Deep RecSys in AI & Artificial Intelligence

Traditional models treat a user's history as a 'Bag of Items'—the order doesn't matter. But in the real world, order is everything. If you just bought a camera, you need a memory card *now*, not next year. Sequential Recommenders (using RNNs, GRUs, or LSTMs) treat the user history as a time-series. They learn 'Short-term Intent' (what are you doing in this session?) and 'Long-term Interest' (what do you usually like?), combining them for hyper-accurate next-item prediction.

Transformers have revolutionized RecSys via the Self-Attention mechanism. Instead of processing the history one item at a time (like an RNN), a Transformer looks at the entire history at once. It calculates weights to decide which past actions are most relevant to the current moment. This allows the model to ignore 'Noise' (the random video you clicked by mistake) and focus on the 'Signals' (the series you are currently binge-watching), even if those signals are buried deep in your history.

How do YouTube or TikTok choose one video from a billion in real-time? They use the Two-Tower Architecture. One neural network (the 'User Tower') compresses the user's complex history into a single vector. Another network (the 'Candidate Tower') does the same for every item. Because the final comparison is just a simple dot product between these two vectors, the system can use Approximate Nearest Neighbor (ANN) search to find the best recommendations in microseconds, even at planetary scale.

Taste is not static; it's a sequence. Deep Learning models like RNNs and Transformers allow us to understand the 'Order' of user actions, predicting what you want *next* based on what you just did.

While CF looks at a bag of items, Sequential Models look at a timeline. If you watch 'Toy Story 1' and 'Toy Story 2', the next prediction should be 'Toy Story 3'.

Transformers and Self-Attention take this further by weighing the importance of every past action. That movie you watched 5 minutes ago might be more relevant than the one you watched 5 years ago.

Checkpoint: Why are 'Sequential' models better for things like news or fashion?

• →They are faster
• →They understand that user interests change over time and that the order of actions matters (trends)

YouTube and TikTok use these 'Two-Tower' and 'Attention' architectures to drive their viral feeds, processing millions of actions in milliseconds.

By mastering Deep Learning for RecSys, you build the 'Infinite Feed'—a system that keeps users engaged by staying perfectly in sync with their evolving interests.

Checkpoint: What is 'Self-Attention' in a recommender system?

• →Random weights
• →A mechanism that allows the model to weigh the importance of different items in a user's history relative to each other

Deep Learning mastered! You've reached the cutting edge. Ready to handle the 'Now' with Session-Based Recommendations?

Predict a Real Rating from Embeddings. Finish predicting a rating as the dot product of a user embedding and an item embedding.