Breast cancer classification (with Deep Learning)

Выполнить проект по Deep Learning в формате Kaggle Competition -
Breast cancer classification.
Требуется знание в предметной области, чтобы к имеющемуся черновику решения
предложить методику оценки производительности различных моделей
(помимо обычных методов Machine Learning обязательно использование
нейронных сетей,
подходящих для решения поставленной задачи.
Входные данные:
2 файла - датасеты для тренинга и теста в формате csv
Код выполняется в Jupiter Notebook, на выходе необходимо предоставить
код и результат классификации в формате csv.
Описание на английском:

Three main goals: 

1. Employ neural networks as a powerful approach to supervised machine learning 

2. Use using state-of-the-art software tools for machine learning, 

3. Investigate the impact of different architectures and hyper-parameters settings on effectiveness 

4. Implement state-of-the-art model(s) and evaluate their performance against baseline deep learning models. 

5. Report and communicate your findings in a clear, transparent and succinct manner.

The problems can be solved using Python, TensorFlow and Keras along with any other packages. You can also use PyTorch.  

This task is to classify tumours into Malignant or Benign type when provided with the tumour's attributes in the dataset.

  More information about the data can be found at: https://archive.ics.uci.edu/ml/datasets/breast+cancer+wisconsin+(diagnostic) 

For the binary classification problem, you will need to: 

• implement a standard machine learning base line model of your choice (but it should be appropriate to the task) 

• implement a dense deep neural network where you need to find a good configuration, i.e. appropriate number of layers, neurons, activation functions, optimiser etc. 

• Implement additional neural networks that follow a different architecture (i.e. such as the wide and deep network, or another one you can find in the literature) 

• Evaluate the performance of the best configurations for each of the different networks. 

Если необходима ссылка на соревнование Kaggle, пишите.