AI playbook

The playbook is a collection of AI projects ranging from Computer Vision to Natural Language Processing:

Reinforcement Learning:

• Reinforcement Learning Racing (includes random racing circuit generator through b-splines using pygame with agent acting on it using either DDQN, A3C or PPO)

Natural Language Processing:

• Speech Recogniser (implements a CNN + RNN architecture with CTCLoss for end-to-end speech recognition)

• POS Tagging (probabilistic graphical model(bayesian) such as HMM and MEMM, and Viterbi algorithm for decoding)

• NER Tagging (NN CRF-layer implementation, work in progress...)

• Word Embeddings (implements word embeddings generation such as Word2Vec and GloVe)

Computer Vision:

• Flower Classifier (includes transfer learning for flower classification with 99% accuracy on dataset)

• Dog Classifier (includes both custom CNN & transfer learning and for dog classification)

• Semantic Segmentation (uses MobileNetV2 with custom decoder to generate a semantic segmentation on COCO dataset)

• Style Transfer

Misc:

• Auction Sale Price Prediction (uses Random Forest and extra pre-processing such as feature importance and engineering)

• Audio Signal Notes

• Bike Sharing Prediction (includes DNN built in NumPy for bike sharing prediction)

• Collaborative Filtering (uses collaborative filtering to predict similar music recommendation based on similar interest of his peers)

• Random Forest (random forest implementation in Pandas and NumPy)

• Disease Linkage (implements ridge regression to detect multicollinearity among patient's characteristics leading to prostate cancer)

• Track Recogniser (implements generation of track's fingerprints through storing the most potent frequency bands with O(1) retrieval through hashing)

Disclaimer

The given structure is chosen as it offers better flexibility to make it more modular. Any feedback is greatly appreciated.