Literature survey, paper reviews, experimental setups and a collection of implementations for baselines methods for predictive uncertainty estimation in deep learning models.

[TNNLS] A Comprehensive Survey of Awesome Visual Transformer Literatures.

A Claude Code plugin that turns your terminal into an autonomous research lab — literature survey, experiment execution, paper writing, all in one pipeline.

Introduction Parkinson’s Disease is the second most prevalent neurodegenerative disorder after Alzheimer’s, affecting more than 10 million people worldwide. Parkinson’s is characterized primarily by the deterioration of motor and cognitive ability. There is no single test which can be administered for diagnosis. Instead, doctors must perform a careful clinical analysis of the patient’s medical history. Unfortunately, this method of diagnosis is highly inaccurate. A study from the National Institute of Neurological Disorders finds that early diagnosis (having symptoms for 5 years or less) is only 53% accurate. This is not much better than random guessing, but an early diagnosis is critical to effective treatment. Because of these difficulties, I investigate a machine learning approach to accurately diagnose Parkinson’s, using a dataset of various speech features (a non-invasive yet characteristic tool) from the University of Oxford. Why speech features? Speech is very predictive and characteristic of Parkinson’s disease; almost every Parkinson’s patient experiences severe vocal degradation (inability to produce sustained phonations, tremor, hoarseness), so it makes sense to use voice to diagnose the disease. Voice analysis gives the added benefit of being non-invasive, inexpensive, and very easy to extract clinically. Background Parkinson's Disease Parkinson’s is a progressive neurodegenerative condition resulting from the death of the dopamine containing cells of the substantia nigra (which plays an important role in movement). Symptoms include: “frozen” facial features, bradykinesia (slowness of movement), akinesia (impairment of voluntary movement), tremor, and voice impairment. Typically, by the time the disease is diagnosed, 60% of nigrostriatal neurons have degenerated, and 80% of striatal dopamine have been depleted. Performance Metrics TP = true positive, FP = false positive, TN = true negative, FN = false negative Accuracy: (TP+TN)/(P+N) Matthews Correlation Coefficient: 1=perfect, 0=random, -1=completely inaccurate Algorithms Employed Logistic Regression (LR): Uses the sigmoid logistic equation with weights (coefficient values) and biases (constants) to model the probability of a certain class for binary classification. An output of 1 represents one class, and an output of 0 represents the other. Training the model will learn the optimal weights and biases. Linear Discriminant Analysis (LDA): Assumes that the data is Gaussian and each feature has the same variance. LDA estimates the mean and variance for each class from the training data, and then uses properties of statistics (Bayes theorem , Gaussian distribution, etc) to compute the probability of a particular instance belonging to a given class. The class with the largest probability is the prediction. k Nearest Neighbors (KNN): Makes predictions about the validation set using the entire training set. KNN makes a prediction about a new instance by searching through the entire set to find the k “closest” instances. “Closeness” is determined using a proximity measurement (Euclidean) across all features. The class that the majority of the k closest instances belong to is the class that the model predicts the new instance to be. Decision Tree (DT): Represented by a binary tree, where each root node represents an input variable and a split point, and each leaf node contains an output used to make a prediction. Neural Network (NN): Models the way the human brain makes decisions. Each neuron takes in 1+ inputs, and then uses an activation function to process the input with weights and biases to produce an output. Neurons can be arranged into layers, and multiple layers can form a network to model complex decisions. Training the network involves using the training instances to optimize the weights and biases. Naive Bayes (NB): Simplifies the calculation of probabilities by assuming that all features are independent of one another (a strong but effective assumption). Employs Bayes Theorem to calculate the probabilities that the instance to be predicted is in each class, then finds the class with the highest probability. Gradient Boost (GB): Generally used when seeking a model with very high predictive performance. Used to reduce bias and variance (“error”) by combining multiple “weak learners” (not very good models) to create a “strong learner” (high performance model). Involves 3 elements: a loss function (error function) to be optimized, a weak learner (decision tree) to make predictions, and an additive model to add trees to minimize the loss function. Gradient descent is used to minimize error after adding each tree (one by one). Engineering Goal Produce a machine learning model to diagnose Parkinson’s disease given various features of a patient’s speech with at least 90% accuracy and/or a Matthews Correlation Coefficient of at least 0.9. Compare various algorithms and parameters to determine the best model for predicting Parkinson’s. Dataset Description Source: the University of Oxford 195 instances (147 subjects with Parkinson’s, 48 without Parkinson’s) 22 features (elements that are possibly characteristic of Parkinson’s, such as frequency, pitch, amplitude / period of the sound wave) 1 label (1 for Parkinson’s, 0 for no