quantdev.blog

Hi there! Welcome to my blog. I’m Quasar. I am a software engineer turned quantitative analyst.

Here, I’ll document my coding conquests, exploring practical modern C++ and financial math, building practical and impractical tools, to playing around with numerics!

C++ Type erasure

Once you instantiate a std::function object, how is it, that you are able to stick objects of different actual…

Implementing `string`

Compiler Explorer

Implementing LRU Cache

A least recently used(LRU) cache is a fixed-size cache that behaves just like a regular lookup table, but remembers the order in which the elements are accessed. Once it’s user defined capacity…

How libstdc++ `std::unordered_map` is implemented?

We all love maps. A map is simply a data-structure that allows you to associate a key with some kind of value. They are fast and help to solve a large number of problems.…

Implementing `vector<T>`

In this blog post, we will write a naive vector<T> training implementation. Coding up these training implementations, handling corner cases, getting your code reviewed, revisiting your design is very effective at…

Lambda Functions

I’ve got my hands on the C++ lambda story by Bartłomiej Filipek and wanted to try out a few of the examples.

Floating-point numbers

Computer memory has a finite capacity. Real numbers in \(\mathbb{R}\), do not, in…

Objects, Pointers and References

An object is something that has a lifetime and occupies storage. Even a humble int is an object, but a function is not.

The C++ casts

Traditionally, C++ has supported four ways to perform those explicit type conversions we call casts - static_cast, dynamic_cast, const_cast and reinterprete_cast. C++11…

Lock-free SPSC Queue

I would like to present my implementation for an SPSC lock-free queue in this blog-post.

Pymalloc - Down the rabit hole

Download the CPython source and build it in the debug mode to step into the interpreter’s internals.

Generators, iterators and asynchronous programming

A generator is a function that uses the yield keyword to return an item. When a generator function is…

Pythonic code

Context managers are a very useful feature in Python. Most of the time, we see contet managers around resource management. For example, in situations, when we open files, we…

Creating a tiny C++ Task Library

A future is an object that represents some undetermined result of a task that will be completed sometime in the future. A promise is the provider of that result.

A crash course in Rust - I

I have been working through the excellent Learn Rust With Entirely Too Many Linked Lists a…

Memory Barriers

Atomic operations are indivisible. An atomic operation is any operation that is guaranteed to execute as a single transaction. At a low-level…

Coroutines

You’ve likely heard about this new C++20 feature, coroutines. I think that this is a really important subject and there are several cool use-cases for coroutines. A coroutine in the simplest terms is just a function that you can pause in the middle. Imagine that you are executing a function top-down and then just at some point in between, you’d…

AAD(Adjoint Algorithmic Differentiation)

I was skimming through Luca Capriotti and Mike Giles’ paper 15 Years of Adjoint Algorithmic Differentiation in Finance and wanted to write a toy implementation in Julia. Algorithmic differentiation (AD) is a set of techniques to accurately and efficiently compute derivatives of a function…

Continuous probability puzzles

Example 1 What is the maximum possible variance of a random variable that takes the values in the set \([-1,1]\)?

Discrete probability and counting puzzles

Example 1 We flip a fair coin until we obtain our first heads. If the first heads occurs on the \(k\)-th flip, we are given \(k\) balls. We put them into 3 bins labeled 1, 2, and 3 at random. Find the…

Uniformly sampling from the unit disk

I was asked this puzzle in a recent quant…

Gaussian Processes

Consider a probability space \((\Omega,\mathcal{F},\mathbb{P})\). We can define several random variables on \(\Omega\). A \(n\)-tuple of random variables on this…

A short note on the Dupire PDE

If the Black-Scholes model were good, the implied volatility \(\hat{\sigma}\) parameter would be the same for all call…

Challenging exercises in Template Metaprogramming

For type metaprogramming, the central data structures are the typelist and std::index_sequence. A TypeList is just a static list of types. A std::index_sequence is an integer sequence known…

deducing `this`

Member functions can be overloaded by cv-qualifiers and reference qualifiers & (ref) and && (ref-ref).

