Scribd
Upload
13 views

Group 3 Report

This document discusses deterministic and probabilistic dynamic programming, covering backward and forward recursive equations. It explains how dynamic programming can be used to break down complex problems into simpler subproblems to maximize reward or minimize cost in deterministic and probabilistic models.

Uploaded by

Aizel Parpado
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
13 views

Group 3 Report

This document discusses deterministic and probabilistic dynamic programming, covering backward and forward recursive equations. It explains how dynamic programming can be used to break down complex problems into simpler subproblems to maximize reward or minimize cost in deterministic and probabilistic models.

Uploaded by

Aizel Parpado
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Formulation of Deterministic and Probabilistic OP:

Backward and Forward Recursive Equations


(Group 3)

Dynamic programming is an optimization technique used to make a sequence of


interrelated decisions to maximize reward or minimize cost. It works by breaking down a
complex problem into simpler sub-problems. Dynamic programming can be formulated
as either deterministic or probabilistic models.

This report will cover:


 Deterministic dynamic programming
 Backward recursive equations
 Forward recursive equations
 Probabilistic dynamic programming
 Backward recursive equations
 Forward recursive equations

Deterministic Dynamic Programming
Deterministic dynamic programming deals with problems where the state transitions
are known and deterministic. In other words, the outcome of each decision is assumed to be
perfectly predictable.

Backward Recursive Equations

The backward recursion approach starts by defining the optimal cost function at the final
state. Then it works backwards to define the cost function at each preceding state, using the
results from later states.

For example, say we had a 5-stage process. We would first define the optimal cost J*(5).
Then use J*(5) to calculate J*(4), and so on back to the first stage J*(1).

Forward Recursive Equations

The forward recursion approach starts from the beginning state and works forwards.
First J*(1) is defined using the initial conditions. Then J*(2) is defined using J*(1), and so on
progressively until the final state J*(5).
So backward recursion works backwards from the final state, while forward
recursion starts from the initial state and works forwards. But both methods yield the
same final results.

Probabilistic Dynamic Programming


In probabilistic dynamic programming, the outcomes of decisions are uncertain.
There is a probability distribution for the transitions between states.

Backward Recursive Equations


The backward equations start by defining the expected value of the optimal cost
function at the final state, taking into account all possible outcomes and probabilities at
the final stage. The result is recursively propagated backwards to preceding states.

Forward Recursive Equations


The forward equations start by defining the probability distribution of being in
each state at stage 1. This distribution is then propagated forwards using the state
transition probabilities to determine the probabilities of being in each state at the next
stage, and so on progressively to the final stage.
The forward equations allow determining the probability of reaching any given
state from the initial conditions. The backward equations allow finding the expected
optimal cost. But both yield equivalent results.

Conclusion
In summary, dynamic programming can be formulated as backward or forward
recursive functions, both for deterministic and probabilistic models. The difference lies in
whether one starts from the final state and works backwards, or starts from the initial
state and works forwards to analyze the optimization problem. But both approaches
yield consistent results.

You might also like