Prospect utility with hyperbolic tangent function
Bibliographic record
Abstract
One branch of safety in reinforcement learning is through integrating risk sensitivity within the Markov Decision Process framework. The objective is to mitigate low-probability events that could lead to severe negative outcomes. Eliminating such risky events is usually done by incorporating a utility function on the expected return; therefore, reshaping the reward structures according to the risk levels associated with different outcomes. The temporal difference learning algorithm can be modified with a utility to capture risk. Notably, such utility functions are either convex or concave depending on the desired risk behavior. Given the outcome space and depending on the risk-sensitivity mode, concave utilities may promote risk-averse behavior and convex utilities may encourage risk-seeking strategies. Such function structure is demonstrated in Prospect Theory, and this motivates a novel formulation using the hyperbolic tangent function called PTanh. Using PTanh, experiments are performed to assess the effect of the diminishing marginal property on the risk-averse policies. It is concluded that there is a correlation between the marginal and selecting the risk-averse parameters. The marginals influence the effectiveness of the averse policies. When the marginals are considered, PTanh can demonstrate better results in terms of a ratio of average reward per prohibited state rate. Furthermore, using empirical evidence, the policy experiments shown with PTanh generalize to other utilities of the Prospect Shape. • Introduces PTanh utility function based on hyperbolic tangent for risk-sensitive RL. • Analyzes impact of diminishing marginals on risk-averse policy effectiveness. • Compares PTanh integrated with Q-learning and Expected SARSA against their base algorithms. • Demonstrates correlation between marginal effects and risk-averse policy parameters.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".