Machine learning in trading: theory, models, practice and algo-trading - page 2969

[mytarmailS]

Maxim Dmitrievsky #:

I started writing code in Python right away, without almost learning it. Then a couple of basic libraries were needed and that was it.

Well, that's fine, it was designed that way....

But why these passive-aggressive attacks on R all the time? It's the best language for its tasks, that's what it was created for,

Python is a general language with a pretence of being easy to learn...

The main thing is not that, but to make beautiful deals with the help of algorithms.

[Maxim Kuznetsov]

mytarmailS #:

because that's not what it's for.

Why doesn't anyone write websites in C++?

Each language has its own task, that's why there are so many of them.

I just chose what was more familiar, more similar to mql, and what was less familiar (not familiar) was called jerky. And you should have studied...

you probably just don't know enough about it...C++ backend is in high demand and it's almost the most expensive jobs.

[mytarmailS]

Maxim Kuznetsov #:

you're probably not fully aware...C++ backend is in high demand and it's almost the most expensive jobs.

you probably just don't fully realise that a website is not only backend ;)

[Renat Fatkhullin]

Not everyone follows the development of MQL5, but it has standard methods for matrices and vectors.

These are the standard methods of the language, not third-party libraries. With the introduction of matrix/vector/complex types, the language has become very powerful for stat analysis and heavy mathematics.

Function	Action	Category
Activation	Calculates the values of the activation function and writes to the passed vector/matrix	Machine learning
ArgMax	Returns the index of the maximum value	Statistics
ArgMin	Returns the index of the minimum value	Statistics
ArgSort	Returns the sorted index	Manipulations
Assign	Copies a matrix, vector or array with automatic conversion	Initialisation
Average	Calculates a weighted average of matrix/vector values.	Statistics
Cholesky	Calculates Cholesky decomposition	Transformations
Clip	Restricts matrix/vector elements to a specified range of acceptable values	Manipulations
Col	Returns a column vector. Writes the vector to the specified column	Manipulations
Cols	Returns the number of columns in the matrix	Characteristics
Compare	Compares elements of two matrices/vectors with specified accuracy	Manipulations
CompareByDigits	Compares elements of two matrices/vectors for matching with precision of significant digits	Manipulations
Cond	Calculates the conditional number of a matrix	Characteristics
Convolve	Returns a discrete linear convolution of two vectors	Derivatives
Copy	Returns a copy of a given matrix/vector	Manipulations
CopyRates	Gets the historical series of the MqlRates structure of the specified symbol-period in the specified amount into a matrix or vector	Initialisation
CopyTicks	Gets ticks from MqlTick structure into a matrix or vector	Initialisation
CopyTicksRange	Gets to a matrix or vector of ticks from the MqlTick structure in the specified date range.	Initialisation
CorrCoef	Calculates Pearson correlation coefficient (linear correlation coefficient)	Derivatives
Correlate	Calculates the cross-correlation of two vectors	Derivatives
Cov	Calculates the covariance matrix	Products
CumProd	Returns the cumulative product of matrix/vector elements, including elements along the given axis.	Statistics
CumSum	Returns the cumulative sum of matrix/vector elements, including elements along the given axis	Statistics
Derivative	Calculates the values of the derivative of the activation function and writes to the passed vector/matrix	Machine Learning
Det	Computes the determinant of a square nondegenerate matrix	Characteristics
Diag	Extracts a diagonal or constructs a diagonal matrix	Manipulations
Dot	Scalar product of two vectors	Derivatives
Eig	Calculates eigenvalues and right eigenvectors of a square matrix	Transformations
EigVals	Calculates eigenvalues of a general matrix	Transformations
Eye	Returns a matrix with ones on the diagonal and zeros elsewhere	Initialisation
Fill	Fills an existing matrix or vector with a given value	Initialisation
Flat	Allows a matrix element to be accessed using a single index instead of two indexes	Manipulations
Full	Creates and returns a new matrix filled with the specified value.	Initialisation
GeMM	General Matrix Multiply of two matrices (General Matrix Multiply)	Products
Hsplit	Horizontal splitting of a matrix into several submatrices. Same as Split with axis=0.	Manipulations
Identity	Creates a single matrix of the specified size	Initialisation
Init	Initialises a matrix or vector	Initialisation
Inner	Inner product of two matrices	Derivatives
Inv	Computes the (multiplicative) inverse of a square nondegenerate matrix using the Jordaan-Gauss method	Solutions
Kron	Returns the Kronecker product of two matrices, a matrix and a vector, a vector and a matrix, or two vectors	Products
Loss	Computes the values of the loss function and writes to the vector/matrix passed in	Machine learning
LstSq	Returns the least squares solution of linear algebraic equations (for non-square or degenerate matrices)	Solutions
LU	LU factorisation of a matrix as the product of a lower triangular matrix and an upper triangular matrix	Transformations
LUP	LUP factorisation with partial permutation, which refers to the LU decomposition with row permutation only: PA=LU	Transformations
MatMul	Matrix product of two matrices	Derivatives
Max	Returns the maximum value in a matrix/vector	Statistics
Mean	Calculates the arithmetic mean of element values	Statistics
Median	Calculates median of matrix/vector elements	Statistics
Min	Returns the minimum value in the matrix/vector	Statistics
Norm	Returns the norm of the matrix or vector	Characteristics
Ones	Creates and returns a new matrix filled with ones	Initialisation
Outer	Calculates the outer product of two matrices or two vectors	Products
Percentile	Returns the specified percentile of the matrix/vector elements or elements along the specified axis.	Statistics
PInv	Computes a pseudo-inverse matrix using the Moore-Penrose method	Solutions
Power	Elevates a square matrix to integer degree	Products
Prod	Returns the product of matrix/vector elements, which can also be performed for a given axis	Statistics
Ptp	Returns the range of matrix/vector values or the given matrix axis	Statistics
QR	Calculates the qr factorisation of a matrix	Transformations
Quantile	Returns the specified quantile of matrix/vector element values or elements along the specified axis	Statistics
Rank	Returns the rank of the matrix using the Gaussian method	Characteristics
RegressionMetric	Calculates the regression metric as the error of deviation from the regression line drawn on the specified data set	Statistics
Reshape	Changes the shape of a matrix without changing its data	Manipulations
Resize	Returns a new matrix with changed shape and size	Manipulations
Row	Returns a vector row. Writes the vector to the specified row	Manipulations
Rows	Returns the number of rows in the matrix	Characteristics
Size	Returns the size of the vector	Characteristics
SLogDet	Calculates sign and logarithm of matrix determinant	Characteristics
Solve	Solves a linear matrix equation or a system of linear algebraic equations	Solutions
Sort	Sort by location	Manipulations
Spectrum	Computes the spectrum of a matrix as the set of its eigenvalues from the product AT*A	Characteristics
Split	Split a matrix into several submatrices	Manipulations
Std	Returns the standard deviation of matrix/vector element values or elements along a given axis.	Statistics
Sum	Returns the sum of matrix/vector elements that can also be performed for the given axis(es)	Statistics
SVD	Singular Value Decomposition	Transformations
SwapCols	Swaps columns in a matrix	Manipulations
SwapRows	Swaps rows in a matrix	Manipulations
Trace	Returns the sum of the diagonals of the matrix	Characteristics
Transpose	Transpose (swaps axes) and returns the modified matrix	Manipulations
Tri	Constructs a matrix with ones on the given diagonal and below and zeros elsewhere.	Initialisation
TriL	Returns a copy of the matrix with zeroed elements over the kth diagonal. Lower triangular matrix	Manipulations
TriU	Returns a copy of the matrix with zeroed elements below the k-th diagonal. Upper triangular matrix	Manipulations
Var	Calculates the variance of the matrix/vector element values	Statistics
Vsplit	Vertical splitting of a matrix into several submatrices. Same as Split with axis=1	Manipulations
Zeros	Creates and returns a new matrix filled with zeros	Initialisation

