ADA coin mining algorithm

1. First of all, what ADA does is the concept of smart contract. I think you should be familiar with smart contract, but for some new people, I still want to explain. The so-called smart contract is to write a contract that the computer can understand through the programming language, because there is no need for a third party, and the executor of the contract is a selfless computer, so it solves the problem of trust between people. For details, please refer to GitHub smart contract
Ethereum is a blockchain smart contract. Now Ethereum ranks 2-3 in the market value ranking, with a market value of 25% of bitcoin. However, Ethereum also has some problems, such as slow trading, high fuel costs and the mining mechanism that has been criticized repeatedly. Familiar? All of bitcoin's problems are with Ethereum. Some time ago, the Ethereum cloud cat game developed on Ethereum went online in the first week, directly blocking Ethereum's trading network. I tried to pay a high gas fee of nearly 100 yuan, but I still couldn't transfer it successfully. This can't help but lower my expectation of Ethereum. Ethereum is a great idea of V God, but after all, it is a pioneer, and there is a long way to go from perfection.

2. Miner is composed of high-end graphics card, a miner to use a lot of high-end graphics card composition. There are a lot of loaders, so the price of graphics card will rise naturally. It's all taken to the loader.

3. You can buy it directly with EBIT. It's very convenient

4. It is to give a certain sample and tell him the expected results. The learning process is too slow by constantly training and modifying weights. In some places, the numerical value changes slowly. There is no theoretical basis for selecting the number of hidden layer nodes. It depends on experience and tends to forget the old learning results

5. AdaBoost algorithm requires weak classifier accuracy & gt; And each weak classifier is independent of each other, but if the error rate of weak classifiers is below 50%, but not completely independent, what kind of consequences will it cause? As discussed with AA today, a weak classifier uses the least square error. The error rate of the least square error should be less than 50%, but adjusting the weight will make the classifier not independent (I don't think so either, because the weak classifier is recalculated after adjusting the weight, and the following time should not be linear correlation. What's more, the description of AdaBoost algorithm itself is also done by adjusting weights. This shows that such weight adjustment will not make weak classifiers relevant. Otherwise, AdaBoost algorithm has such a big loophole that it can't be widely used.) AA says that it will make the error rate of combined classifiers & gt; But I don't think so. The iterative algorithm should be able to make it better and better, but it may not be able to reach the error rate; The state of the value near 0. BTW AdaBoost can't get the result... Crying

6. AdaBoost

AdaBoost algorithm (an algorithm that iterates weak classifiers to proce final strong classifiers) has more definitions & gt& gt;

[network phrase]

AdaBoost, AdaBoost, adaptive enhancement
gentle AdaBoost, gentle AdaBoost
AdaBoost algorithm number recognition

7. You can limit some conditions to rece the extracted features, such as features with small values, and set threshold or subsampling

8. I haven't used it. I think the explanation can be used to do some mathematical modeling

9. There are also many, one-step secant algorithm, quasi Newton algorithm, quantized conjugate gradient method, elastic gradient descent method and so on. The details can be viewed in the help file training function of MATLAB, and the path is: neural network toolbox & gt; Functions> Training functions, you can see the functions and detailed introction of various algorithms

10. These algorithms are derived from the original algorithm
Yitong Wantong