Analysis of a Movie Recommendation System Built from Scratch - Machine Learning

The project consists of building a movie recommendation system from scratch and analysis of different parameters like cost function, optimum feature size declaration etc.  Figure 2.2.2 shows that, the RMS Loss vs. Iteration curve for varied feature sizes shows that all loss curves start to level out with higher interaction rates. The loss is initially clearly higher because of the bigger feature size, and the gradient is also higher, similar to the mean loss curve. However, this graph differs in that the curves are positioned slightly farther apart. Figure 2.2.3 dipicts that, The RMS Loss vs. Iteration curve for fixed feature size shows that all loss curves begin to flatten as the iteration count rises. Each loss curve starts out with a single value, and as the learning rate varies, the steepness of the curve also changes. At a learning rate of 2e-5, the curves become more accurate. Figure 2.4.2 shows that, our model performs better for a specific value of feature size, according to the RMS Loss(Cost) vs. No of Feature Size plot. Yet, the crucial value of the feather size is also influenced by learning rate. With this model, four characteristics are about the right number. The impact of the regularization coefficient is also depicted in figure 3.4.2, where it causes the ideal feature size to change depending on the value of lambda. Data dependency and model parameter are trade-offs. 

Designing an ‘7-bit Hamming code encoder and decoder’ using the cadence tool -VLSI

The Hamming code serves the purpose of detecting and correcting single-bit errors that may occur during data transmission. This project proposes a Hamming code encoder and decoder with a configuration of 4 data bits and 3 parity bits. The RTL layout for this project is designed using Verilog Coding, and physical design is also conducted. Various analyses such as Slack Time, Nano Routine, Post Timing, and DRC checks are performed throughout the project. The Cadence software is utilized, utilizing BUET's Server, to implement and develop this project.