Machine Learning for Physical Sciences

Key Points

Data Sources	Essential libaries for data online data sources Data retrieval from the CDS Toolbox netCDF and GRIB data formats
Data Preprocessing and Visualization
Machine Learning Fundamentals	Data representations are crucial for ML in science, including spatial data (vector, raster), point clouds, time series, graphs, and more ML algorithms like linear regression, k-nearest neighbors,support vector Machine, xgboost and random forests are vital algorithms Supervised learning is a popular ML approach, with decision trees, random forests, and neural networks being widely used Fundamentals of data engineering are crucial for building robust ML pipelines, including data storage, processing, and serving
Deep Learing Fundamentals	Deep Learning algorithms are often represented as graph computation We have different non-linear activation functions that help in learning different relationships to solve handle non-linearity in nn problems.
Recurrent Neural Networks (RNNs) Models	LSTMs and GRUs are advanced RNN architectures designed to handle long-term dependencies in sequential data. The application of deep learning, particularly through RNNs and their variants like LSTM, GRU, and TFT, holds significant promise for time series forecasting in the physical sciences LSTM, GRU and TFT models leverage advanced mechanisms for superior predictive performance in physical sciences applications.
Machine Learning For Material Science	Artificial intelligence and machine learning are being increasingly used in scientific domains such as computational science RDKit can be used to calculate various molecular properties for machine learning in materials science PyMatGen help to query structures from MP repository Descriptor-based ML approaches focus on using numerical features (descriptors) that summarize key aspects of the material Graph-based ML techniques are particularly suitable when dealing with materials that can be represented as networks or graphs In practice, hybrid approaches that combine aspects of both graph-based and descriptor-based methods are also gaining traction, leveraging the strengths of each depending on the task at hand. Crystal Hamiltonian Graph neural Network is pretrained on the GGA/GGA+U static and relaxation trajectories from Materials Project Charge-informed molecular dynamics can be simulated with pretrained CHGNet through ASE python interface CHGNet can perform fast structure optimization and provide site-wise magnetic moments. This makes it ideal for pre-relaxation and MAGMOM initialization in spin-polarized DFT

Glossary

additive color model: A way to represent colors as the sum of contributions from primary colors such as red, green, and blue.
argument: A value given to a function or program when it runs. The term is often used interchangeably (and inconsistently) with parameter.
assertion: An expression which is supposed to be true at a particular point in a program. Programmers typically put assertions in their code to check for errors; if the assertion fails (i.e., if the expression evaluates as false), the program halts and produces an error message. See also: invariant, precondition, postcondition.
assign: To give a value a name by associating a variable with it.
body: (of a function): the statements that are executed when a function runs.
call stack: A data structure inside a running program that keeps track of active function calls.
case-insensitive: Treating text as if upper and lower case characters of the same letter were the same. See also: case-sensitive.
case-sensitive: Treating text as if upper and lower case characters of the same letter are different. See also: case-insensitive.
comment: A remark in a program that is intended to help human readers understand what is going on, but is ignored by the computer. Comments in Python, R, and the Unix shell start with a # character and run to the end of the line; comments in SQL start with --, and other languages have other conventions.
compose: To apply one function to the result of another, such as f(g(x)).
conditional statement: A statement in a program that might or might not be executed depending on whether a test is true or false.
comma-separated values: (CSV) A common textual representation for tables in which the values in each row are separated by commas.
default value: A value to use for a parameter if nothing is specified explicitly.
defensive programming: The practice of writing programs that check their own operation to catch errors as early as possible.
delimiter: A character or characters used to separate individual values, such as the commas between columns in a CSV file.
docstring: Short for “documentation string”, this refers to textual documentation embedded in Python programs. Unlike comments, docstrings are preserved in the running program and can be examined in interactive sessions.
documentation: Human-language text written to explain what software does, how it works, or how to use it.
dotted notation: A two-part notation used in many programming languages in which thing.component refers to the component belonging to thing.
empty string: A character string containing no characters, often thought of as the “zero” of text.
encapsulation: The practice of hiding something’s implementation details so that the rest of a program can worry about what it does rather than how it does it.
floating-point number: A number containing a fractional part and an exponent. See also: integer.
for loop: A loop that is executed once for each value in some kind of set, list, or range. See also: while loop.
function: A group of instructions (i.e., lines of code) that transform some input arguments to some output.
function call: A use of a function in another piece of software.
immutable: Unchangeable. The value of immutable data cannot be altered after it has been created. See also: mutable.
import: To load a library into a program.
in-place operators: An operator such as += that provides a shorthand notation for the common case in which the variable being assigned to is also an operand on the right hand side of the assignment. For example, the statement x += 3 means the same thing as x = x + 3.
index: A subscript that specifies the location of a single value in a collection, such as a single pixel in an image.
inner loop: A loop that is inside another loop. See also: outer loop.
integer: A whole number, such as -12343. See also: floating-point number.
invariant: An expression whose value doesn’t change during the execution of a program, typically used in an assertion. See also: precondition, postcondition.
library: A family of code units (functions, classes, variables) that implement a set of related tasks.
loop variable: The variable that keeps track of the progress of the loop.
member: A variable contained within an object.
method: A function which is tied to a particular object. Each of an object’s methods typically implements one of the things it can do, or one of the questions it can answer.
object: A collection of conceptually related variables (members) and functions using those variables (methods).
outer loop: A loop that contains another loop. See also: inner loop.
parameter: A variable named in the function’s declaration that is used to hold a value passed into the call. The term is often used interchangeably (and inconsistently) with argument.
pipe: A connection from the output of one program to the input of another. When two or more programs are connected in this way, they are called a “pipeline”.
postcondition: A condition that a function (or other block of code) guarantees is true once it has finished running. Postconditions are often represented using assertions.
precondition: A condition that must be true in order for a function (or other block of code) to run correctly.
regression: To re-introduce a bug that was once fixed.
return statement: A statement that causes a function to stop executing and return a value to its caller immediately.
RGB: An additive model that represents colors as combinations of red, green, and blue. Each color’s value is typically in the range 0..255 (i.e., a one-byte integer).
sequence: A collection of information that is presented in a specific order. For example, in Python, a string is a sequence of characters, while a list is a sequence of any variable.
shape: An array’s dimensions, represented as a vector. For example, a 5×3 array’s shape is (5,3).
silent failure: Failing without producing any warning messages. Silent failures are hard to detect and debug.
slice: A regular subsequence of a larger sequence, such as the first five elements or every second element.
stack frame: A data structure that provides storage for a function’s local variables. Each time a function is called, a new stack frame is created and put on the top of the call stack. When the function returns, the stack frame is discarded.
standard input: A process’s default input stream. In interactive command-line applications, it is typically connected to the keyboard; in a pipe, it receives data from the standard output of the preceding process.
standard output: A process’s default output stream. In interactive command-line applications, data sent to standard output is displayed on the screen; in a pipe, it is passed to the standard input of the next process.
string: Short for “character string”, a sequence of zero or more characters.
syntax error: A programming error that occurs when statements are in an order or contain characters not expected by the programming language.
test oracle: A program, device, data set, or human being against which the results of a test can be compared.
test-driven development: The practice of writing unit tests before writing the code they test.
traceback: The sequence of function calls that led to an error.
tuple: An immutable sequence of values.
type: The classification of something in a program (for example, the contents of a variable) as a kind of number (e.g. floating-point, integer), string, or something else.
type of error: Indicates the nature of an error in a program. For example, in Python, an IOError to problems with file input/output. See also: syntax error.
while loop: A loop that keeps executing as long as some condition is true. See also: for loop.