docs: update all docs

A/README.md

@@ -5,21 +5,3 @@
 Binary classification task (using PneumoniaMNIST dataset). The objective
 is to classify an image onto "Normal" (no pneumonia) or "Pneumonia"
 (presence of pneumonia)
-
-## TODOs:
-
-* [x] Design model for task A
-    * [x] Choose a model: [XGBoost](https://github.com/dmlc/xgboost)
-    * [x] Understand [boosted trees](https://xgboost.readthedocs.io/en/stable/tutorials/model.html)
-    * [x] Try a demo of XGBoost
-* [x] Report training, validation, and testing errors / accuracies, along with describe any hyper-parameter tunice process.
-    * [ ] [Cross Validation Reference](https://scikit-learn.org/dev/modules/cross_validation.html)
-    * [x] Add tools to [estimate models](https://xgboost.readthedocs.io/en/stable/python/sklearn_estimator.html)
-    * [x] [Get the best model and all metrics](https://xgboost.readthedocs.io/en/stable/python/examples/sklearn_examples.html#sphx-glr-python-examples-sklearn-examples-py) using [cross validation](https://scikit-learn.org/stable/modules/cross_validation.html#cross-validation)
-    * [x] Try [early stopping](https://xgboost.readthedocs.io/en/stable/python/sklearn_estimator.html#early-stopping)
-    * [x] Save and load [models](https://xgboost.readthedocs.io/en/stable/tutorials/saving_model.html)
-    * [ ] Plot all figures needed
-
-* The assessment will predominantly concentrate on how you articulate about the choice of models, how
-you develop/train/validate these models, and how you report/discuss/analyse the
-results.

A/solution.py

@@ -2,12 +2,10 @@
 # -*- coding: utf-8 -*-
 import os
 import sys
-from typing import List
 
 # hack here
 sys.path.append(os.path.join(os.path.dirname(__file__), ".."))
 
-from utils.logger import logger
 from utils.solution import Solution
 from constants import TASK_A_DIR
 

B/README.md

@@ -4,8 +4,3 @@
 
 Multi-class classification task (using PathMNIST dataset): The objective is
 to classify an image onto 9 different types of tissues.
-
-## TODOs:
-
-* [x] Design model for task B
-* [x] Report training, validation, and testing errors / accuracies, along with describe any hyper-parameter tunice process.

B/solution.py

@@ -2,7 +2,6 @@
 # -*- coding: utf-8 -*-
 import os
 import sys
-from typing import List
 
 # hack here
 sys.path.append(os.path.join(os.path.dirname(__file__), ".."))

README.md

@@ -15,32 +15,40 @@
 
 ## Overview
 
-* [ ] TODO: Add repo description here.
+This repo is for ELEC0134: Applied Machine Learning Systems 23/24.
 
-## Repo Structure
+It is using [XGBoost](https://github.com/dmlc/xgboost) to accomplish the binary classification task and the multi-class classificaion task.
 
-* [ ] TODO: Explain the role of each file.
+## Repo Structure
 
 ```bash
-# TODO: update repo structure
 $ tree
 .
 ├── A
 │   ├── README.md
 │   ├── config.json
 │   ├── solution.py                 # Solution for task A
+│   ├── task_a_early_stopping_rounds_3_evals_result.json
+│   ├── task_a_early_stopping_rounds_None_evals_result.json
 │   └── task_a_training_model.json  # Model saved
 ├── B
 │   ├── README.md
 │   ├── config.json
 │   ├── solution.py                 # Solution for task B
+│   ├── task_b_early_stopping_rounds_3_evals_result.json
+│   ├── task_b_early_stopping_rounds_None_evals_result.json
 │   └── task_b_training_model.json  # Model saved
 ├── Datasets                        # An empty dir for Datasets
 ├── Makefile
 ├── README.md
 ├── constants.py                    # global constants
 ├── environment.yml                 # conda environment config
+├── images
+│   ├── flowchart.png               # draw.io flowchart
+│   ├── task_a_learning_curve.png
+│   └── task_b_learning_curve.png
 ├── main.py                         # Entrypoint of this repo
+├── plot.ipynb                      # Plots
 ├── requirements.txt                # python package requirements
 └── utils
     ├── dataset.py                  # Dataset Loader
@@ -75,10 +83,9 @@ Datasets
 
 ## Requirements
 
-* [ ] TODO: Add assignment requirements.
+* `cuda 11.8`: If you want to use `py-xgboost-gpu`, GPU and `cuda` are required
 
