build based on 341ed4a

Author: Documenter.jl

dev/.documenter-siteinfo.json

@@ -1 +1 @@
-{"documenter":{"julia_version":"1.12.5","generation_timestamp":"2026-03-25T15:16:01","documenter_version":"1.17.0"}}
+{"documenter":{"julia_version":"1.12.5","generation_timestamp":"2026-03-26T14:14:37","documenter_version":"1.17.0"}}

dev/api/index.html

@@ -3,4 +3,4 @@
 \fbox{Neural network $\varphi_w$}
 \xrightarrow[\text{Predicted values}]{\hat{\theta}}
 \fbox{\texttt{one\_hot\_argmax}}
-\xrightarrow[\text{Decision}]{y}\]</p><p>The neural network predicts item values <span>$\hat{\theta} \in \mathbb{R}^n$</span> from the feature vector <span>$x \in \mathbb{R}^{n+d}$</span>. The default architecture is <code>Dense(n+d =&gt; n; bias=false)</code>, which can exactly recover the optimal linear predictor <span>$[I_n \mid W]$</span>, so a well-trained model should reach near-zero gap.</p><h3 id="SAA-Policy"><a class="docs-heading-anchor" href="#SAA-Policy">SAA Policy</a><a id="SAA-Policy-1"></a><a class="docs-heading-anchor-permalink" href="#SAA-Policy" title="Permalink"></a></h3><p><span>$y_{\text{SAA}} = \operatorname{argmax}\bigl(\frac{1}{S}\sum_s \xi^{(s)}\bigr)$</span> — the exact SAA-optimal decision for linear argmax, accessible via <code>generate_baseline_policies(bench).saa</code>.</p></article><nav class="docs-footer"><a class="docs-footer-prevpage" href="../argmax/">« Argmax</a><a class="docs-footer-nextpage" href="../dvsp/">Dynamic Vehicle Scheduling »</a><div class="flexbox-break"></div><p class="footer-message">Powered by <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> and the <a href="https://julialang.org/">Julia Programming Language</a>.</p></nav></div><div class="modal" id="documenter-settings"><div class="modal-background"></div><div class="modal-card"><header class="modal-card-head"><p class="modal-card-title">Settings</p><button class="delete"></button></header><section class="modal-card-body"><p><label class="label">Theme</label><div class="select"><select id="documenter-themepicker"><option value="auto">Automatic (OS)</option><option value="documenter-light">documenter-light</option><option value="documenter-dark">documenter-dark</option><option value="catppuccin-latte">catppuccin-latte</option><option value="catppuccin-frappe">catppuccin-frappe</option><option value="catppuccin-macchiato">catppuccin-macchiato</option><option value="catppuccin-mocha">catppuccin-mocha</option></select></div></p><hr/><p>This document was generated with <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> version 1.17.0 on <span class="colophon-date" title="Wednesday 25 March 2026 15:15">Wednesday 25 March 2026</span>. Using Julia version 1.12.5.</p></section><footer class="modal-card-foot"></footer></div></div></div></body></html>
+\xrightarrow[\text{Decision}]{y}\]</p><p>The neural network predicts item values <span>$\hat{\theta} \in \mathbb{R}^n$</span> from the feature vector <span>$x \in \mathbb{R}^{n+d}$</span>. The default architecture is <code>Dense(n+d =&gt; n; bias=false)</code>, which can exactly recover the optimal linear predictor <span>$[I_n \mid W]$</span>, so a well-trained model should reach near-zero gap.</p><h3 id="SAA-Policy"><a class="docs-heading-anchor" href="#SAA-Policy">SAA Policy</a><a id="SAA-Policy-1"></a><a class="docs-heading-anchor-permalink" href="#SAA-Policy" title="Permalink"></a></h3><p><span>$y_{\text{SAA}} = \operatorname{argmax}\bigl(\frac{1}{S}\sum_s \xi^{(s)}\bigr)$</span> — the exact SAA-optimal decision for linear argmax, accessible via <code>generate_baseline_policies(bench).saa</code>.</p></article><nav class="docs-footer"><a class="docs-footer-prevpage" href="../argmax/">« Argmax</a><a class="docs-footer-nextpage" href="../dvsp/">Dynamic Vehicle Scheduling »</a><div class="flexbox-break"></div><p class="footer-message">Powered by <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> and the <a href="https://julialang.org/">Julia Programming Language</a>.</p></nav></div><div class="modal" id="documenter-settings"><div class="modal-background"></div><div class="modal-card"><header class="modal-card-head"><p class="modal-card-title">Settings</p><button class="delete"></button></header><section class="modal-card-body"><p><label class="label">Theme</label><div class="select"><select id="documenter-themepicker"><option value="auto">Automatic (OS)</option><option value="documenter-light">documenter-light</option><option value="documenter-dark">documenter-dark</option><option value="catppuccin-latte">catppuccin-latte</option><option value="catppuccin-frappe">catppuccin-frappe</option><option value="catppuccin-macchiato">catppuccin-macchiato</option><option value="catppuccin-mocha">catppuccin-mocha</option></select></div></p><hr/><p>This document was generated with <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> version 1.17.0 on <span class="colophon-date" title="Thursday 26 March 2026 14:14">Thursday 26 March 2026</span>. Using Julia version 1.12.5.</p></section><footer class="modal-card-foot"></footer></div></div></div></body></html>

