Basics of Python Programming
——————————————-
a. Lists, Tuples, Dictionaries, Conditionals, Loops, etc...
https://lnkd.in/gWRbc3J
b. Data Structures & Algorithms
https://lnkd.in/gYKnJWN
d. NumPy Arrays:
https://lnkd.in/geeFePh
c. Regex:
https://lnkd.in/gzUahNV

Practice Coding Challenges
—————————————
a. Hacker Rank:
https://lnkd.in/gEufBUu
b. Codeacademy:
https://lnkd.in/gGQ7cuv
c. LeetCode:
https://leetcode.com/

Data Manipulation
————————-
a. Pandas:
https://lnkd.in/gxSgfuQ
b. Pandas Cheatsheet:
https://lnkd.in/gfAdcpw
c. SQLAlchemy:
https://lnkd.in/gjvbm7h

Data Visualization
————————
a. Matplotlib:
https://lnkd.in/g_3fx_6
b. Seaborn:
https://lnkd.in/gih7hqz
c. Plotly:
https://lnkd.in/gBYBMXc
d. Python Graph Gallery:
https://lnkd.in/gdGe-ef

Machine Learning / Deep Learning
————————————————
a. Skcikit-Learn Tutorial:
https://lnkd.in/gT5nNwS
b. Deep Learning Tutorial:
https://lnkd.in/gHKWM5m
c. Kaggle Kernels:
https://lnkd.in/e_VcNpk
d. Kaggle Competitions:
https://lnkd.in/epb9c8N

Has Area Under the ROC Curve (AUC-ROC) become Data Science & AI/ML community’s P-Value?

Just returned from day 1 of Intelligent Health AI conference - and while there were some great speakers & talks - one thing stood out. Of the multiple talks reporting machine learning model performance, all except one talk reported AUC-ROC as the only metric - even for unbalanced datasets. It appears that the AUC-ROC metric is being misused similar to how the P-value has been misused & misinterpreted.

There is more to model evaluation than a single number. In addition to AUC-ROC, we have the Precision-Recall (PR) curve, Sensitivity (Recall), Specificity, F1-score, Positive/Negative Predictive Values, Matthews Correlation Coefficient, Calibration, and many other metrics. The graphic below presents a good summary of the various model performance / evaluation metrics (see articles & book for more details):

Regression Metrics:
https://lnkd.in/eRWvRVc

Classification Metrics:
https://lnkd.in/dpYnvGh

Evaluating Machine Learning Models (open-access book):
https://lnkd.in/dHcfZdP

Meta-Learning with Implicit Gradients

Aravind Rajeswaran, Chelsea Finn, Sham Kakade, Sergey Levine : https://lnkd.in/g9H6mZ2

#MachineLearning #ArtificialIntelligence #Optimization #Control #MetaLearning

Awesome victory for #DeepLearning 👏🏻

GE Healthcare wins FDA clearance for #algorithms to spot type of collapsed lung!

Here’s how the AI algorithm works
————————————————
1. A patient image scanned on a device is automatically searched for pneumothorax.
2. If pneumothorax is suspected, an alert with the original chest X-ray, is sent to the radiologist to review.
3. That technologist would also receive an on-device notification to highlight prioritized cases.
4. Algorithms would then analyze and flag protocol and field of view errors and auto rotate images on device.

Article is here:
https://lnkd.in/daNYHfP

#machinelearning

Video of the ACM Tech Talk webinar I gave on 2018/07/11.

ACM says this is one of the most popular Tech Talks ever.

https://youtu.be/zikdDOzOpxY

Hierarchical Decision Making by Generating and Following Natural Language Instructions

“Experiments show that models using natural language as a latent variable significantly outperform models that directly imitate human actions.”

https://arxiv.org/abs/1906.00744

Counterfactual Story Reasoning and Generation”, presents the TimeTravel dataset that tests causal reasoning capabilities over natural language narratives.

Paper:
https://arxiv.org/abs/1909.04076
Code+Data:
https://github.com/qkaren/Counterfactual-StoryRW

What Kind of Language Is Hard to Language-Model?

Mielke et al.: https://lnkd.in/eDUGmse

#ArtificialIntelligence #MachineLearning #NLP

CvxNets: Learnable Convex Decomposition

Boyang Deng, Kyle Genova, Soroosh Yazdani, Sofien Bouaziz, Geoffrey Hinton, Andrea Tagliasacchi : https://lnkd.in/eGUqxjz

Facebook Research at Interspeech 2019

https://ai.facebook.com/blog/facebook-research-at-interspeech-2019/

Sequence-to-Sequence Speech Recognition with Time-Depth Separable Convolutions

https://research.fb.com/publications/sequence-to-sequence-speech-recognition-with-time-depth-separable-convolutions/

Unsupervised Singing Voice Conversion

https://research.fb.com/publications/unsupervised-singing-voice-conversion/

The largest publicly available language model: CTRL has 1.6B parameters and can be guided by control codes for style, content, and task-specific behavior.


code: https://github.com/salesforce/ctrl

article: https://einstein.ai/presentations/ctrl.pdf

https://blog.einstein.ai/introducing-a-conditional-transformer-language-model-for-controllable-generation/

What makes a good conversation?
How controllable attributes affect human judgments

A great post on conversation scoring.

Link:
http://www.abigailsee.com/2019/08/13/what-makes-a-good-conversation.html
Paper:
https://www.aclweb.org/anthology/N19-1170

#NLP #NLU #DL

❇️ @ai_python_en

🔹 Python examples of popular machine learning algorithms with interactive Jupyter demos and math being explained👌

📌 https://github.com/trekhleb/homemade-machine-learning

❇️ @AI_Python

NVIDIA Announces TensorRT 6; Breaks 10 millisecond barrier for BERT-Large

https://news.developer.nvidia.com/tensorrt6-breaks-bert-record/

🔍 DeepPavlov: An open-source library for end-to-end dialogue systems and chatbots

article: https://medium.com/tensorflow/deeppavlov-an-open-source-library-for-end-to-end-dialog-systems-and-chatbots-31cf26849e37

research: https://colab.research.google.com/github/deepmipt/dp_notebooks/blob/master/DP_tf.ipynb

code: https://github.com/deepmipt/DeepPavlov

⭐️Fine-Tuning GPT-2 from Human Preferences

#OpenAI team fine-tuned 774M parameters model to achieve better scores in #summarization and stylistic text continuation in terms of human understanding.

Article definately worths reading (approx 15 min.) with Challenges and lessons learned section and examples.

Link: https://openai.com/blog/fine-tuning-gpt-2/
Paper: https://arxiv.org/abs/1909.08593
Code: https://github.com/openai/lm-human-preferences

#NLP #NLU #finetuning

🗣 Using AI-generated questions to train NLP systems

https://ai.facebook.com/blog/research-in-brief-unsupervised-question-answering-by-cloze-translation/

code:
https://github.com/facebookresearch/UnsupervisedQA

paper:
https://research.fb.com/publications/unsupervised-question-answering-by-cloze-translation/

Neural networks in NLP are vulnerable to adversarially crafted inputs.

We show that they can be trained to become certifiably robust against input perturbations such as typos and synonym substitution in text classification:

https://arxiv.org/abs/1909.01492

Depth Hints are complementary depth suggestions which improve monocular depth estimation algorithms trained from stereo pairs

code:
https://github.com/nianticlabs/depth-hints

paper:
https://arxiv.org/abs/1909.09051

dataset :
https://lmb.informatik.uni-freiburg.de/resources/datasets/SceneFlowDatasets.en.html