Storage Durations in C++

Two key properties of an object in C++ are storage and linkage.

The `virtual` keyword

The virtual keyword specifies that a non-static member function is virtual and supports dynamic dispatch. It may only appear in the…

Quanto Options

A quanto option is…

Custom iterators and Iterator concepts

An iterator is a generalization of a pointer. C++ STL containers usually expose iterators as part of their…

A note on `make_shared<T>(Args&&...)` and `make_unique<T>(Args&&...)`

Since C++14, unique_ptr<T> has been accpompanied by the factory function make_unique<T>(Args&&...) that perfectly forwards its arguments to the constructor of T. Why standard library implementors provide a separate…

`unique_ptr` - A custom implementation

In this post, I try to write a simple homegrown version of std::unique_ptr<T>. This post is partly inspired by the fantastic book C++ Memory Management by Patrice Roy. Tghe toy examples in this book are very instructive and I highly reckon you order a copy. Our goal is just to build intuition for the kind of code required to write such a…

Implementing `shared_ptr<T>`

shared_ptr<T> is tricky to implement, since it is a wrapper over raw underlying pointer of type T* and a reference counter.

`numpy` and `pandas` CheatSheet

np.arange(start, stop, step) returns evenly spaced values in a given interval.

DIY asyncio

In single-core processors, the machine can only perform one task at a time, but can…

A thread-safe queue implementation

In the producer-consumer problem, we have two classes of threads, producers and consumers and a buffer containing a fixed number of slots. A producer thread attempts to put something into the next empty…

Thread-Safe Stack Implementation

We can use C++ synchronization primitives to implement a basic thread-safe stack and queue.

Margrabe’s formula

Let \(S_1(t)\) and \(S_2(t)\) denote the prices of two risky assets which have dynamics:

Levenberg-Marquardt Algorithm

The Levenberg-Marquardt(LM) method consists of an iterative least-squares minimization of a function based on a modification of the Newton method. It’s a super-intuitive…

C++ Ranges

A range is a programmatic abstraction for a sequence of elements, bounded by two iterators(one to the first element of the sequence and one to the last element).

Collateralized Discounting

In the past, standard derivatives pricing theory assumed the existence of a risk-free rate for derivatives discounting. Until the…

IRS, Caps, Floors and Swaptions

I review here a few basic definitions relevant to the interest-rate world.

CRTP(Curiously recurring template pattern)

CRTP consists in

Rule of Five

The rule-of-five states that, if a class C has one of user-declared

Currying and partial function application

Let \(u=f(x_1,\ldots,x_n)\) be a function of \(n\) variables. curry<n>(f) is an operator form of the function \(f\), such that:

Copy-and-swap idiom

The canonical way to write the copy assignment operator is the following:

SFINAE

When we write templates, we sometimes need to restrict the template arguments. For instance, we have a function that should work for any…

C++20 concepts

A class template, function template (including lambdas) may be associated with a constraint, which specifies requirements on the template arguments. This…

A gentle introduction to the Girsanov Theorem - Back to the basics

One of the most popular technical tools in…

Fun with numeraires!

A proficiency in the change-of-measure technique is useful to the working quant. An excellent summary of the important results is the note Girsanov, Numeraires and all that, by Andrew Lesniewski. In this post, I would like to derive relevant results and then we can enjoy pricing some payoffs…

Class Template Argument Deduction(CTAD)

Type Traits 101

Meta-programs are programs that treat other programs as data. They could be other programs or itself. A meta-function is not a function, but a class or struct. Metafunctions are not part of the language and have no formal language support. They exist purely as an idiomatic use of the existing language features. Now, since their use is not…

Tridiagonal Systems

The special case of a system of linear equations that is tridiagonal, that is, has non-zero elements…

Interpolation and Approximation

In this blog post, I would like to implement some interpolation algorithms using modern C++. It’s important we understand how and why these algorithms work. Because there are nuances, exceptions and corner-cases, we need to understand. It’s a great learning experience!…