dev/benchmarks/argmax/index.html

@@ -3,4 +3,4 @@
 \fbox{Neural network $\varphi_w$}
 \xrightarrow[\text{Prizes}]{\theta}
 \fbox{Prize-collecting VSP}
-\xrightarrow[\text{Routes}]{a_t}\]</p><p><strong>Components</strong>:</p><ol><li><strong>Neural Network</strong> <span>$\varphi_w$</span>: Takes current state features as input and predicts customer prizes <span>$\theta = (\theta_1, \ldots, \theta_n)$</span>, one value per postponable customer.</li><li><strong>Optimization Layer</strong>: Solves the prize-collecting vehicle scheduling problem to determine optimal routes given the predicted prizes, by maximizing total collected prizes minus travel costs:  <code>math  \max_{a_t\in \mathcal{A}(s_t)} \sum_{r \in a_t} \left( \sum_{i \in r} \theta_i - \sum_{(i,j) \in r} d_{ij} \right)</code>  This can be modeled as a flow linear program on a directed acyclic graph (DAG) and is solved using standard LP solvers.</li></ol><p>The neural network architecture adapts to the feature dimensionality:</p><ul><li><strong>2D features</strong>: <code>Dense(2 =&gt; 1)</code>, applied in parallel to each postponable customer</li><li><strong>Full features</strong>: <code>Dense(27 =&gt; 1)</code> applied in parallel to each postponable customer</li></ul><p><strong>Note:</strong> one can also use more complex architectures such as a deeper MLP or a graph neural network for better performance.</p></article><nav class="docs-footer"><a class="docs-footer-prevpage" href="../contextual_stochastic_argmax/">« Contextual Stochastic Argmax</a><a class="docs-footer-nextpage" href="../dynamic_assortment/">Dynamic Assortment »</a><div class="flexbox-break"></div><p class="footer-message">Powered by <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> and the <a href="https://julialang.org/">Julia Programming Language</a>.</p></nav></div><div class="modal" id="documenter-settings"><div class="modal-background"></div><div class="modal-card"><header class="modal-card-head"><p class="modal-card-title">Settings</p><button class="delete"></button></header><section class="modal-card-body"><p><label class="label">Theme</label><div class="select"><select id="documenter-themepicker"><option value="auto">Automatic (OS)</option><option value="documenter-light">documenter-light</option><option value="documenter-dark">documenter-dark</option><option value="catppuccin-latte">catppuccin-latte</option><option value="catppuccin-frappe">catppuccin-frappe</option><option value="catppuccin-macchiato">catppuccin-macchiato</option><option value="catppuccin-mocha">catppuccin-mocha</option></select></div></p><hr/><p>This document was generated with <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> version 1.17.0 on <span class="colophon-date" title="Wednesday 25 March 2026 15:15">Wednesday 25 March 2026</span>. Using Julia version 1.12.5.</p></section><footer class="modal-card-foot"></footer></div></div></div></body></html>
