Add Diffusion models (#70)

* add diffusion model notebook

* diffusion model frrameworks

* diffusion process

* add diffusion examples

* update

* diffusion naive model

* naive diffusion model

* update diffusion models

* encoder and decoder

* add diffusion model

* md lint

Author: emptymalei

dl/notebooks/diffusion_model.py

@@ -0,0 +1,11 @@
+# ---
+# jupyter:
+#   jupytext:
+#     text_representation:
+#       extension: .py
+#       format_name: light
+#       format_version: '1.5'
+#       jupytext_version: 1.15.2
+# ---
+
+# # Diffusion Process

dl/notebooks/diffusion_model_timegrad.py

@@ -0,0 +1,50 @@
+# ---
+# jupyter:
+#   jupytext:
+#     text_representation:
+#       extension: .py
+#       format_name: light
+#       format_version: '1.5'
+#       jupytext_version: 1.15.2
+#   kernelspec:
+#     display_name: .venv
+#     language: python
+#     name: python3
+# ---
+
+# # TabPFN as Forecaster
+
+import pandas as pd
+from ts_dl_utils.datasets.pendulum import Pendulum
+
+pen = Pendulum(length=100)
+
+df = pd.DataFrame(pen(300, 30, initial_angle=1, beta=0.00001))
+
+
+def time_delay_embed(df: pd.DataFrame, window_size: int) -> pd.DataFrame:
+    """embed time series into a time delay embedding space
+
+    Time column `t` is required in the input data frame.
+
+    :param df: original time series data frame
+    :param window_size: window size for the time delay embedding
+    """
+    dfs_embedded = []
+
+    for i in df.rolling(window_size):
+        i_t = i.t.iloc[0]
+        dfs_embedded.append(
+            pd.DataFrame(i.reset_index(drop=True))
+            .drop(columns=["t"])
+            .T.reset_index(drop=True)
+            # .rename(columns={"index": "name"})
+            # .assign(t=i_t)
+        )
+
+    df_embedded = pd.concat(dfs_embedded[window_size - 1 :])
+
+    return df_embedded
+
+
+df = time_delay_embed(df, 11)

dl/notebooks/diffusion_process.py

@@ -61,4 +61,5 @@ Terminologies for time series data may be different in different fields[^Hyndman
 [^Das2019]: Das P. Econometrics in Theory and Practice. Springer Nature Singapore; [doi:10.1007/978-981-32-9019-8](https://link.springer.com/book/10.1007/978-981-32-9019-8)
 [^Dawid2022]: Dawid P, Tewari A. On learnability under general stochastic processes. Harvard Data Science Review. 2022;
 [^Hyndman2020]: Hyndman R. Rob J Hyndman - Terminology matters. In: Rob J Hyndman [Internet]. 26 Jun 2020 [cited 9 Nov 2023]. Available: https://robjhyndman.com/hyndsight/terminology-matters/#same-concept-different-terminology
-4. [doi:10.1162/99608f92.dec7d780](https://hdsr.mitpress.mit.edu/pub/qixx99zn/release/1)
+
+1. [doi:10.1162/99608f92.dec7d780](https://hdsr.mitpress.mit.edu/pub/qixx99zn/release/1)

dl/notebooks/tabpfn.py

@@ -262,6 +262,7 @@ nav:
       - "Generate Time Series Using Statistics": notebooks/time-series-data-generation.py
       - "Generate Time Series Using VAE": notebooks/time_vae.py
       - "Comparing Time Series": notebooks/timeseries-comparison.py
+      - "Diffusion Models": notebooks/diffusion_model.py
     - "Small Yet Powerful Concepts":
       - concepts/index.md
       - "Entropy": concepts/entropy.md

dl/time-series/timeseries-data.analysis.md

@@ -33,6 +33,7 @@ jupytext = "^1.15.2"
 [tool.poetry.group.visualization.dependencies]
 seaborn = "^0.13.2"
 opentsne = "^1.0.2"
+plotly = "^6.1.1"
 
 
 [tool.poetry.group.data.dependencies]