This is an excellent overview on the state of the art methods for NLP (natural language processing). An exciting area of research with wide applications.

https://lnkd.in/eKt-fKK

#analytics #machinelearning #datascience #nlp

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

How to fail the Data Science business case
2. Recruiting as quick-fix — “I am looking to recruit 150 Data Scientists in the next 12 months”. I am not kidding: I did get that phone call, and it was about recruiting ‘Data-Scientists-as-consultants’. Yes, expertise matters. Yes, there is a shortage of talent. And, yes, as companies struggle to build up data science capabilities they likely will be keen on consultancy services. However, the shortage of experts is real. Moreover, a senior data scientist likely prefers building products over project work, and impact with customers over project management meetings. Overall, I have seen quite a few attempts at using recruiting-as-a-fix, often failing at implementation already, either because of an unrealistic ‘unicorn’ wishlist or because the case couldn’t be made as to why an experienced Data Scientists should join the company. Moreover, Data Scientists frequently report that they are interviewed by non-experts.
#interviews #datascientist #recruiting #machinelearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Machine Learning is Fun! Part 3: Deep Learning and Convolutional Neural Networks

by Adam Geitgey: https://lnkd.in/gZ6sdPW

#artificialintelligence #deeplearning #machinelearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

My ML Tip of the Week on Overfitting

💡 What is overfitting?

Overfitting is when a model makes much better predictions on known data (data included in the training set) than unknown data (data not included in the training set).

💡 How can you combat overfitting?

👉 A few ways of combating overfitting are:
• simplify the model by use fewer parameters
• simply the model by changing the hyperparameters
• simplify the model by introducing regularization
• select a different model
• use more training data
• gather better quality data

#datascience #machinelearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Text preprocessing principles:

- Text normalization
- Tokenization
- Removing stop words
- Stemming
- Lemmatization
- Part-of-speech tagging (POS)
- Chunking (shallow parsing)
- Named entity recognition
- Collocation extraction
- Relationship extraction

This article describes the points mentioned above in more details
https://lnkd.in/dzeFR7e

#NLP #DeepLearning
#MachineLearning

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

This slide from lex fridman is amazing. How many of these do you know?


🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Here is a simple explanation of what a Q-Learning is.

You see here Q-learning with value iteration, which is a reinforcement learning technique used for learning the optimal policy in a Markov Decision Process (MDP)

The authors explain the technique which works by introducing a game where a reinforcement learning agent tries to maximize points, and through this.

Small introduction on Q-tables and the trade-off between exploration and exploitation is also given.

Credits: deeplizard , Please see full list here
https://lnkd.in/dmuaGsw
#artificialintelligence #reinforcementlearning #deeplearning #datascience

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

We offer a number of fully-funded fellowship opportunities to students applying for the Master of Computer Science or PhD program in the Faculty of Computer Science at Dalhousie University, Canada. As seen on their website👇🏿
https://deepsense.ca/fellowships/

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

AIMS is accepting applications for fully-funded UniofOxford PhD places to work on machine learning, vision, robotics, sensor networks and more. The deadline is 25 Jan. More details and the application site: https://www.ox.ac.uk/admissions/graduate/courses/autonomous-intelligent-machines-and-systems?wssl=1 … Please spread the word!

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Start the new year applying for a fully-funded 4-year PhD position in my lab on formal verification of ethical principles of artificial intelligence: https://umu.mynetworkglobal.com/en/what:job/jobID:241808/ … Still a few days to apply! #AI #AIethics #responsibleAI

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

One of the best I’ve seen in a LONG time!

Except, The Java 8 streaming API / Java Lamda functions have completely changed how I look at the language.

Nevertheless, Python was and is my first love so... #FunctionalProgramming

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

If you enjoy maintaining a well-organized Bookmark lists for the work you do e.g. #datascience or #machinelearning, then you can easily share that those resources with everyone in the world.

Just import the bookmark into an HTML -> Use an #HTML-to-#markdown converter -> copy-paste that Markdown text into a Readme.md of a #Github repo and you are done.

If you have a good collection, I bet your Github will attract stars in no time.

Here is mine (there are many repetitions but I will clean them up in a few weeks)
https://github.com/tirthajyoti/Data-science-best-resources
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

The Doctoral College "Imagint the Mind" is inviting applications for 12 fully funded PhD studentships in cognitive (neuro-)science, psychology, biology, medicine/neurology, or computational neuroscience. More info on this amazing programme: https://phdim.ccns.sbg.ac.at/

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

How do you detect outliers in #data?

Do you use a blanket rule of anything outside 3 standard deviations?

Or do you use a more robust method?

If you have a resource you learned from or one you created. I'd love to reference it in my article on exploratory data analysis.

If you want to read it, there's a link in the comments. #EDA is one of the areas I've learned the most over the past year.

I remember things best if I write about them. So that's what I did.

PS There's more pretty pictures like this one in there too 🎨

#datascience

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

YES! YOU CAN MAKE COMPUTER CAN HEAR! (WITH PYTHON)

Science behind Google Duplex, Google Home, Amazon Alexa, Amazon Echo, Apple Siri, Microsoft Cortana etc!

Audio Processing has massive potential in #DeepLearning. Here are 7 articles to get started with audio processing #algorithms and how to implement them in #Python:

1. Librosa, library for voice recognition
https://lnkd.in/fb6CPYJ

2. 10 Audio Processing Tasks to get started with Deep Learning Applications - https://lnkd.in/fDDfkyw

3. Getting Started with Audio Data Analysis using Deep Learning - https://lnkd.in/fMQWNwv

4. Heart Sound Segmentation using Deep Learning (A Doctor in the Making?) - https://lnkd.in/fsqtfBQ

5. Learn Audio Beat Tracking for Music Information Retrieval - https://lnkd.in/fG4PWyX
Faizan Shaikh

6. Join Voice Information Retreival community
https://lnkd.in/fu-Qgt5

7. Complete tutorial for Speech Recognition on Python
https://lnkd.in/fszVwie

#analytics #artificialintelligence

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

DATA VISUALIZATION DO AND DON'TS

1. Time axis. When using time in charts, set it on the horizontal axis, from left to right. Don't skip values (time periods).

2. Proportional values. The numbers in charts should be proportional with size.

3. Data-Ink Ratio. Remove excess information, lines, colors, and text from charts that doesn't add value.

4. Sorting. For column and bar charts, sort your data in ascending/descending order by the value, not alphabetically.

5. Labels. Use labels directly on the line, column, bar, pie whenever possible, avoid indirect look-up!

6. Inflation adjustment. When using long-term series monetary values, adjust for inflation!

7. Colors.
a. Don’t use more than six colors.
b. For comparison at different time periods, use the same color (different intensity).
c. For different categories, use different colors.
d. Keep the same color palette, axes and labels for similar charts.
e. If you can't distinguish color differences on greyscale print, change hue and saturation of colors!
f. Ensure that charts are readable for color-blind people. Use Vischeck test or color palettes that color-blind friendly.

8. Data Complexity.
Don’t add too much information to a single chart. Split into two if needed.

9. Expressions. Avoid using technical expression.

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

If you're new to Python or Web Development with Django or Machine Learning with Python, then you must visit Sentdex.

Sentdex is a popular programming YouTube channel run by Harrison Kinsley.

He's also referred to as Sentdex (because he named his channel Sentdex).

He has a very cool way of introducing topics and the best part is that he builds stuff and teaches it together.

His videos are very conversational and easy to understand if you're just starting out or even if your experienced.

Another interesting thing is that he uses Windows for all his programming, which is not seen in programming tutorials very often.

You can check him out here:
https://lnkd.in/fu9rQBU

#datascience #machinelearning #python #programming

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

3-yr post-doc in computational neuroscience in LMU Munich Germany Our new project funded by the NOMIS foundation To examine how diversity of opinion affects collective decisions and belief-change. Deadline 15 Feb https://www.cvbe.philosophie.uni-muenchen.de/vacancies-job-offers/index.html
🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

Seeking a computational postdoc interested in relating representational (embedding) spaces, including those from brain, real-world behaviour, and convolutional networks, at UCL (http://bradlove.org ). Neuro/ML. Any nationality, open search, all welcome

https://atsv7.wcn.co.uk/search_engine/jobs.cgi?amNvZGU9MTc4NTc5NyZ2dF90ZW1wbGF0ZT05NjUmb3duZXI9NTA0MTE3OCZvd25lcnR5cGU9ZmFpciZicmFuZF9pZD0wJmpvYl9yZWZfY29kZT0xNzg1Nzk3JnBvc3RpbmdfY29kZT0yMjQ%3D&jcode=1785797&vt_template=965&owner=5041178&ownertype=fair&brand_id=0&job_ref_code=1785797&posting_code=224

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN

7 STEPS MAKE AN OBJECT DETECTION USING DEEP LEARNING

1. Collect Data
These deep learning machines that have been working so well need fuel — lots of fuel; that fuel is data. The more labelled data we have, the better our model performs.
https://lnkd.in/gNa78_Q

2. Labeling Bounding Box
If your data isn't labelled, you can try LabelImg is a graphical image annotation tool and label object bounding boxes in images
Youtube: https://lnkd.in/gJhhd9R
Github: https://lnkd.in/gDW8GEb

3. Feature Engineering
Histogram of Image - https://lnkd.in/gjDUa7F

4. Deep Learning
Convolutional Neural Network https://lnkd.in/gZ74d4W

5. YOLO
https://lnkd.in/g6EHch2

6, Add some extra needs,
In this vidio use empty parking spot detector

7. Putting ALL Together
Adam Geitgey used deep learning to detect when a parking space becomes available and then also a message will be sent to him. He calculated the intersection over union (IoU) on the car’s bounding box to see if it is overlapping with a parking spot’s bounding box. Finally, he used Twilio to send a message to him when there's a parking spot available. Code is also provided. #deeplearning #machinelearning

Article: https://lnkd.in/dB2QacM

🗣 @AI_Python_Arxiv
✴️ @AI_Python_EN