martes, 26 de mayo de 2015

Plans to make new market simulator.

Some ideas to write a new agent-based market simulator in ANSI C. This would be acontinuation of my previous market simulator.







Notes on approach for new forex predictor

Use frequency of specific bitonic patterns because in forex, algorithmic trading is predominant. Forex bots watch for rise and falls, so it would make sense that there should be a regular, somewhat predictable frequency, for specific bitonic sections of the wave.



domingo, 24 de mayo de 2015

Designing an oscilloscope

Started with initial design to communicate with RAM using PIC interface. Work further in interface design.




domingo, 17 de mayo de 2015

Reinventing the wheel

Turns out my "new Huffman" is just the good old Sannon-Fano. Maybe there is a way to disturb a Shannon-Fano tree to Huffmanize it. Anyway: Shannon-Fano seems easy to do in parallel. Consider writing article about that instead.

More notes about new Huffman encoding algorithm

Some additional notes on the new Huffman encoding algorithm. The algorithm could be used to parallelize Huffman encoding of large alphabets because there is no dependency between left and right subtrees. At any given level of the full binary tree, all subtrees could be calculated at the same time using SUM(p), which is easily calculated in O(n).









Notes 4 - 14

Ideas for a new algorithm to produce Huffman coding. The secret is in recursively generating the full binary tree by building the left and right subtrees for recursive halves of the sorted probability distribution of the symbols.

Each subtree should have an amount of nodes proportional to the probability sum of it's corresponding half of the probability distribution.

1. Proportion subtree nodes to halves of the probability distribution.
2. Proportion subsubtrees to halves' halves.
3. Repeat until done.












sábado, 16 de mayo de 2015

Notes 1, 2, 3


Check to see if there is a continuity in the distribution of peaks sizes in FOREX time series. Knowing WHEN a specific peak size will happen is trickier. Borwnian motion?