Numerical Integration

We are interested in an approximate calculation of the definite integral

Template programming

C++11 introduced variadic templates which permit functions to accept a variable number of arguments. They also permit template types such as std::tuple that can hold a variable number of elements. The main language mechanism…

A hitchhiker’s guide to move semantics and perfect forwarding

To understand the basic principles of move semantics, let’s look at the execution of a small piece of code. I’ve written a toy Vector class. I choose to…

Norms

Consider geometric vectors \(\mathbf{x}, \mathbf{y} \in \mathbf{R}^2\). The scalar product(dot-product) of these two vectors is defined by:

Singular Value Decomposition(SVD)

Rectangular matrices do not have eigenvalues. However, we might look at the eigenvalues of the symmetric, positive semidefinite square Gram matrix \(K=AA^T\). Perhaps…

Eigenthingies and Diagonalizability

Each square matrix possesses a collection of one or more complex scalars, called eigenvalues and associated vectors called eigenvectors. A matrix is a concrete realization of a linear transformation on a vector space. The eigenvectors indicate the directions of pure stretch and the…

The Spectral Theorem

Every real, symmetric matrix is orthogonally diagonalizable.

Ito Calculus

Exercise 1 (A strange martingale) Let \((B_t,t\geq 0)\) be a standard Brownian motion. Consider the process:

Martingales

In elementary probability, the conditional expectation of a variable \(Y\) given another random variable \(X\) refers to the expectation of \(Y\) give…

The Markov Property

Let’s start by exhibiting the Markov property of Brownian motion. To see this, consider \((\mathcal{F}_t,t\geq 0)\), the natural filtration of the Brownian motion \((B…

Ito Processes and Stochastic Differential Equations

Let’s start with the definition of Ito processes.

Multivariate Ito Calculus

We can generalize the theory to functions of several brownian motions. This unleashes the full power of Ito calculus.

The distribution of the first passage time of Brownian Motion

The distribution of Brownian motion enjoys many interesting symmetries. The reflection of a Brownian motion about any time \(s\) is also a Brownian motion.

Borel-Cantelli Lemmas

Borel-Cantelli Lemmas.

Positive Definiteness

Definition 1. A real-valued symmetric matrix \(A\) is positive-definite (written as \(A \succ 0\)), if for every real valued vector \(\mathbf{x}…

Optimization Algorithms

Machine Learning

Definition. Let \(f:\mathbf{R}^n \to \mathbf{R}\) be a scalar-valued function. The gradient vector of \(f\) is defined as:

Backpropogation

Machine Learning

With a randomly initialized model, or even a model initialized with more sophisticated approaches, our goal is to train, or…

Coding a neural network layer

Machine Learning

In 1943, McCulloch and Pitts introduced artificial intelligence to the world. Their idea was…

Classification Algorithms

Machine Learning

Let’s say that you have input data \(\mathcal{D}=\{(\mathbf{x}_i,y_i): i=1,2,\ldots,N\}\), and suppose that \(y_i \in \{0,1\}\). Each input has \(D\)-features, so \(\mathbf{x}_i \in…

Derivation of the Least Squares Estimate Beta in Linear Regression

Machine Learning

The following post is going to derive the least squares estimate of the coefficients of linear regression. Our data consists of \(p\) predictors or features \(X_1,\ldots,X_p\) and a…

Standard CDS Pricing Theory

I summarize below standard CDS pricing formulas.

Cox-Ingersoll-Ross (CIR) model

The short rate under the CIR model has the dynamics:

Black Scholes Formula for a European Call

The mean rate of growth of all assets under the risk-neutral measure \(\mathbb{Q}\) is risk-free rate \(r\).

Exploring Option Greeks

I derived the Black-Scholes formula for European style vanilla FX options in a previous post here. The Black-Scholes model \(Bl(S_0,K,T,r_{DOM},r_{FOR},\sigma)\) equipped with a single flat…

Properties of Brownian Motion

Let \(B(t)\) be a fixed Brownian motion. We give below some simple properties that follow directly from the definition of the Brownian Motion.

Implementing Vanna Volga

It is possible to calculate analytically the values of…