chasemathsi/BayesianInference

1import gradio as gr2import numpy as np3import matplotlib.pyplot as plt4from scipy.stats import norm, binom, beta, gamma, poisson5 6import pandas as pd7 8class Prior():9    def __init__(self, distribution, param1, param2):10        self.distribution = distribution11        self.param1 = param112        self.param2 = param213        self.left = distribution.ppf(0.0001, param1, param2)14        self.right = distribution.ppf(0.9999, param1, param2)15 16    def plot(self):17        x = np.linspace(self.left, self.right, 100)18        y = self.distribution.pdf(x, self.param1, self.param2)19        20        # Set common axis limits for both plots21        plt.xlim(self.left, self.right)22        plt.ylim(0, max(y) * 1.1)23        24        return plt.plot(x, y)25 26 27class SamplingModel():28    def __init__(self, distribution, data):29        self.data = pd.read_csv(data.name, header=None).to_numpy().squeeze()30        self.distrubution = distribution31 32    def plot(self):33        plt.hist(self.data, density=True)34        35        # Set common axis limits for both plots36        left, right = plt.xlim()37        plt.xlim(left, right)38        39        plt.show()40 41 42 43 44class PosteriorModel():45    def __init__(self, distribution):46        pass47 48        49 50 51 52 53def run(data, prior_dist, sampling_dist, param1, param2):54    plt.figure(figsize=(8*3, 5*3))55    dists = {56        "Normal":norm,57        "Beta":beta,58        "Gamma":gamma59    }60 61    # set up prior and sampling model62    prior = Prior(dists[prior_dist], param1, param2)63    model = SamplingModel(sampling_dist, data)64    prior.plot()65    model.plot()66    plt.grid()67    plt.autoscale(tight = True)68    69    # Save and return the plot70    plt.tight_layout()71    return plt72 73 74# demo = gr.Interface(75#     fn=run,76#     inputs=[77#         gr.components.File(label="Upload CSV"),78#         gr.components.Dropdown(79#             choices=["Normal", "Binomial", "Beta", "Gamma", "Poisson"], 80#             label="Choose Prior Distribution"81#         ),82#         gr.components.Dropdown(83#             choices=["Normal", "Binomial", "Beta", "Gamma", "Poisson"], 84#             label="Choose Sampling Distribution"85#         ),86#         gr.components.Number(label=r"\theta"),87#         gr.components.Number(label=r"\sigma^2")88#     ],89#     outputs="plot"90# )91 92def predict(img):93    x = torch.tensor(img, dtype=torch.float32).unsqueeze(0).unsqueeze(0) / 255.94    with torch.no_grad():95        out = model(x)96    probabilities = torch.nn.functional.softmax(out[0], dim=0)97    values, indices = torch.topk(probabilities, 5)98    confidences = {LABELS[i]: v.item() for i, v in zip(indices, values)}99    return confidences100 101gr.Interface(fn=predict,102             inputs="sketchpad",103             outputs="label",104             live=True).launch()105 106# demo.launch()107