These are my notes about the training session "Blockchain + Neuronal Networks" we had today.

Blockchain I - Bitcoin Basics

• 18.5.2010 10000 BTC were payed for pizzas (40 USD)
• Virtual cryptocurrencies are controlled by the community:
  • Developer
  • Mining Pools
  • User
  • Service Provider (Exchanges, Payment Processors, Online Markets)
• Currency overview: https://coinmarketcap.com/

Basics Blockchain and Bitcoin

• Authenticated data structure
• Verifiable transaction history (checks whether a transaction can be executed)
• Public Ledger
• Consensus
• Arbitrators (Vermitler): Nodes of an open and asynchronous overlay network with weak identities -> no central party
• Method: Arbitrators add cryptographic authentication markers at "regular" intervals and get rewards-> blocks
• There are wallets, full node clients (which make up the P2P network), and miners.
• A block is a list of transactions (at Bitcoin)
• The Bitcoin hash is 64 Byte big

Transaction

Anyone who has the software for it can perform a transaction. However, it is not yet in the block chain.
The miners take those with the most fees from the list of all transactions first and form the next block.

The price the users are willing to pay, https://jochen-hoenicke.de/queue/#1,24h

Architecture Layer of Blockchain Systems

Pseudonyms -> Application -> State Machine -> Ledger -> Consensus Protocol -> Overlay-Net (Peer to Peer) -> Nodes (weak Identities (IP-hosts))

Properties of Bitcoin

• Based on public protocols/open source software and public key cryptography
• No central instance
• Decentralisation based on P2P network infrastructure

ATMs are machines that change FIAT Money into Bitcoin and vice versa. They are located all over the world, except of some developing countries (like Germany (I'm really disappointed how far we are behind other countries)).

https://coinatmradar.com/ - One can see that Germany is a white space

Mining - Proof of Work

Compute some moderately expensive, but not intractable, function in order to get access to new coins (Naor Dwork)
Challenges in case of Bitcoin:

• Find a new block with a specific count of leading zeros in the block hash
• Required count of leading zeros determines difficulty to new block
  • Cryptographic hash function SHA-256 is used for calculating block hashes
• Bitcoin mining started on CPU, optimizations led to GPU-/FPGA- and ASICs-mining
• Today, only highly optimised ASICS in use

Mining Alternatives

• Proof of Work
• Proof of Space
• Proof of Stake
• Proof of Stake
• Proof of Burn (Burn coins by sending them to unspendable addresses and receive new coins, I.e. XCP)
• Proof of Authority

Bitpay.com allows to pay with bitcoin.

Steem

The teacher allowed me to talk 10 minutes about the Steem-Blockchain and the https://steemit.com platform. (I stood in front of another class (25 people) than tuesday.)
I talked about:

• Proof of brain
• Steem is a bank, where you lend your money and divide your interest to others by giving them upvotes
• Steemit.com is a possibility to earn money by writing posts and articles
• You can earn curation-money by setting your upvotes to good content posts
• There are other platforms, like actifit, where you can earn money by uploading your proof of activity
  
  For more details follow me on https://steemit.com/@achimmertens

Demonstration of Bitcoin:

We got access to a Virtual machine, where we tested to send and receive Bitcoin
http://demo.dence.de/#/

Bitcoin Core Wallet

That's how a Bitcoin Block looks like

Abuse of Virtual Currencies

• Bitcoin is no official money and therefore "cannot be stolen". (There are no laws about it).
• Spy out private keys
• Double Spending
• Double Receiving
• Mining on attacked computer/hardware
• Money Laundering - Bitcoin Mixer

Architecture of DAPPS on Ethereum

DAPP: Distributed Application, based on blockchain
Plugin in browser (i.e. Metamask)
Youtube:

When do you need blockchain: https://medium.com/@sbmeunier/when-do-you-need-blockchain-decision-models-a5c40e7c9ba1

Artificial Intelligence

• Systems, that act like human
  • Turing Test
• Machine seeing
• Systems that think like humans
• Systems that think rationally
• Systems that act rationally
• Pattern recognition: Recurring structures

Machine learning

• Sensor (input) -> Preprocessing -> Characteristic extraction -> Learning (test data) -> Classification
• Semantic analysis of an observation:
• Given:
  • Observation as real representation x
  • Set of meanings (classes)
  • Model of the context

Analyse strategies

Bottom Up

The sensor observes something and tries to assign it to a class. It looks which reference patterns fit best.
Example: Recognition of a digit from 1-9

Top Down

Splitting of the picture into individual parts

Top down and bottom up are often combined.

Statistical detection

Neuronal Nets

With strategic networks one knows how it works, with neural networks not necessarily.

Structure of a neuron

A cell decides whether it forwards information or not, depending on the strength and frequency of the incoming information and a set threshold value.

Decision when to forward an impulse

http://playground.tensorflow.org/ - With this tool on can play and understand how neuronal layers influence the pattern recognition rate
It is interesting to see, that each time one let the software learn to recognize the pattern, we get other results.

Structurs of neuronal networks

• Feed Forward nets
• Recurrent nets

Perceptrons are depicted neurons.
As an example, a list of American universities can be checked to see which criteria all lead to a school being private or not.
There is a AI-programming language called "R".
https://katacoda.com/basiafusinska/courses/r-basics/r-environment
We have seen a test-software, that has analysed a big table of data. Each rowset contains a school and a field "Private". The program observes with two neuronal layers the patterns of the data:

Demonstration of a dataset from American Universities - It is compared, which Universities are private and the KI is trained to find the combination of parameters with 2 layers of neuronal networks

Correlation of data with universities, that are private

Decision Tree, if a University is private or not

Regards, Achim Mertens