[Maxim Dmitrievsky]

mytarmailS #:

Well great, that's the way it was intended.....

But why these passive aggressive attacks on R all the time? It's the best language for its tasks, that's what it was designed for,

Python is a generic language with the pretence of being easy to learn...

The main thing is not that, but to make beautiful deals with the help of algorithms.

I don't need your deals, I need backtests!

Trading is not about science, it's more situational. Today's trades are +, tomorrow's are -.

There was a normal arbitrage when you could buy flowers for women and cars for yourself, and then somehow it was deflated. And no matter what statistics you use, Mikola won't make a stone flower.

Another problem with neurons is the choice of models. One has been working for a year, the others are blown away. Then some Egyptian yells at you that you cheated him. What are you gonna do? That's the nature of things. And the statistics somehow pass by again. So the choice was obvious - for maximum simplicity and convenience.

[mytarmailS]

Maxim Dmitrievsky #:
I don't need your deals, I need your backtests!

Trading is not about science, it's more situational. Today's trades are up, tomorrow's

With all my knowledge, I don't know how to algorithmise it, I just have an understanding of the situation and that's all...

The MO is like an indicator, and 99% of the intellectual robot behind me...

1) Either it's like this and it's working.

2) or everything is automatic and it never works.

for now I'm sitting on 1) but dreaming about 2)

[Maxim Dmitrievsky]

mytarmailS #:

For all my knowledge, I don't know how to algorithmise it, I just have an understanding of the situation and that's it....

IO is like an indicator, and 99% of the intelligent robot is behind me...

1) It's either that, and it works.

2) Or it's all on automatic and it never works.

I'm sitting on 1) but I'm dreaming of 2).

You figure out how to validate your FF crafting better and it'll be automatic. It's a cool idea.

[mytarmailS]

Maxim Dmitrievsky #:
You figure out how to validate your FF craft better and it'll be automatic. That's a cool idea.

You just validate it like a normal algorithm.

I'm not talking about it as a discovery, I invented it more than a year ago...

The man asked how to train AMO for profit, I just showed him how...

[Maxim Dmitrievsky]

mytarmailS #:

you just validate it like a normal algorithm.

I'm not talking about it as a discovery, I invented this thing more than a year ago....

The man asked how to train AMO for profit, I just showed you how.

It's not gonna work like a regular one. You need some tricky auto-tuning for unknown data, screening out noise signals, etc. to do about the same thing as a human.

[СанСаныч Фоменко]

mytarmailS #:

The man asked me how to train AMO for profit, I just showed him how.

I can not understand what, but something protests in me against your idea with a target in the form of a balance.

Everything is very similar to trading on history: here bought, and here sold ... and you sit there all smart and rich.

And then you go to real trading, buy, and the market turns around and a hundred pips in the other direction. This is exactly what I observe in my TS. There are few such cases, no more than 10 per cent of all, but everything goes under the tail.

It follows from your idea that you can bypass, i.e. actually predict strong market movements at the expense of the penalty, but it is impossible, because strong movements are the basis of non-stationarity of financial markets.

PS.

The mentioned zigzag is the same balance, but marked into longs and shorts.