With packages like caret and sci-kit-learn, the implementation of machine learning algorithms is quite easy. The most challenging part of machine learning is to understand the underlying model metrics, parameter tuning conditions and choosing the right model evaluation metrics.
For example, If you're working on a regression problem, metrics like MSPE, MAPE, R-square, and Adj. R-square is valued more than accuracy per se. In the case of the classification problem, metrics like Precision-Recall, ROC-AUC, Accuracy, and Log-loss play a vital role.
Choosing the right parameters/metrics to create and evaluate models is the most important of machine learning implementation than just using a package or function is to create a model with no intent. The capability as mentioned earlier requires a lot of hands-on experience, domain knowledge, and research.
Are you evaluating your models effectively?
Share your thoughts and insights with the community in the comments below.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
For example, If you're working on a regression problem, metrics like MSPE, MAPE, R-square, and Adj. R-square is valued more than accuracy per se. In the case of the classification problem, metrics like Precision-Recall, ROC-AUC, Accuracy, and Log-loss play a vital role.
Choosing the right parameters/metrics to create and evaluate models is the most important of machine learning implementation than just using a package or function is to create a model with no intent. The capability as mentioned earlier requires a lot of hands-on experience, domain knowledge, and research.
Are you evaluating your models effectively?
Share your thoughts and insights with the community in the comments below.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
"For many companies, deploying AI is slower and more expensive than it might seem."
Recent article in MIT Technology Review.
https://www.technologyreview.com/s/612897/this-is-why-ai-has-yet-to-reshape-most-businesses/
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Recent article in MIT Technology Review.
https://www.technologyreview.com/s/612897/this-is-why-ai-has-yet-to-reshape-most-businesses/
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
image_2019-02-21_22-40-24.png
3.3 MB
Download this very useful Infographic - it includes a step-by step process of cleaning text data in python using a Twitter case study. #Python
@AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
@AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
How deep learning is being used by a Canadian dairy farmer, a Kenyan microbiologist, a 73-year old starting a second career, an Australian accountant expanding use of solar power, a son of refugees in cybersecurity, & a cancer genomics researcher:
https://www.fast.ai/2019/02/21/dl-projects/
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
https://www.fast.ai/2019/02/21/dl-projects/
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
9 YouTube Channels That Will Teach You Everything You Need to Know About #artificialintelligence
🌎 Link
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
🌎 Link
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
In time-series analysis, a good-fitting and reasonable model may actually be misleading.
It's usually wise to test for "regime switching" since different processes (models) may be operating in different sections of the data series.
This is not merely a geeky issue and is very important in fields such as marketing, where consumer behavior and responses to marketing can change over time, sometimes suddenly.
In cross-sectional data analysis as well we must be careful about total sample models, since segments of consumers, some hidden, can react differently to new products and other marketing activity.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
It's usually wise to test for "regime switching" since different processes (models) may be operating in different sections of the data series.
This is not merely a geeky issue and is very important in fields such as marketing, where consumer behavior and responses to marketing can change over time, sometimes suddenly.
In cross-sectional data analysis as well we must be careful about total sample models, since segments of consumers, some hidden, can react differently to new products and other marketing activity.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Forwarded from DLeX: AI Python (Farzad🦅🐋🐕🦏🐻)
Innovation Nation: AI godfathers gave Canada an early edge — but we could end up being left in the dust
"Canada is hanging on to the lead by 'our fingernails' as the gold rush to commercialize artificial intelligence goes global"
By James McLeod: https://lnkd.in/emxjPNS
#artificialintelligence #business #deeplearning
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
"Canada is hanging on to the lead by 'our fingernails' as the gold rush to commercialize artificial intelligence goes global"
By James McLeod: https://lnkd.in/emxjPNS
#artificialintelligence #business #deeplearning
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Week 2 of AI for Everyone is all about identifying and building AI projects. Check out what you’ll learn:
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
In the context of algorithm fairness in AI. From the O'Reilly Data Show podcast, Sharad Goel says: "You need domain expertise. There's no silver bullet. There's no black box you stick your algorithm into and it gives you thumbs up or down. It always depends on domain and context".
Examples in the podcast and paper are from loan approvals and the elusive idea of "fairness".
It's encouraging that people are socializing the challenges and limitations of data driven approaches and the role domain context plays in it.
This also has ramifications for organizations trying to create effective analytics strategies: it's hard delegate analytics because the needed domain context originates outside of the analytics space (which also explains why outsourcing analytics/data science offshore has not been successful).
Finally, on the idea of fairness, it's worth pointing out this is a bit if old news. Thomas Lumley points out in a similar article "This is old news in medical diagnostics, but appears not to have been considered in some other areas".
O'Reilly Data Show - O'Reilly Media Podcast -
https://lnkd.in/e9ryTZc
#analytics #datascience #machinelearning #statistics
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Examples in the podcast and paper are from loan approvals and the elusive idea of "fairness".
It's encouraging that people are socializing the challenges and limitations of data driven approaches and the role domain context plays in it.
