State–action–reward–state–action

:$Q^\{new\}(S\_t,\; A\_t)\; \backslash leftarrow\; (1\; -\; \backslash alpha)\; Q(S\_t,A\_t)\; +\; \backslash alpha\; \backslash ,\; [R\_\{t+1\}\; +\; \backslash gamma\; \backslash ,\; Q(S\_\{t+1\},\; A\_\{t+1\})]$

A SARSA agent interacts with the environment and updates the policy based on actions taken, hence this is known as an on-policy learning algorithm. The Q value for a state-action is updated by an error, adjusted by the learning rate α. Q values represent the possible reward received in the next time step for taking action a in state s, plus the discounted future reward received from the next state-action observation.

Watkin's Q-learning updates an estimate of the optimal state-action value function $Q^*$ based on the maximum reward of available actions. While SARSA learns the Q values associated with taking the policy it follows itself, Watkin's Q-learning learns the Q values associated with taking the optimal policy while following an exploration/exploitation policy.

Some optimizations of Watkin's Q-learning may be applied to SARSA.{{Cite journal|last1=Wiering|first1=Marco|last2=Schmidhuber|first2=Jürgen|date=1998-10-01|title=Fast Online Q(λ)|url=https://link.springer.com/content/pdf/10.1023%2FA%3A1007562800292.pdf|journal=Machine Learning|language=en|volume=33|issue=1|pages=105–115|doi=10.1023/A:1007562800292|s2cid=8358530|issn=0885-6125|doi-access=free}}

The learning rate determines to what extent newly acquired information overrides old information. A factor of 0 will make the agent not learn anything, while a factor of 1 would make the agent consider only the most recent information.

The discount factor determines the importance of future rewards. A discount factor of 0 makes the agent "opportunistic", or "myopic", e.g.,{{cite web |title=Arguments against myopic training |date=9 July 2020 |url=https://www.lesswrong.com/posts/GqxuDtZvfgL2bEQ5v/arguments-against-myopic-training |access-date=17 May 2023 |language=en}} by only considering current rewards, while a factor approaching 1 will make it strive for a long-term high reward. If the discount factor meets or exceeds 1, the $Q$ values may diverge.

Since SARSA is an iterative algorithm, it implicitly assumes an initial condition before the first update occurs. A high (infinite) initial value, also known as "optimistic initial conditions",{{Cite web|url=http://incompleteideas.net/book/ebook/node21.html|title=2.7 Optimistic Initial Values|website=incompleteideas.net|access-date=2018-02-28}} can encourage exploration: no matter what action takes place, the update rule causes it to have higher values than the other alternative, thus increasing their choice probability. In 2013 it was suggested that the first reward $r$ could be used to reset the initial conditions. According to this idea, the first time an action is taken the reward is used to set the value of $Q$. This allows immediate learning in case of fixed deterministic rewards. This resetting-of-initial-conditions (RIC) approach seems to be consistent with human behavior in repeated binary choice experiments.{{cite journal | last1 = Shteingart | first1 = H | last2 = Neiman | first2 = T | last3 = Loewenstein | first3 = Y | date = May 2013 | title = The Role of First Impression in Operant Learning | journal = J Exp Psychol Gen | volume = 142 | issue = 2| pages = 476–88 | doi = 10.1037/a0029550 | pmid = 22924882 | url = http://ratio.huji.ac.il/sites/default/files/publications/dp626.pdf }}

• Prefrontal cortex basal ganglia working memory
• Sammon mapping
• Constructing skill trees
• Q-learning
• Temporal difference learning
• Reinforcement learning

{{reflist}}

{{Artificial intelligence navbox}}

{{DEFAULTSORT:State-action-reward-state-action}}

Category:Machine learning algorithms