100DaysofMLCode

This repository is a part of #100DaysofMLCode challenge. That means, I pledge to code machine learning for at least an hour everyday for the next 100 days. Challenge is starting from July 06, 2018.

Day 1: July 06, 2018

What I did ?

1. Perform the <b>classification task</b> on <b>Breast Cancer</b> dataset.
2. By Using, Keras (Tensorflow as a backend).
3. Achieve the accuracy near to <b>96%</b> after using <b>'Relu'</b> as an activation function.

What is Keras ?

<b>Keras</b> is a high-level machine learning framework for Python under the MIT license. To know more, see Keras official [documentation](https

Day 2: July 07, 2018

What I did ?

1. Explore the <b>Twitter Dataset</b> of around <b>3653 tweets</b> of PyData Connference, London.
2. Find out maximum number of characters in a tweet = <b>152</b>
3. Plot the histogram of <b>Frequency of characters in tweets</b>.
4. Find out most common words in tweets : <b>Python, datascience</b>
5. Find out most common callouts : <b>@KirkDBorne, @kdnuggets</b>
6. Find out most common hashtags : <b>#DataScience, #BigData</b>

What is pickle format ?

By default, the pickle data format uses a relatively compact binary representation. If you need optimal size characteristics, you can efficiently compress pickled data. To know more, see Pickle official documentation. Or you can follow up the Introduction-To-Pickle tutorial.

Day 3: July 08, 2018

What I did ?

1. Build up a Deep Learning Rest API using Keras
2. Predict the pictures of Dogs, Cats, Rabbit specific to their bread.
3. To setup your own environment and the source code Look Here

Day 4: July 09, 2018

What I did ?

1. Go through a series of lectures by <b>Lex Fridman</b>, a research scientist at MIT.
2. These lectures include the content about Deep Reinforcement Learning and Self-Driving Cars
3. To know more about, what I learned Look Here

Day 5: July 10, 2018

1. Build up a Deep learning model for classifing MNIST Dataset using Keras.
2. Tried different model and predict the accuracy:

| Model | Testing Accuracy (%) | | ------------- |:----------------: | | Basic Model | 11.35 | | He Normal | 75.52 | | Relu | 92.88 | | Adam | 92.59 | | Adam & Relu | 93.85 | | Batchnorm | 95.46 | | Batchnorm & Relu | 94.71 | | Dropout | 11.35 | | Ensemble | 27.93 |

3. Most methods improve the model training & test performance. That's why we will use them all together.
4. After applying all the models together we get the accuracy near to <b>98%</b>. Look Here

Day 6: July 11, 2018

What I did?

Go through a series of lectures on self driving cars by Udacity.

Day 7: July 12, 2018

Started learning about Recommendation Engines.