This also has ramifications for organizations trying to create effective analytics strategies: it's hard delegate analytics because the needed domain context originates outside of the analytics space (which also explains why outsourcing analytics/data science offshore has not been successful).
Finally, on the idea of fairness, it's worth pointing out this is a bit if old news. Thomas Lumley points out in a similar article "This is old news in medical diagnostics, but appears not to have been considered in some other areas".
O'Reilly Data Show - O'Reilly Media Podcast -
https://lnkd.in/e9ryTZc
#analytics #datascience #machinelearning #statistics
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
💡 What is multi-armed bandit testing and why is it useful?
Multi-armed bandit (or simply "bandit") testing is similar to multivariate and A/B testing, but the sampling distribution for variants change gradually over time as feedback is received.
For example, with traditional A/B tests, we could test 2 email subject lines: A and B. We would initially send out emails to 200 customers, sending 100 A variations and 100 B variations. After some set period of time, say 24 hours, we would observe which email variant was opened by more customers. We would then send that variant to all customers moving forward.
With bandit testing, we would set some learning rate for the distribution of variants to change over time. Perhaps 60 customers opened the A variant emails and only 50 customers opened the B variant emails. We could then shift the distribution from (50% A, 50% B) to (55% A, 45% B) for the next round of emails.
Using this approach, we can continuously monitor the response from our audience and shift our actions accordingly. This is particularly useful in marketing or any industry where people's preferences and opinions may change rapidly since it continuously tests and learns preferences and can adapt very quickly.
#datascience #dsdj #multiarmedbandit
#QandA
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Multi-armed bandit (or simply "bandit") testing is similar to multivariate and A/B testing, but the sampling distribution for variants change gradually over time as feedback is received.
For example, with traditional A/B tests, we could test 2 email subject lines: A and B. We would initially send out emails to 200 customers, sending 100 A variations and 100 B variations. After some set period of time, say 24 hours, we would observe which email variant was opened by more customers. We would then send that variant to all customers moving forward.
With bandit testing, we would set some learning rate for the distribution of variants to change over time. Perhaps 60 customers opened the A variant emails and only 50 customers opened the B variant emails. We could then shift the distribution from (50% A, 50% B) to (55% A, 45% B) for the next round of emails.
Using this approach, we can continuously monitor the response from our audience and shift our actions accordingly. This is particularly useful in marketing or any industry where people's preferences and opinions may change rapidly since it continuously tests and learns preferences and can adapt very quickly.
#datascience #dsdj #multiarmedbandit
#QandA
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
14 NLP Research breakthroughs you can apply to your business:
1. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
2. Sequence Classification with Human Attention
3. Phrase-Based & Neural Unsupervised Machine Translation
4. What you can cram into a single vector: Probing sentence embeddings for linguistic properties
5. SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference
6. [ELMo] Deep contextualized word representations
7. Meta-Learning for Low-Resource Neural Machine Translation
8. Linguistically-Informed Self-Attention for Semantic Role Labeling
9. A Hierarchical Multi-task Approach for Learning Embeddings from Semantic Tasks
10. Know What You Don’t Know: Unanswerable Questions for SQuAD
11. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
12. Universal Language Model Fine-tuning for Text Classification
13. Improving Language Understanding by Generative Pre-Training
14. Dissecting Contextual Word Embeddings: Architecture and Representation
Source: TopBot https://lnkd.in/eCYFvFE
#nlp #deeplearning
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
1. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
2. Sequence Classification with Human Attention
3. Phrase-Based & Neural Unsupervised Machine Translation
4. What you can cram into a single vector: Probing sentence embeddings for linguistic properties
5. SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference
6. [ELMo] Deep contextualized word representations
7. Meta-Learning for Low-Resource Neural Machine Translation
8. Linguistically-Informed Self-Attention for Semantic Role Labeling
9. A Hierarchical Multi-task Approach for Learning Embeddings from Semantic Tasks
10. Know What You Don’t Know: Unanswerable Questions for SQuAD
11. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
12. Universal Language Model Fine-tuning for Text Classification
13. Improving Language Understanding by Generative Pre-Training
14. Dissecting Contextual Word Embeddings: Architecture and Representation
Source: TopBot https://lnkd.in/eCYFvFE
#nlp #deeplearning
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Here's a comprehensive learning path for any person wanting to get into the field of deep learning. This path contains plenty of resources, links, ideas and suggestions to get you on your way! Also download the below infographic - A very handy resource for every data science professional.
https://lnkd.in/f3yVDkW
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
https://lnkd.in/f3yVDkW
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Time Series forecasting & modeling plays an important role in data analysis. The best way to learn #TimeSeries techniques is by applying them on time series data!
