rlcmab-sampler 0.1.0__tar.gz

rlcmab_sampler-0.1.0/PKG-INFO

@@ -0,0 +1,84 @@
+Metadata-Version: 2.3
+Name: rlcmab-sampler
+Version: 0.1.0
+Summary: Add your description here
+Author: Saswata Sarkar
+Author-email: Saswata Sarkar <sarkarsaswata01@gmail.com>
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+
+# rlcmab-lab3-sampler
+
+A reward sampler for Contextual Bandit news recommendation systems.
+
+## Installation
+
+```bash
+pip install rlcmab-lab3-sampler
+```
+
+Or install from source:
+
+```bash
+cd sampler
+pip install .
+```
+
+## Quick Start
+
+```python
+from sampler import sampler
+
+# Initialize with your roll number (i)
+reward_sampler = sampler(i=42)
+
+# Get a reward from arm j
+reward = reward_sampler.sample(j=5)
+print(reward)
+
+# Sample a few arms
+for j in range(3):
+    print(j, reward_sampler.sample(j))
+```
+
+## API Reference
+
+### `sampler(i, n_arms=12)`
+
+Initialize a new sampler instance.
+
+**Parameters:**
+
+- `i` (int): Student ID used to seed the random number generator
+- `n_arms` (int, optional): Number of arms/articles. Default: 12
+
+**Attributes:**
+
+- `means`: List of mean values for each Gaussian distribution
+- `stds`: List of standard deviations for each Gaussian distribution
+- `distributions`: List of scipy.stats.norm distribution objects
+
+### `sample(j)`
+
+Sample a reward from the specified arm.
+
+**Parameters:**
+
+- `j` (int): Arm index (must be between 0 and 11)
+
+**Returns:**
+
+- `float`: Reward value sampled from the arm's Gaussian distribution
+
+**Raises:**
+
+- `ValueError`: If `j` is not in the valid range [0, 11]
+
+## Reproducibility
+
+For the same student ID `i`, the generated arm distributions remain the same
+across runs. Different student IDs map to different distributions.
+
+## License
+
+[MIT License](LICENSE)

rlcmab_sampler-0.1.0/README.md

@@ -0,0 +1,75 @@
+# rlcmab-lab3-sampler
+
+A reward sampler for Contextual Bandit news recommendation systems.
+
+## Installation
+
+```bash
+pip install rlcmab-lab3-sampler
+```
+
+Or install from source:
+
+```bash
+cd sampler
+pip install .
+```
+
+## Quick Start
+
+```python
+from sampler import sampler
+
+# Initialize with your roll number (i)
+reward_sampler = sampler(i=42)
+
+# Get a reward from arm j
+reward = reward_sampler.sample(j=5)
+print(reward)
+
+# Sample a few arms
+for j in range(3):
+    print(j, reward_sampler.sample(j))
+```
+
+## API Reference
+
+### `sampler(i, n_arms=12)`
+
+Initialize a new sampler instance.
+
+**Parameters:**
+
+- `i` (int): Student ID used to seed the random number generator
+- `n_arms` (int, optional): Number of arms/articles. Default: 12
+
+**Attributes:**
+
+- `means`: List of mean values for each Gaussian distribution
+- `stds`: List of standard deviations for each Gaussian distribution
+- `distributions`: List of scipy.stats.norm distribution objects
+
+### `sample(j)`
+
+Sample a reward from the specified arm.
+
+**Parameters:**
+
+- `j` (int): Arm index (must be between 0 and 11)
+
+**Returns:**
+
+- `float`: Reward value sampled from the arm's Gaussian distribution
+
+**Raises:**
+
+- `ValueError`: If `j` is not in the valid range [0, 11]
+
+## Reproducibility
+
+For the same student ID `i`, the generated arm distributions remain the same
+across runs. Different student IDs map to different distributions.
+
+## License
+
+[MIT License](LICENSE)

rlcmab_sampler-0.1.0/pyproject.toml

@@ -0,0 +1,14 @@
+[project]
+name = "rlcmab-sampler"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+authors = [
+    { name = "Saswata Sarkar", email = "sarkarsaswata01@gmail.com" }
+]
+requires-python = ">=3.12"
+dependencies = []
+
+[build-system]
+requires = ["uv_build>=0.9.11,<0.10.0"]
+build-backend = "uv_build"

rlcmab_sampler-0.1.0/src/rlcmab_sampler/__init__.py

@@ -0,0 +1,50 @@
+import numpy as np
+import scipy.stats as stats
+
+
+class sampler:
+    """
+    Gaussian reward sampler for Contextual Bandit news recommendation.
+
+    Properties:
+    - Same i → same hidden distributions
+    - Strong separation across students
+    - 12 Gaussian arms
+    - Continuous rewards (old behavior preserved)
+    """
+
+    def __init__(self, i, n_arms=12):
+        self.i = int(i)
+        self.n_arms = n_arms
+
+        seed = self._derive_seed(self.i)
+        self.rng = np.random.default_rng(seed)
+
+        self.means, self.stds = self._generate_gaussian_parameters()
+
+        self.distributions = [
+            stats.norm(m, s) for m, s in zip(self.means, self.stds)
+        ]
+
+    def _derive_seed(self, i):
+        return (i * 2654435761) % (2**32)
+
+    def _generate_gaussian_parameters(self):
+        base_means = np.linspace(-8.0, 8.0, self.n_arms)
+        noise = self.rng.normal(0.0, 2.0, size=self.n_arms)
+
+        means = base_means + noise
+        self.rng.shuffle(means)
+
+        stds = self.rng.uniform(0.8, 1.4, size=self.n_arms)
+        return means.tolist(), stds.tolist()
+
+    def sample(self, j):
+        if not (0 <= j < self.n_arms):
+            raise ValueError("Index j must be between 0 and 11.")
+        return self.distributions[j].rvs()
+
+
+# Explicit public API
+__all__ = ["sampler"]
+__author__ = "Saswata Sarkar"