Writing Code for NLP Research

Slides by the Allen Institute for Artificial Intelligence: https://lnkd.in/eubgSGY

#naturallanguageprocessing #NLP #research

If you like our channel, i invite you to share it with your friends

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Here are three nice posts/pages regarding Gaussian Processes which help illuminate some Bayesian concepts.
#datascience

Gaussian Processes are Not so Fancy
https://bit.ly/2F8jaRu
(Python implementation)

A Visual Exploration of Gaussian Processes
https://bit.ly/2BQGwXA
(Cool interactive features)

Robust Gaussian Processes in Stan
https://bit.ly/2COMgme
(R implementation using stan library)

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

I won't claim to be an authority on neural nets, but here are some books on ANN I can recommend:

- Neural Network Design (Hagan)

- Deep Learning and Neural Networks (Heaton)

- Deep Learning (Goodfellow et al.)

- Deep Learning with R (Chollet and Allaire)

- Neural Networks and Deep Learning: A Textbook (Aggarwal)

- Neural Network Methods in Natural Language Processing (Goldberg)

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

10 Predictions for #DeepLearning in 2019 – Intuition Machine

https://bit.ly/2F26VG4

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

GitHub Free users now get unlimited private repositories
link
❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

#DeepSpeech --> how a speech application works.

Project DeepSpeech is an open source Speech-To-Text engine, using a model trained by machine learning techniques, based on Baidu, Inc.'s Deep Speech research paper.

Project DeepSpeech uses Google's TensorFlow project to make the implementation easier.

Paper: https://lnkd.in/di6kSyB
Github: https://lnkd.in/dka5rWn

#ArtificialIntelligence #NLP #speechrecognition #DeepLearning #machinelearning

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

How would a rockstar 🎸would improve their machine learning models?
چگونه مدلهای یادگیری ماشین را بهبود دهیم؟

To get better at playing the guitar, you play the guitar more. You try different songs, different cords. Practice, practice, practice.

All the practice adds up to more experience, more examples of different notes.

And to try something totally different, you might merge two songs together. Or even take a song written originally for the piano but play it on your guitar.

After a while, you're ready to play a show. But the show won't some any good if all the speakers are set to different settings. Steve the sound guy takes care of this.

How does this relate #machinelearning?

1. More practice = more data

More examples of playing different notes = more data. Machine learning models love more data.

2. Combining different songs = feature engineering

If the #data you have isn't in the form you want, transforming into a different shape may be a better way of looking at it.

3. Tuning the speakers = hyperparameter tuning

There's a reason tuning the speakers is the last step in playing a rock show. Working speakers don't mean anything without all the practice (collecting data) and songwriting (feature engineering). If you've done 1 and 2 right, this is the easy part.

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Here is another Learning path of Deep Learning. Check it out.


Link to complete article: https://lnkd.in/fQTtuex

#deeplearning


❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

What Should I Do Now? Marrying Reinforcement Learning and Symbolic Planning

Paper by Gordon et al.: https://lnkd.in/eVqQ4BB

#artificialintelligence #machinelearning #reinforcementlearning

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Great new resource with code, math, explanations, textbook, upcoming videos: Berkeley's Spring 2019 Introduction to Deep Learning

These are the topics:

- A Taste of Deep Learning

- Deep Learning Basics

- Deep Learning Computation

- Convolutional Neural Networks

- Recurrent Neural Networks

- Optimization Algorithms

- Computational Performance

- Computer Vision

- Natural Language Processing

https://lnkd.in/fhVEJWm

#deeplearning #machinelearning #artificialintelligence

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

CS224n: Natural Language Processing with Deep Learning
#NLP #deeplearning
Stanford / Winter 2019

Public lecture videos: Once the course has completed, we plan to also make the videos publicly available on YouTube.

http://web.stanford.edu/class/cs224n/


This year, CS224n will be taught for the first time using #PyTorch rather than #TensorFlow

🙏Thanks to: @cyberbully_gng
❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

A guide to deep learning in healthcare

https://www.nature.com/articles/s41591-018-0316-z

❇️ @AI_Python
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Artificial Intelligence used to create inexpensive heart disease detector.

https://bit.ly/2CXeewh

#IoT #BigData #MachineLearning #ML #fintech #tech #blockchain #DeepLearning #DataScience #cyberecuin



🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Facilitate proactive #cybersecurity threat hunting, detection, & analysis with game-changing technical capabilities (#BigData Analytics + #MachineIntelligence)

https://www.oreilly.com/ideas/modernizing-cybersecurity-approaches

#DataScience #BehavioralAnalytics #MachineLearning #AI

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

FML: Face Model Learning from Videos Multi-frame video-based self-supervised learning to reconstruct 3D faces, without strong face priors.

https://gvv.mpi-inf.mpg.de/projects/FML19/

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

The power of correctly framing your data science results...

Let's say you create a model that predicts sales leads with 3% accuracy, while only 1% of the population are true leads.

1. Your model is wrong 97% percent of the time. (Ouch.)
2. Your model is right 300% as often as a baseline approach. (Nice.)

Which one sounds better?

They're both saying the exact same thing, but presented in a different way. They have been framed differently.

And people will react to them very differently.

👉 What does this mean for data scientists?

It means that when you have a good result, it is not enough to simply present the numbers. You must frame them appropriately for the audience to understand the value of the work.

When presenting, make sure that you understand the needs and expectations of your audience so that you can communicate in way that presents your results in a favorable light.

✅ Focus on the positive, not the negative.

✅ Focus on improvements, not shortcomings.

✅ Focus on opportunities, not problems.

✅ Focus on what you learned, not where you failed.

#datascience #cognitivebiases #communication

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) Network

Paper by Alex Sherstinsky: https://lnkd.in/enTMCDH

#RecurrentNeuralNetwork #RNN #LongShortTermMemory #LSTM #NeuralNetworks

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Best Paper Awards in Computer Science (since 1996)

A well maintained list: https://lnkd.in/e6_ks3E

#artificialintelligence #machinelearning #papers #research

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

How to get started with data science if you don't like learning theory:

The best thing to do in your situation is to find a project and start working on it.

➡️ Grab a dataset and formulate a problem that you think you can solve using the data.

Then, begin working through solving it.

As you get stuck, hop online or grab a book and learn what you need to keep pushing the project forward.

Then, once you finish the project, you can evaluate weaknesses and find areas that can be improved.

Again, go back and learn what you need to learn to improve your project.

You can repeat this iterative process as many times as you need to until you've got something that really makes you a standout candidate and you can start showing it off in your portfolio.

➡️ Here are a few places to find datasets to get you started:

Kaggle datasets - https://lnkd.in/gzz_ZWd
UCI dataset repo - https://lnkd.in/g_f8sag
Google dataset search - https://lnkd.in/egee4gR

#datascience #machinelearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

A new study published in Nature shows we can predicting the genetic disorder directly from the face using deep learning. The network was trained on a dataset of 17,000 patient images representing more than 200 syndromes. The paper reports that model achieves 91% top-10-accuracy in identifying the correct syndrome on 502 images and outperformed expert clinicians in three experiments. The method can be used to diagnose dysmorphology syndromes which typically affect roughly 1 in 30,000 people.
While this work has a great potential to improve discovering rare diseases, insurance companies may use this technology to deny providing medical insurance or increase the policy fees for people with specific genes.
paper: https://lnkd.in/fUEpYRt
#artificialintelligence #syndrome #ai #deeplearning #research

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

PhDs linked to #DataScience or #artificialintelligence

Turing Institute PhD Studentships

https://www.qmul.ac.uk/scholarships/items/turing-institute-phd-studentships.html

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN