Real Eyes Realize 101

From Charles Darwin's Theory of Evolution to Zahavi's handicap principle. The essentially ground braking ideas revolve very much beyond the horizon of plain sight. Such is the assumption of my efforts to iterate, reiterate and discover in essence of a format of comprehension, implementation in the form of programming ranging from a mathematical expressions to AGI.

The ideas of designing systems that fall into a corner of class of systems that scratch the surface of systemic awareness and comprehension.

The lens of past in a limited manner does provide the path to imaginary future and its aspects. The proportions may be off and directed by forces not just vast in intensity rather vastness of the their number.

One essence of neural networks arises from utilizing the vastness of the number of forces, or in other words incremental nature of dimensions. The extension of cause and effect playing in cohesion where ample number of forces accumulate to ascertain an affect.

Auto-encoders and the variations of the class are fundamentally a very powerful idea as they maintain the symmetry of a circle, in terms of data life cycles. Their nature of beginning with an element and transforming into components and elemental states to again revert back to its original state is essentially a supra-effective methodology of logic and symmetry and information storage.

The simple metaphors that help me imagine their capabilities are visualization of concentric circles with a common midpoint(head). pragmatically ellipses with focal points.

The other manner to imagine the auto encoders would be stationary waves with varying harmonic states.

courtesy https://www.cyberphysics.co.uk

The value of lossless symmetry can lead to whole class of models addressing a singular objective in n, which can be optimized in different pathways of gradient descent.

Courtesy https://cloud4scieng.org

A mere visual inspections of the above two images spark a thought of structural overlap of Auto-encoders, view with a lens of optics and wave nature could reveal mysteries in the effectiveness and help us design and iterate over a whole plethora of probabilistic solution pathways.

The Variational Autoendoders, GANS and other derivatives have already helped us make powerful strides in Image processing and Natural Language Processing based domains, dealing with homogeneous and heterogeneous data alike. Notably their value in handling heterogeneous knowledge spaces.

They seems to be able to deals all 4 forms of variable . Especially the Unknown Unknowns

1.Known Knowns

2.Known Unknowns

3.Unknown Knowns

4.Unknown Unknowns

Driving back my original point, Machine Learning, Artificial Intelligence techniques at the very essence are knitting together the zones of comprehensible and the obscure. Combined with Physics principles and biological system inspirations. The pace of growth into complex learning systems that tally solutions, their pathways and integrate that which lies boldly concluded across AI systems. The overlap of emotions or its equivalent into the mix of artificial intelligence

We never know what the horizon is hiding in plain sight.