Here's a really cool #dataset where you need to
#forecast the traffic for a startup's product. Close to 10,500 data scientists have taken this challenge - can you climb up the leaderboard?https://lnkd.in/fyZiCJt
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Here's a really cool #dataset where you need to
#forecast the traffic for a startup's product. Close to 10,500 data scientists have taken this challenge - can you climb up the leaderboard?https://lnkd.in/fyZiCJt
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
François Chollet
The complaints "Python is slow" or "Python is unsafe" seem misguided. The point of Python isn't to be fast or safe, it's to be flexible and hackable, and to interface well with everything else. It has become successful by serving as a frontend from which to call other libraries.
#Python is more of an interface than it is a development language. It's a UX.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
The complaints "Python is slow" or "Python is unsafe" seem misguided. The point of Python isn't to be fast or safe, it's to be flexible and hackable, and to interface well with everything else. It has become successful by serving as a frontend from which to call other libraries.
#Python is more of an interface than it is a development language. It's a UX.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Yann LeCun
Video of my talk at the Institute of Advanced Studies workshop "Deep Learning: Alchemy or Science?", organized by Sanjeev Arora Friday Feb 22, 2019. The audience was very diverse, so I focused on the early history and dynamics of ideas in neural..
🌎 The epistemology of Deep Learning"
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Video of my talk at the Institute of Advanced Studies workshop "Deep Learning: Alchemy or Science?", organized by Sanjeev Arora Friday Feb 22, 2019. The audience was very diverse, so I focused on the early history and dynamics of ideas in neural..
🌎 The epistemology of Deep Learning"
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Google Doubles Down On Spammers With #TensorFlow. #BigData #Analytics #MachineLearning #DataScience #AI #NLProc #IoT #IIoT #PyTorch #Python #RStats #Java #JavaScript #ReactJS #GoLang #CloudComputing #Serverless #DataScientist #Linux #TransferLearning
🌎 http://bit.ly/2U4ZSAf
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
🌎 http://bit.ly/2U4ZSAf
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Knowing that people judge you by your books I picked out this selection for our new IKEA book shelves. Now I’m just waiting for any statistician to come visit.
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
✴️ @AI_Python_EN
❇️ @AI_Python
🗣 @AI_Python_arXiv
Forwarded from DLeX: AI Python (Farzad🦅🐋🐕🦏🐻)
♨️ Free Self-Study Books on Mathematics, Machine Learning & Deep Learning
🔶1. Matrix Computations
✅ Free Book: Download here
🔶 2. A Probabilistic Theory of Pattern Recognition
💠 Free Book: Download here
✅3. Advanced Engineering Mathematics
💠 Free Book: Download here
✅ 4. Probability and Statistics Cookbook
Free Book: Download here
Machine Learning & Deep Learning Books
➡️ 1. An Introduction to Statistical Learning (with applications in R)
🖇 Free Book: Download here
➡️ 2. Probabilistic Programming and Bayesian Methods for Hackers
👉 Free Book: Download here
➡️3. The Elements of Statistical Learning
👉 Free Book: Download here
➡️4. Bayesian Reasoning and Machine Learning
👉 Free Book: Download here
❇️@AI_Python
❇️@Data_Experts
✔️ 5. Information Theory, Inference, and Learning Algorithms
👉 Free Book: Download here
✔️ 6. Deep Learning
🔗Free Book: Download here
📚 7. Neural Networks and Deep Learning
🔗 Free Book: Download here
📚 8. Supervised Sequence Labelling with Recurrent Neural Networks
🔗Free Book: Download here
📚 9. Reinforcement Learning: An Introduction
🔗 Free Book: Download here
#کتاب #هوش_مصنوعی #یادگیری_عمیق #یادگیری_تقویتی #آموزش #آمار #احتمال
❇️@AI_Python
❇️@Data_Experts
🔶1. Matrix Computations
✅ Free Book: Download here
🔶 2. A Probabilistic Theory of Pattern Recognition
💠 Free Book: Download here
✅3. Advanced Engineering Mathematics
💠 Free Book: Download here
✅ 4. Probability and Statistics Cookbook
Free Book: Download here
Machine Learning & Deep Learning Books
➡️ 1. An Introduction to Statistical Learning (with applications in R)
🖇 Free Book: Download here
➡️ 2. Probabilistic Programming and Bayesian Methods for Hackers
👉 Free Book: Download here
➡️3. The Elements of Statistical Learning
👉 Free Book: Download here
➡️4. Bayesian Reasoning and Machine Learning
👉 Free Book: Download here
❇️@AI_Python
❇️@Data_Experts
✔️ 5. Information Theory, Inference, and Learning Algorithms
👉 Free Book: Download here
✔️ 6. Deep Learning
🔗Free Book: Download here
📚 7. Neural Networks and Deep Learning
🔗 Free Book: Download here
📚 8. Supervised Sequence Labelling with Recurrent Neural Networks
🔗Free Book: Download here
📚 9. Reinforcement Learning: An Introduction
🔗 Free Book: Download here
#کتاب #هوش_مصنوعی #یادگیری_عمیق #یادگیری_تقویتی #آموزش #آمار #احتمال
❇️@AI_Python
❇️@Data_Experts