arXiv subjects

Jibin Joseph

Publications and source records attributed to Jibin Joseph.

3 recordsLinked to original sources

arXiv2026

MiniGPT: Rebuilding GPT from First Principles

This paper presents MiniGPT, a compact from-scratch implementation of GPT-style autoregressive language modeling in PyTorch. The aim is to rebuild the core GPT pipeline from first principles after studying the design of nanoGPT by Andrej Karpathy, while keeping the model and training code independently written in a single notebook. MiniGPT implements token and positional embeddings, causal multi-head self-attention, pre-LayerNorm Transformer blocks, residual connections, feed-forward MLP layers, next-token cross-entropy training (teacher forcing), validation tracking, checkpoint selection, and autoregressive text generation. This paper evaluates the implementation on Tiny Shakespeare dataset using character-level tokenization. A baseline 0.83M-parameter model reaches a validation loss of 1.7236 after 3000 training iterations. A stronger 10.77M-parameter configuration, using a larger context length and improved training settings, reaches a best validation loss of 1.4780 and generates text with recognizable Shakespeare-style dialogue structure. MiniGPT does not introduce a new language-model architecture. Instead, it documents a clear and reproducible implementation path from raw text to trained character-level generation, including design choices, training behavior, generation quality, and practical limitations.

cs.CL↗

arXiv2025

Machine Learning to Predict Digital Frustration from Clickstream Data

Many businesses depend on their mobile apps and websites, so user frustration while trying to complete a task on these channels can cause lost sales and complaints. In this research, I use clickstream data from a real e-commerce site to predict whether a session is frustrated or not. Frustration is defined using certain rules based on rage bursts, back and forth navigation (U turns), cart churn, search struggle, and long wandering sessions, and applies these rules to 5.4 million raw clickstream events (304,881 sessions). From each session, I build tabular features and train standard classifier models. I also use the full event sequence to train a discriminative LSTM classifier. XGBoost reaches about 90% accuracy, ROC AUC of 0.9579, while the LSTM performs best with about 91% accuracy and a ROC AUC of 0.9705. Finally, the research shows that with only the first 20 to 30 interactions, the LSTM already predicts frustration reliably.

cs.LG↗

arXiv2023

Measuring the Numerical Viscosity in Simulations of Protoplanetary Disks in Cartesian Grids -- The Viscously Spreading Ring Revisited

Hydrodynamical simulations solve the governing equations on a discrete grid of space and time. This discretization causes numerical diffusion similar to a physical viscous diffusion, whose magnitude is often unknown or poorly constrained. With the current trend of simulating accretion disks with no or very low prescribed physical viscosity, it becomes essential to understand and quantify this inherent numerical diffusion, in the form of a numerical viscosity. We study the behavior of the viscous spreading ring and the spiral instability that develops in it. We then use this setup to quantify the numerical viscosity in Cartesian grids and study its properties. We simulate the viscous spreading ring and the related instability on a two-dimensional polar grid using PLUTO as well as FARGO, and ensure convergence of our results with a resolution study. We then repeat our models on a Cartesian grid and measure the numerical viscosity by comparing results to the known analytical solution, using PLUTO and Athena++. We find that the numerical viscosity in a Cartesian grid scales with resolution as approximately $ν_{num}\proptoΔx^2$ and is equivalent to an effective $α\sim10^{-4}$ for a common numerical setup. We also show that the spiral instability manifests as a single leading spiral throughout the whole domain on polar grids. This is contrary to previous results and indicates that sufficient resolution is necessary in order to correctly resolve the instability. Our results are relevant in the context of models where the origin should be included in the computational domain, or when polar grids cannot be used. Examples of such cases include models of disk accretion onto a central binary and inherently Cartesian codes.

astro-ph.IM↗