Connecting #finance and #technology |#MachineLearning #AI in making sense of different data |Valuable decision making is taking over the news and events' speed.
Joined September 2018
- Tweets 612
- Following 340
- Followers 206
- Likes 1,598
40 Photos and videos
Pinned Tweet
17 Jan 2019
We work on #AI #assistants processing #financial, #economical, #news and #events data. Economical behaviour can't always be explained directly. Our fast models work with #explicit, #hidden or #implicit relations and can forecast.
#AI #MachineLearning #money #investments #fintech
8
FinBlock.AI retweeted
18 Oct 2021
Kernels is all you need:
"Kernel methods don’t overfit and do generalize well to test data without any need for regularization, similar to neural networks and contrary to what you’d expect from traditional learning theory."
#AI #MachineLearning
quantamagazine.org/a-new-lin…
5
17
FinBlock.AI retweeted
29 Apr 2021
Nice publication with academic decorations.
But decomposition into components, tricks with kernels always gave very good results. Better and faster with online KPCA tricks.
#AI #MachineLearning
github.com/google-research/g…
1
7
7
29 Apr 2021
RT @AppliedMLdev :
Performers.
Using only linear space and time complexity.
It looks like that's all.
#AI #MachineLearning
arxiv.org/abs/2009.14794
1
FinBlock.AI retweeted
18 Jan 2021
Artificial intelligence is just applied automation, including machine learning techniques
5
18
FinBlock.AI retweeted
18 Jun 2020
According to @lopezdeprado: In 1899 K. Pearson wrote a paper warning that Correlation is not Causation. While most statistics books incorporate this warning in the first chapter, subsequent chapters tend to ignore it. Theories matter.
#investments #finance
Everything is the same, and that looks bad trib.al/EFqqCvh
10
21
FinBlock.AI retweeted
10 Jun 2020
Extrapolating the spectacular performance of GPT3 into the future suggests that the answer to life, the universe and everything is just 4.398 trillion parameters.
64
640
3,739
FinBlock.AI retweeted
3 Jun 2020
Just follow the cash flow.
#investments
3 Jun 2020
A great graph from an article illustrating supply and demand. The main discovery of the last century: the Brownian motion based on the analysis of the diffusion of supply and demand was made in 1900 by the financier Louis Bachelier.
#investments #AI #ML.
smart-lab.ru/mobile/topic/62…
2
6
FinBlock.AI retweeted
3 Jun 2020
The financial analysis, as acknowledged by its author B. Graham, actually became useless in 1975.
The time of quantitative analysis methods had begun. The question is data quality and a return to the origins of game theory and speculation - Louis Bachelier.
#AI #ML #investments
3 Jun 2020
A great graph from an article illustrating supply and demand. The main discovery of the last century: the Brownian motion based on the analysis of the diffusion of supply and demand was made in 1900 by the financier Louis Bachelier.
#investments #AI #ML.
smart-lab.ru/mobile/topic/62…
5
9
FinBlock.AI retweeted
1 Jun 2020
Pure numerical methods are how to enter through the window when there is a door. Well, thanks to statistics, you can be at the wrong time and in the wrong place
1
4
8
10 Apr 2020
RT @Real__Finance : Which side : up or down?
3
6 Apr 2020
RT @AppliedFinMeth: Modern financial markets have turned into isolated tournaments, distant from real processes. Prices will fly due to the slightest sneeze and tweets without regard to macro indicators. Actually about this was said back in 1975 by B.Graham - Buffett's teacher.
5
3 Apr 2020
3 Apr 2020
Well, the classic "buy the rumor, sell the fact" also works well. Trump tweeted, oil flew away. The number of hyperparameters can be reduced - in fact, the vector of the center of the group of words of synonyms or analogies: Trump, help to markets, oil...
x.com/realDonaldTrump/status…
4
FinBlock.AI retweeted
20 Mar 2020
In fact, the hype created and used to enrich.
The Senate’s newest member sold off seven figures’ worth of stock holdings in the days and weeks after a private, all-senators meeting on the novel coronavirus that subsequently hammered U.S. equities.
thedailybeast.com/sen-kelly-…
20 Mar 2020
Text-based misinformation capitalising on coronavirus panic, and the pure velocity at which it spread, has shocked experts. US officials believe the campaign bears hallmarks pointing to Russia or China on.ft.com/2U42y3f
4
7
FinBlock.AI retweeted
19 Mar 2020
Compared to the year 1921, regulators simply learned to mitigate the fall.
19 Mar 2020
This is precisely the weakness of the attempt to analytically or using fitting methods to link the increase in bond yields and the fall in stock prices. It is impossible to know exactly how and who will liquidate the position.
1
4
FinBlock.AI retweeted
19 Mar 2020
There is only one pattern. With the constant release by the American casino of new chips, which are called dollars, the amplitude of fluctuations of everything that is subject to monetary valuation increases. Compare, for example, with the situation between 1921 and 1975.
19 Mar 2020
Question.
Tell me last time stocks & Treasuries sold off as much as today?
Never!
The S&P 500 is down 16% in the last 7 days while 10-year yields more than doubled!
Some risk parity funds are already down 40 % in a month!
Now think of this on a large scale.
2
9
FinBlock.AI retweeted
17 Mar 2020
Sorry, but what is usd?
17 Mar 2020
CME should launch a Toilet Paper Futures contract
5
8
FinBlock.AI retweeted
5 Jan 2020
Green - is a typical overfitting,
black - SVM/SVR.
x.com/aimldeveloper/status/1…
Typical example. No python, C SIMD optimized, 10.000 data points: SVR takes 1 min, kernel regression 1 second, LSTM 1 hour without result. 1.000.000 data points: SVR 1 hour, KR some seconds.
Good luck.
29 Sep 2019
Read about ML in the original. Vapnik and Chervonenskis. They immediately begin with an overfit model theorem by a combination of polynomials. But ML includes methods that exclude such a fit. Yes, if you still need to really overfit in with slightly variable data, then use NN.
2
7
14
FinBlock.AI retweeted
23 Dec 2019
The fact is that the bet is paid while there is uncertainty and there is a desire to pay for viewing the next card. Perhaps this is the most correct way to explain the difference between the analysis of physical processes and cash games. Slightly different rules come into force.
1
2
11
25 Dec 2019
Not only stocks
22 Dec 2019
@TIME: "It's a Matter of When." How Machines Are Taking Over the World's Stock Markets
time.com/5751190/investing-m…
3
7