AI & Quantum @aiquan - Tumblr Blog

Blog Moved

(Tumblr's smart search is not working with single word, so hard to search).

Iteration Vars and Count Vars

Letters in alphabet order:

i, j, k, l, m, n

l is confusing with 1, avoid it

Iteration vars (from left to right)

i, j, k

Count vars (from right to left)

n, m, k

5 posts!

Neuralnet Designer's Works

1. Decide compositon of network

Eg. Both Lstm and Convo1D can learn sound, which is better

2. Decide network shape and size

This is hard to estimate, smallest network for fast training n feed

Pros and cons of neuralnet

Pro: Algorithm approximator which gives result fast after 1 feed while classic algos take some orders of n to complete

Con: Not accurate

High Level in AI/ML Carrier

Amateurs

Do basic and standard, well-known solution, neuralnets

Pros

Do with additional new designs of neuralnets to make different solutions

Experts

Can make highly-efficient and powerful neuralnet designs with top output results

In AI/ML carrier, this post can be: Neuralnet Designer

What is Combinatory Model

We've all heard about Boosted Model, it's a bunch of same-design week models to do strong works.

Combinatory Model is similar, but the week models are of different designs.

TensorFlow vs PyTorch

TensorFlow fits with biology and mathematics, both, terming the word Tensor in mathematics.

PyTorch has a wrong design comparing to biology, the activation function. It's at axon, in a neuron, can't be outside of layer.

Artificial Things

AI = Instincts (Genetics) + ML

Those instincts can be done quite in regular algorithms.

ML is done mainly with neuralnet.

AI & Quantum