 ```bash
-# TODO: update requirements.txt
 $ cat requirements.txt
 # format python files
 black
@@ -91,9 +98,11 @@ scikit-learn
 py-xgboost-gpu
 # save results
 pandas
+# plot
+matplotlib
 ```
 
-**NOTE**: `py-xgboost-gpu` will use GPU. If you want to use CPU only, use [`py-xgboost-cpu`](https://xgboost.readthedocs.io/en/stable/install.html#conda) and set the device using `--device cpu`.
+**NOTE**: `py-xgboost-gpu` will use GPU. If you want to use CPU instead, use [`py-xgboost-cpu`](https://xgboost.readthedocs.io/en/stable/install.html#conda) and set the device using `--device cpu`.
 
 ## Usage
 

environment.yml

@@ -8,4 +8,5 @@ dependencies:
   - numpy
   - py-xgboost-gpu
   - pandas
+  - matplotlib
 prefix: /home/uceezl8/.conda/envs/amls-final-zhaoyanlu

requirements.txt

@@ -9,3 +9,5 @@ scikit-learn
 py-xgboost-gpu
 # save results
 pandas
+# plot
+matplotlib

utils/dataset.py

@@ -21,15 +21,19 @@ def __init__(self, path: str) -> None:
         logger.info(
             "We concatenate the training set and validation set together for a better cross validation."
         )
+        # old_X_train is used for training
         self.old_X_train = data.get("train_images", np.asarray([]))
         self.old_X_train = self._reshape_to_2_dims(self.old_X_train)
         X_val = data.get("val_images", np.asarray([]))
         self.X_val = self._reshape_to_2_dims(X_val)
+        # X_train is used for cross validation
         self.X_train = np.concatenate((self.old_X_train, self.X_val), axis=0)
         logger.debug(f"X_train shape: {self.X_train.shape}")
 
+        # old_y_train is used for training
         self.old_y_train = data.get("train_labels", np.asarray([]))
         self.y_val = data.get("val_labels", np.asarray([]))
+        # y_train is used for cross validation
         self.y_train = np.concatenate((self.old_y_train, self.y_val), axis=0)
         logger.debug(f"y_train shape: {self.y_train.shape}")
 
@@ -42,6 +46,7 @@ def __init__(self, path: str) -> None:
         logger.info(f"Dataset loaded: {path}")
 
     def _reshape_to_2_dims(self, ndarray: np.ndarray) -> np.ndarray:
+        """Reshape a ndarry to the shape (n_samples, n_features)."""
         from functools import reduce
         import operator
 

utils/solution.py

@@ -72,6 +72,7 @@ def solve(self, stages: List[str]):
         logger.info(f"----------{self.task_name} finished!-----------")
 
     def val(self):
+        """Cross Validation"""
         logger.info(f"[{self.task_name}] [Validation] Running on {self.device}...")
         val_config: dict = self.config.get("validation", {})
         param_grid: dict = val_config.get("param_grid", {})
@@ -111,6 +112,7 @@ def val(self):
             df.to_csv(cv_result_path)
 
     def train(self):
+        """Training"""
         logger.info(f"[{self.task_name}] [Training] Running on {self.device}...")
         training_config: dict = self.config.get("training", {})
         random_state: int = self.config.get("random_state", DEFAULT_RANDOM_STATE)
@@ -161,6 +163,7 @@ def train(self):
         logger.info(f"classification report for training:\n{report}")
 
     def test(self):
+        """Testing"""
         logger.info(f"[{self.task_name}] [Testing] Running on {self.device}...")
 
         if not self.classifier: