Risk as a Science of Fluctuation

Loading...

Do curso por New York University Tandon School of Engineering

Overview of Advanced Methods of Reinforcement Learning in Finance

3 classificações

New York University Tandon School of Engineering

Overview of Advanced Methods of Reinforcement Learning in Finance

3 classificações

Curso 4 de 4 em Specialization Machine Learning and Reinforcement Learning in Finance

In the last course of our specialization, Overview of Advanced Methods of Reinforcement Learning in Finance, we will take a deeper look into topics discussed in our third course, Reinforcement Learning in Finance. In particular, we will talk about links between Reinforcement Learning, option pricing and physics, implications of Inverse Reinforcement Learning for modeling market impact and price dynamics, and perception-action cycles in Reinforcement Learning. Finally, we will overview trending and potential applications of Reinforcement Learning for high frequency trading, cryptocurrencies, peer-to-peer lending, and more.

Na lição

Black-Scholes-Merton model, Physics and Reinforcement Learning

Reinforcement Learning and Ptolemy's Epicycles5:12

PDEs in Physics and Finance5:48

Competitive Market Equilibrium Models in Finance5:28

I Certainly Hope You Are Wrong, Herr Professor!7:03

Risk as a Science of Fluctuation3:45

Markets and the Heat Death of the Universe3:37

Conheça os instrutores

Igor Halperin

At this point, I would like to take a short pause and ask a question.

How would Mark Twain approach option-pricing?

What are the wrong words in this list that should be crossed out, should be,

No arbitrage, or maybe it should be the notion of Risk-less hedge, et cetera?

To understand this a bit better,

let's talk again about risk.

Risk is a story about fluctuations,

that in our case will be fluctuations of portfolio value.

One of the simplest measure of risk is variance.

For example, in the Markowitz portfolio theory,

portfolio risk is proportional to the variance of the total portfolio return.

One multiplied by risk aversion parameter lambda,

if user risk penalty,

that is subtracted from expected return to get a risk-adjusted return.

There are other specifications of risk that can be

defined in terms of higher moments of portfolio value distribution.

In physics, fluctuations are known to play a key role in many physical effects,

for example in phase transitions.

There exists equilibrium fluctuations and non-equilibrium fluctuations.

Equilibrium fluctuations describe small deviations of

absorbable thermodynamic quantities such as pressure,

volume or entropy from their constant equilibrium values.

If our system is of infinite volume,

all fluctuations die off,

and thermodynamic absorbables become strictly constant in time, in this limit.

This limit is called the thermodynamic limit,

so fluctuations are Sonic that requires that we work with a finite system,

not an infinite one.

There also known equilibrium fluctuations,

these are fluctuations around the non-equilibrium steady-state over system.

An example of a non-equilibrium system would be an electric circuit,

where electric current running in wires produces heat.

Physics has developed various approaches to deal with fluctuations,

both equilibrium and non-equilibrium ones.

Now, the thing with the Black-Scholes model is that it throws away

fluctuations in the replicating portfolio of an option.

This happens because in the Black-Scholes equation,

we work from the start in continuous time.

Because of that instantaneously,

risk-free hedging becomes mathematically possible.

This means that by continuously hedging,

the Black-Scholes model kills all fluctuations in the option replicating portfolio.

So, by taking a non-physical or better say,

maybe non-financial limit of continuous time,

the Black-Scholes model gets a non-physical result,

that option hedging and hence the option itself is riskless.

This is entirely equivalent to taking the thermodynamic limit in physics.

Instead of taking the volume of a system to infinity,

here we take time steps to zero,

but the net result is the same,

all fluctuations die off in this limit.

Explore nosso catálogo

Registre-se gratuitamente e obtenha recomendações, atualizações e ofertas personalizadas.

O Coursera proporciona acesso universal à melhor educação do mundo fazendo parcerias com as melhores universidades e organizações para oferecer cursos on-line.

© 2018 Coursera Inc. Todos os direitos reservados.

Baixar na App Store

Baixar no Google Play