals.
(https://lnkd.in/fTGDkju)

11) How to Become a Data Scientist - On your own.
(https://lnkd.in/f_Zhpzf)

#datascience #neverstoplearning

✴️ @AI_Python_EN

Fashion++: Minimal Edits for Outfit Improvement https://arxiv.org/pdf/1904.09261.pdf

✴️ @AI_Python_EN

Swift + TensorFlow

Create a simple NN and CNN.

Notebook by Zaid Alyafeai: https://lnkd.in/e5zWxZ5

#ArtificialIntelligence #DeepLearning #NeuralNetworks

✴️ @AI_Python_EN

Many people working in data analysis believe that there's something special about Python (or R, or Scala). They will tell you that you have to use one of those because otherwise, you will not get the best result.

That is of course not true. The choice of language should be based on two factors:

1) How well your final product will integrate with the existing ecosystem.

2) The availability of production-grade data analysis libraries.

Currently, almost any popular language has one or more powerful libraries for data analysis. Java is an excellent example, where the development of everything hot is happening right now because of a multitude of existing JVM languages.

C++ historically has a huge choice of implemented algorithms. Even proprietary ecosystems such as .Net today contain an implementation of most state-of-the-art algorithms and machine learning paradigms.

So, if the person you consider hiring to work on your data analysis project tells you that only #Python is the way to go, I would be skeptical and look for someone who embraces diversity.

✴️ @AI_Python_EN

an ultra-short micro-course that gives you a fast way to try Python and start using it for data visualization.
This micro-course won’t teach you computer science, and it skips most parts of the Python programming language. But you’ll learn enough to impress colleagues or potential employers with nicer graphics than anyone makes in Excel.



https://www.kaggle.com/learn/data-visualization-from-non-coder-to-coder

✴️ @AI_Python_EN

https://github.com/pytorch/pytorch/releases
Official TensorBoard Support, Attributes, Dicts, Lists and User-defined types in JIT / TorchScript, Improved Distributed
✴️ @AI_Python_EN

Current State of Deep Learning from Francois Chollet (Keras creator)


While very strong future perspective, it's still a data-fitting model and require lots and lots of data. In contrast, human brain can work with fewer data but use abstract model of the world.

Video link here

#deeplearning

✴️ @AI_Python_EN

George Box's observation that "Essentially, all models are wrong, but some are useful" is one of the most quoted of all statistical proverbs.

However, we need to ask "Useful for what?"

There are many #machinelearning #algorithms and statistical models that are difficult to interpret but useful in predictive analytics. Sometimes all we need are predictions and classifications that are sufficiently accurate for decision purposes.

Often in the business world there is little or no theory to guide statisticians and data scientists. Moreover, we may not have the data necessary for a good understanding of why some customers are heavier purchasers of our product than others, for instance.

That said, predictions and classifications that are "accurate enough" often aren't good enough. We may need a reasonable - if imperfect - understanding of the Why for these predictions and classifications to be useful.

This is obvious in "hard" scientific research but just as true in the behavioral and social sciences, marketing included.

Being able to design primary studies and having a good grasp of causal analysis, IMO, is necessary to be a full stack analytics professional, as opposed to being a full stack programmer or IT professional. These are different occupations.


"Causation: The Why Beneath The What," an interview with Tyler VanderWeele, a Harvard epidemiologist and authority on causal analysis might be of interest
http://www.greenbookblog.org/2017/07/17/causation-the-why-beneath-the-what/

✴️ @AI_Python_EN

Transfer Learning is a boon to #DeepLearning when you don't have much data of your own.

This allows you to succeed with trained datasets that have worked hard on solving similar problems in #computer #vision or #nlp

Higher start, higher slope and higher asymptote are key ways to know that your model will be performing better.

#performance #machinelearning #transferlearning #model

✴️ @AI_Python_EN

In the context of analytics, the terms longitudinal and time-series refer to data covering more than one time-period (i.e., not cross-sectional data).

The terms are often seen as interchangeable but time-series tends to be used more often when there are many periods, e.g., four years of weekly sales data. On the other hand, some see longitudinal as a generic term that includes high-frequency data.

Whatever you call it, there's a lot more of it now than ever and more ways to analyze it than ever, including neural networks architectures such as LSTM.

Here are some books I've found helpful which cover this topic from a statistical angle:

- Longitudinal Analysis (Hoffman)
- Longitudinal Structural Equation Modeling (Newsom)
- Growth Modeling (Grimm et al.)
- Age-Period-Cohort Analysis (Yang and Land)
- Analysis of Longitudinal Data (Diggle et al.)
- Applied Longitudinal Data Analysis for Epidemiology (Twisk)
- Modeling Dynamic Relations (Pauwels)
- Time Series Analysis and Its Applications (Shumway and Stoffer)
- Time Series Analysis (Wei)

(Some titles are abbreviated.)

There are many more, including those focused on fields such meteorology, environmental studies and financial econometrics, but these are good places to start if you'd like to learn more about this topic.


Richly Parameterized Linear Models: Additive, Time Series, and Spatial Models Using Random Effects (Hodges) is a very interesting if (for me challenging) book that might be of interest to some of you -

https://www.crcpress.com/Richly-Parameterized-Linear-Models-Additive-Time-Series-and-Spatial/Hodges/p/book/9781439866832

✴️ @AI_Python_EN

Cornell University - Machine Learning for Intelligent Systems (CS4780/ CS5780)

I highly recommend the Cornell University's "Machine Learning for Intelligent Systems (CS4780/ CS5780)" course taught by Associate Professor Kilian Q. Weinberger.


Youtube Video Lectures:
https://www.youtube.com/playlist?list=PLl8OlHZGYOQ7bkVbuRthEsaLr7bONzbXS
Course Lecture Notes:
http://www.cs.cornell.edu/courses/cs4780/2018fa/lectures/

#artificialintelligence #machinelearning #deeplearning #AI #algorithms #computerscience #datascience

✴️ @AI_Python_EN

This is lecture 3 in the series on Wasserstein #GAN. In this lecture, basic understanding of Wasserstein Generative

Adversarial Network (WGAN) is discussed

videos


✴️ @AI_Python_EN

✴️ @AI_Python_EN

#Mathematics for #artificialintelligence

https://rubikscode.net/2019/04/29/mathematics-for-artificial-intelligence-linear-algebra/