+\xrightarrow[\text{Routes}]{a_t}\]</p><p><strong>Components</strong>:</p><ol><li><strong>Neural Network</strong> <span>$\varphi_w$</span>: Takes current state features as input and predicts customer prizes <span>$\theta = (\theta_1, \ldots, \theta_n)$</span>, one value per postponable customer.</li><li><strong>Optimization Layer</strong>: Solves the prize-collecting vehicle scheduling problem to determine optimal routes given the predicted prizes, by maximizing total collected prizes minus travel costs:  <code>math  \max_{a_t\in \mathcal{A}(s_t)} \sum_{r \in a_t} \left( \sum_{i \in r} \theta_i - \sum_{(i,j) \in r} d_{ij} \right)</code>  This can be modeled as a flow linear program on a directed acyclic graph (DAG) and is solved using standard LP solvers.</li></ol><p>The neural network architecture adapts to the feature dimensionality:</p><ul><li><strong>2D features</strong>: <code>Dense(2 =&gt; 1)</code>, applied in parallel to each postponable customer</li><li><strong>Full features</strong>: <code>Dense(27 =&gt; 1)</code> applied in parallel to each postponable customer</li></ul><p><strong>Note:</strong> one can also use more complex architectures such as a deeper MLP or a graph neural network for better performance.</p></article><nav class="docs-footer"><a class="docs-footer-prevpage" href="../contextual_stochastic_argmax/">« Contextual Stochastic Argmax</a><a class="docs-footer-nextpage" href="../dynamic_assortment/">Dynamic Assortment »</a><div class="flexbox-break"></div><p class="footer-message">Powered by <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> and the <a href="https://julialang.org/">Julia Programming Language</a>.</p></nav></div><div class="modal" id="documenter-settings"><div class="modal-background"></div><div class="modal-card"><header class="modal-card-head"><p class="modal-card-title">Settings</p><button class="delete"></button></header><section class="modal-card-body"><p><label class="label">Theme</label><div class="select"><select id="documenter-themepicker"><option value="auto">Automatic (OS)</option><option value="documenter-light">documenter-light</option><option value="documenter-dark">documenter-dark</option><option value="catppuccin-latte">catppuccin-latte</option><option value="catppuccin-frappe">catppuccin-frappe</option><option value="catppuccin-macchiato">catppuccin-macchiato</option><option value="catppuccin-mocha">catppuccin-mocha</option></select></div></p><hr/><p>This document was generated with <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> version 1.17.0 on <span class="colophon-date" title="Thursday 26 March 2026 14:14">Thursday 26 March 2026</span>. Using Julia version 1.12.5.</p></section><footer class="modal-card-foot"></footer></div></div></div></body></html>

