Wednesday 8 January 2020

Machine Learning Applied

Here are examples of some interesting practical applications of Machine Learning.
You might find here an inspiration for some of your future AI projects.

I'll keep this post updated.





Goal: Kapa.ai learns from your technical resources to generate an LLM-powered chatbot that answers developer questions automatically and helps you find gaps in your docs.

Input: technical documentation

Output: computer's sentences in interactive conversation/chat

ML Method used: 
LLM

URL: https://www.kapa.ai/



Goal: chatbots humanizing computer interactions, improving foreign language practice, and making relatable interactive movie and videogame characters.

Input: computer's partner input sentences in interactive conversation/chat

Output: computer's sentences in interactive conversation/chat

ML Method used: 
recurrent neural network

Google AI Blog: Towards a Conversational Agent that Can Chat About…Anything
[1506.05869] A Neural Conversational Model
2001.09977.pdf




Goal: pose estimation


Input: 

Output:

ML Method used:

smellslikeml/ActionAI: custom human activity recognition modules by pose estimation and cascaded inference using sklearn API



Artificial intelligence tracks deadly viruses back to their animal hosts
Machine learning helps to identify the animals that deadly viruses call home
Predicting reservoir hosts and arthropod vectors from evolutionary signatures in RNA virus genomes | Science
OP-VEVO180037 1..12
matthewcotten/cotten_myphan_coronavirus_classification_tool - Docker Hub
Main factors influencing recovery in MERS Co-V patients using machine learning - ScienceDirect

Goal: Detect the source carrier of the virus

Input: virus DNA samples

Output: animal carrier classification

ML Method used: Random Forest




Using machine learning to guide targeted and locally-tailored empiric antibiotic prescribing in a children's hospital in Cambodia

Input: blood culture, clinical, demographic and living condition information (time from admission to blood culture, patient age, hospital versus community-acquired infection, and age-adjusted weight score).

Output: predict Gram stains and whether bacterial pathogens could be treated with one of common empiric antibiotics from a predefined set.

ML Method used: Random Forest



https://www.move-lab.com/
opendatacam/opendatacam: An open source tool to quantify the world
Beat The Traffic X - Trailer on Vimeo
OpenDataCam: An open source tool to quantify the world - NVIDIA Developer Forums

ML Method used: CNN (YOLO, tinyYOLO, YOLOv3)

Technologies used: Docker, Node.JS

Hardware: NVIDIA Jetson (TX2, Nano, Xavier)



Feature extraction and classification of heart sound using 1D convolutional neural networks | EURASIP Journal on Advances in Signal Processing | Full Text

Input:  phonocardiogram (PCG), heart auscultation sound

Output: classification of heart sound signals (normal vs abnormal)

ML Method used: 1-D CNN



(02/01/2019)

AI 'outperforms' doctors diagnosing breast cancer - BBC News

The results showed that the AI model was as good as the current double-reading system of two doctors. And it was actually superior at spotting cancer than a single doctor.



Cambridge Analytica

Cambridge Analytica: the Geotargeting and Emotional Data Mining Scripts
BojanKomazec/cambridgeAnalytica


Input:
  FB users data

Output: Tailored FB adds (political campaign)

ML Method used: Emotional Data Mining



Netflix

75% of Netflix users select films recommended to them by the company’s machine learning algorithms.

source: Roundup Of Machine Learning Forecasts And Market Estimates, 2020

Input:  Films user has been watching

Output:  Films offered to user to watch

ML Method used: Recommender System





No comments: