Landscape Similarity and ML Methods

 Natural similarity is everywhere 

In Landscapes , that contain the Sky


The only items that are linear are human made objects



                                     Humans use "our memory" and similarity to recognize nature. 


Many image processing algorithms are biased towards certain features: straight edges or objects associated with humans like faces, eyes, etc.  

Common machine learning algorithms utilize

      1) exact image template matching   



       As described in this article, utilizing OpenCV

        https://towardsdatascience.com/image-matching-with-opencvs-template-matching-5df577a3ce2e 

      2) reduce the image into histograms 

        https://towardsdatascience.com/histograms-in-image-processing-with-skimage-python-be5938962935

      3) generate new features , like Convolution Neural Networks, 

        https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html

      

Memory based theories are associated with how humans are processing information.  "On Intelligence" by Jeff Hawkins is an excellent book describing the brain's structural theory, supporting evidence and applications.   https://en.wikipedia.org/wiki/On_Intelligence 




Comments

Post a Comment

Popular posts from this blog

New Wheels, Old Wheels, Threading the Needle

Image
  Business focal personnel pursue driving efficiencies higher to reduce costs enabling increased marketplace performance.  Technology often helps automate complex processes, reducing costs. Wheels are synonymous with Tech, we can choose to make new wheels, use old wheels or integrate. Threading the Needle requires minimizing the organization delta, using Old Wheels, adding a New Wheel (/ML/Data Science), integration to... Generate growth !!! $100 M plus in 5 Minutes a 10X ROI starting in 2010. Growth shows benefit of New Wheel; New Wheels can be replaced with cheaper Newer Wheels.
Read more

Bahill ( Eye Tracking & Baseball Hall of Fame ), Grad School and Systems Engineering Curricula

Bahill's Lab Research Assistant and Departments System Admin supporting professors research. Bahill is a Systems Control Theory INCOSE fellow http://sysengr.engr.arizona.edu/ https://sie.engineering.arizona.edu/news-events/bahills-baseball-vision-study-still-hit  https://duckduckgo.com/?q=bahill+systems+theory&t=chromentp&ia=web   https://duckduckgo.com/?q=bahill+systems+theory+incose&t=chromentp&ia=web Stark was Bahill's Professor -  https://duckduckgo.com/?q=stark+bahill&t=chromentp&ia=web Together at China Lake they invented night vision capabilities.  In Bahill's lab I supported his eye tracking research group and the optical equipment generating targets, an infrared eye detector glasses, an the analytics packages that recorded the target state space trajectories, the horizontal, vertical and rotational eye movement - state space trajectories, along with the computed differentials between the target and what the eye was looking at.  T...
Read more

Portfolio Derivative Analysis

Image
  Tick data analysis indicating consistent non-linear stock market correlations across multiple days: Gaps indicate opportunity, as stocks are moving opposite -  delta in gap is possible profit  if you short one side and buy high side. Same stock pairs, similar correlation patterns.: Correlated Stocks are chosen by finding most similar profiles in large graphs. This technique can be applied to many different types of fintech systems: stocks, derivatives, crypto. forex and other funds,  Derivatives create co-associated graphs correlated and evolving across time.  In the graphs below the metric is normalized logarithmic (0.0-1.0) generated every three seconds.  The underlying infrastructure combines EMQX, an MQTT broker capable of sustaining 100 M connections fed into a HPC / MPI streaming infrastructure capable of running 100 M active cooperative agent processes coordinated by its own scheduler.  it has been shown to scale up to handle 5000 financial st...
Read more