dev/benchmarks/contextual_stochastic_argmax/index.html

@@ -9,4 +9,4 @@
 \fbox{Neural network $\varphi_w$}
 \xrightarrow[\text{Cost vector}]{\theta}
 \fbox{Top K}
-\xrightarrow[\text{Assortment}]{a_t}\]</p><p><strong>Components</strong>:</p><ol><li><strong>Neural Network</strong> <span>$\varphi_w$</span>: Takes the current state <span>$s_t$</span> as input and predicts item utilities <span>$\theta = (\theta_1, \ldots, \theta_N)$</span></li><li><strong>Optimization Layer</strong>: Selects the top <span>$K$</span> items with highest predicted utilities to form the assortment <span>$a_t$</span></li></ol><h2 id="Reference"><a class="docs-heading-anchor" href="#Reference">Reference</a><a id="Reference-1"></a><a class="docs-heading-anchor-permalink" href="#Reference" title="Permalink"></a></h2><p>Based on the paper: <a href="https://arxiv.org/abs/2505.19053">Structured Reinforcement Learning for Combinatorial Decision-Making</a></p></article><nav class="docs-footer"><a class="docs-footer-prevpage" href="../dvsp/">« Dynamic Vehicle Scheduling</a><a class="docs-footer-nextpage" href="../fixed_size_shortest_path/">Shortest paths »</a><div class="flexbox-break"></div><p class="footer-message">Powered by <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> and the <a href="https://julialang.org/">Julia Programming Language</a>.</p></nav></div><div class="modal" id="documenter-settings"><div class="modal-background"></div><div class="modal-card"><header class="modal-card-head"><p class="modal-card-title">Settings</p><button class="delete"></button></header><section class="modal-card-body"><p><label class="label">Theme</label><div class="select"><select id="documenter-themepicker"><option value="auto">Automatic (OS)</option><option value="documenter-light">documenter-light</option><option value="documenter-dark">documenter-dark</option><option value="catppuccin-latte">catppuccin-latte</option><option value="catppuccin-frappe">catppuccin-frappe</option><option value="catppuccin-macchiato">catppuccin-macchiato</option><option value="catppuccin-mocha">catppuccin-mocha</option></select></div></p><hr/><p>This document was generated with <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> version 1.17.0 on <span class="colophon-date" title="Wednesday 25 March 2026 15:16">Wednesday 25 March 2026</span>. Using Julia version 1.12.5.</p></section><footer class="modal-card-foot"></footer></div></div></div></body></html>
+\xrightarrow[\text{Assortment}]{a_t}\]</p><p><strong>Components</strong>:</p><ol><li><strong>Neural Network</strong> <span>$\varphi_w$</span>: Takes the current state <span>$s_t$</span> as input and predicts item utilities <span>$\theta = (\theta_1, \ldots, \theta_N)$</span></li><li><strong>Optimization Layer</strong>: Selects the top <span>$K$</span> items with highest predicted utilities to form the assortment <span>$a_t$</span></li></ol><h2 id="Reference"><a class="docs-heading-anchor" href="#Reference">Reference</a><a id="Reference-1"></a><a class="docs-heading-anchor-permalink" href="#Reference" title="Permalink"></a></h2><p>Based on the paper: <a href="https://arxiv.org/abs/2505.19053">Structured Reinforcement Learning for Combinatorial Decision-Making</a></p></article><nav class="docs-footer"><a class="docs-footer-prevpage" href="../dvsp/">« Dynamic Vehicle Scheduling</a><a class="docs-footer-nextpage" href="../fixed_size_shortest_path/">Shortest paths »</a><div class="flexbox-break"></div><p class="footer-message">Powered by <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> and the <a href="https://julialang.org/">Julia Programming Language</a>.</p></nav></div><div class="modal" id="documenter-settings"><div class="modal-background"></div><div class="modal-card"><header class="modal-card-head"><p class="modal-card-title">Settings</p><button class="delete"></button></header><section class="modal-card-body"><p><label class="label">Theme</label><div class="select"><select id="documenter-themepicker"><option value="auto">Automatic (OS)</option><option value="documenter-light">documenter-light</option><option value="documenter-dark">documenter-dark</option><option value="catppuccin-latte">catppuccin-latte</option><option value="catppuccin-frappe">catppuccin-frappe</option><option value="catppuccin-macchiato">catppuccin-macchiato</option><option value="catppuccin-mocha">catppuccin-mocha</option></select></div></p><hr/><p>This document was generated with <a href="https://github.com/JuliaDocs/Documenter.jl">Documenter.jl</a> version 1.17.0 on <span class="colophon-date" title="Thursday 26 March 2026 14:14">Thursday 26 March 2026</span>. Using Julia version 1.12.5.</p></section><footer class="modal-card-foot"></footer></div></div></div></body></html>