Parkinson’s) Project Pipeline pipeline Summary of Procedure Split the Oxford Parkinson’s Dataset into two parts: one for training, one for validation (evaluate how well the model performs) Train each of the following algorithms with the training set: Logistic Regression, Linear Discriminant Analysis, k Nearest Neighbors, Decision Tree, Neural Network, Naive Bayes, Gradient Boost Evaluate results using the validation set Repeat for the following training set to validation set splits: 80% training / 20% validation, 75% / 25%, and 70% / 30% Repeat for a rescaled version of the dataset (scale all the numbers in the dataset to a range from 0 to 1: this helps to reduce the effect of outliers) Conduct 5 trials and average the results Data a_o a_r m_o m_r Data Analysis In general, the models tended to perform the best (both in terms of accuracy and Matthews Correlation Coefficient) on the rescaled dataset with a 75-25 train-test split. The two highest performing algorithms, k Nearest Neighbors and the Neural Network, both achieved an accuracy of 98%. The NN achieved a MCC of 0.96, while KNN achieved a MCC of 0.94. These figures outperform most existing literature and significantly outperform current methods of diagnosis. Conclusion and Significance These robust results suggest that a machine learning approach can indeed be implemented to significantly improve diagnosis methods of Parkinson’s disease. Given the necessity of early diagnosis for effective treatment, my machine learning models provide a very promising alternative to the current, rather ineffective method of diagnosis. Current methods of early diagnosis are only 53% accurate, while my machine learning model produces 98% accuracy. This 45% increase is critical because an accurate, early diagnosis is needed to effectively treat the disease. Typically, by the time the disease is diagnosed, 60% of nigrostriatal neurons have degenerated, and 80% of striatal dopamine have been depleted. With an earlier diagnosis, much of this degradation could have been slowed or treated. My results are very significant because Parkinson’s affects over 10 million people worldwide who could benefit greatly from an early, accurate diagnosis. Not only is my machine learning approach more accurate in terms of diagnostic accuracy, it is also more scalable, less expensive, and therefore more accessible to people who might not have access to established medical facilities and professionals. The diagnosis is also much simpler, requiring only a 10-15 second voice recording and producing an immediate diagnosis. Future Research Given more time and resources, I would investigate the following: Create a mobile application which would allow the user to record his/her voice, extract the necessary vocal features, and feed it into my machine learning model to diagnose Parkinson’s. Use larger datasets in conjunction with the University of Oxford dataset. Tune and improve my models even further to achieve even better results. Investigate different structures and types of neural networks. Construct a novel algorithm specifically suited for the prediction of Parkinson’s. Generalize my findings and algorithms for all types of dementia disorders, such as Alzheimer’s. References Bind, Shubham. "A Survey of Machine Learning Based Approaches for Parkinson Disease Prediction." International Journal of Computer Science and Information Technologies 6 (2015): n. pag. International Journal of Computer Science and Information Technologies. 2015. Web. 8 Mar. 2017. Brooks, Megan. "Diagnosing Parkinson's Disease Still Challenging." Medscape Medical News. National Institute of Neurological Disorders, 31 July 2014. Web. 20 Mar. 2017. Exploiting Nonlinear Recurrence and Fractal Scaling Properties for Voice Disorder Detection', Little MA, McSharry PE, Roberts SJ, Costello DAE, Moroz IM. BioMedical Engineering OnLine 2007, 6:23 (26 June 2007) Hashmi, Sumaiya F. "A Machine Learning Approach to Diagnosis of Parkinson’s Disease."Claremont Colleges Scholarship. Claremont College, 2013. Web. 10 Mar. 2017. Karplus, Abraham. "Machine Learning Algorithms for Cancer Diagnosis." Machine Learning Algorithms for Cancer Diagnosis (n.d.): n. pag. Mar. 2012. Web. 20 Mar. 2017. Little, Max. "Parkinsons Data Set." UCI Machine Learning Repository. University of Oxford, 26 June 2008. Web. 20 Feb. 2017. Ozcift, Akin, and Arif Gulten. "Classifier Ensemble Construction with Rotation Forest to Improve Medical Diagnosis Performance of Machine Learning Algorithms." Computer Methods and Programs in Biomedicine 104.3 (2011): 443-51. Semantic Scholar. 2011. Web. 15 Mar. 2017. "Parkinson’s Disease Dementia." UCI MIND. N.p., 19 Oct. 2015. Web. 17 Feb. 2017. Salvatore, C., A. Cerasa, I. Castiglioni, F. Gallivanone, A. Augimeri, M. Lopez, G. Arabia, M. Morelli, M.c. Gilardi, and A. Quattrone. "Machine Learning on Brain MRI Data for Differential Diagnosis of Parkinson's Disease and Progressive Supranuclear Palsy."Journal of Neuroscience Methods 222 (2014): 230-37. 2014. Web. 18 Mar. 2017. Shahbakhi, Mohammad, Danial Taheri Far, and Ehsan Tahami. "Speech Analysis for Diagnosis of Parkinson’s Disease Using Genetic Algorithm and Support Vector Machine."Journal of Biomedical Science and Engineering 07.04 (2014): 147-56. Scientific Research. July 2014. Web. 2 Mar. 2017. "Speech and Communication." Speech and Communication. Parkinson's Disease Foundation, n.d. Web. 22 Mar. 2017. Sriram, Tarigoppula V. S., M. Venkateswara Rao, G. V. Satya Narayana, and D. S. V. G. K. Kaladhar. "Diagnosis of Parkinson Disease Using Machine Learning and Data Mining Systems from Voice Dataset." SpringerLink. Springer, Cham, 01 Jan. 1970. Web. 17 Mar. 2017.

The goal of this survey is two-fold: (i) to present recent advances on adversarial machine learning (AML) for the security of RS (i.e., attacking and defense recommendation models), (ii) to show another successful application of AML in generative adversarial networks (GANs) for generative applications, thanks to their ability for learning (high-dimensional) data distributions. In this survey, we provide an exhaustive literature review of 74 articles published in major RS and ML journals and conferences. This review serves as a reference for the RS community, working on the security of RS or on generative models using GANs to improve their quality.

Automatically literature survey/review with GPT! An intelligent research assistant leveraging GPT-3.5 /GPT-4 to find, analyze, and rank relevant academic papers from Google Scholar based on user-provided search queries and topics

Literature Survey of Information Extraction, especially Relation Extraction, Event Extraction, and Slot Filling.

A comprehensive, categorized collection of hundreds of research papers and surveys in Machine Learning and Natural Language Processing. This repository organizes papers by topics with direct links for easy access. Continuously updated to help researchers, students, and practitioners quickly find relevant literature across the ML landscape.

A book surveying the literature on user interface software and technology.

Literature survey of order execution strategies implemented in python

The project that we worked on this summer internship falls in the domain of research in IoT (Internet of Things). Initially, the mentor asked us to find real-life problems, which we would attempt to solve by using the tools of Information Technology. We were allowed to discuss and work in a group of three. We picked the problem of devising an attendance monitoring system, which would mark the presence of the students in a big room, in a non-intrusive manner using image recognition, for e.g. an auditorium or our college’s lecture theatre. Our project was divided into two phases, which would be illustrated in the subsequent passages. The first phase involved doing a literature survey on the tools and technologies through various authentic research papers and the existing libraries, which would enable us to devise a backend structure for our project. We, then developed a flowchart, which comprised of two modules of processes, through which the procedure would pass through. The first module involves the initial training of a machine learning based classifier by training it with the various images of a specific person. The second module involves the testing part in the real environment, which involves face detection and face recognition. A camera would take the frames/image of a live audience. Then, these frames would be pre-processed (involves grey-scaling and image resizing) for achieving better performance in the subsequent face detection module. The face-detection algorithm would detect all the faces present in the frame, and would crop the detected faces, and would pass them to the face recognition classifier for testing. The classifier would classify the cropped images and would mark the attendance accordingly. The libraries used for face-detection were that of OpenCV, and a convolutional neural network was trained for the image recognition part. The libraries which were used for training the convolutional neural network was Keras. The second phase involved the implementation part, where we had to gather the data for training the neural network, and find out the parameters of the image, for which we are getting better accuracy performance. We trained the neural network with the images of about 64 students, with about 20 images per student, covering different angles and brightness levels. We trained the network with 70 percent of the image corpus, and used the remaining 30 percent for testing. We got an accuracy of 93 percent. For testing the face detection part, we took a video of a classroom of about 40 students. Then, we generated frames from the video and passed it to the face detection algorithm. We extrapolated that the accuracy of an individual frame was not that high, but if we consider all the detected members in all the frames, we are covering almost every student. Hence, considering multiple frames for testing is crucial to get a high detection accuracy. We are currently trying to figure out the camera and its mounting position, which would be conducive for the algorithm, to give us accurate results.

TMLR | This survey presents a comprehensive and structured synthesis of memory in LLMs and MLLMs, organizing the literature into a cohesive taxonomy comprising implicit, explicit, and agentic memory paradigms.

Semi-automatic Literature Survey (SaLS)

Literature Survey about the available Diffusion Models capable of Time-Series Forecasting

A conversational agent or a chatbot is piece of software which can communicate with human users with the help of natural language processing (NLP). Modelling conversation is a very crucial task in natural language processing and artificial intelligence (AI). Since the discovery of artificial intelligence, creating a good chatbot is one of the field’s hardest and complex challenges. Chatbots can be used for various tasks such as make phone calls, provide reminders etc; in general they have to understand users’ utterances and provide relevant responses for the problem in hand. Previously, methods which were used for constructing chatbot architectures relied on hand-written rules, templates or simple statistical methods. Rising and innovating field of deep learning have replaced previous models with trainable neural network models. The recurrent encoder-decoder model is the dominating model in the field modelling conversations. Multiple variations and features have been presented that have changed the quality of the conversation that chatbots are capable of. In our project, we have surveyed recent literatures published, examining various publications related to chatbots. We started with taking Cornell movie dialogue corpus as our dataset then after training our model with it and fine tuning it with various parameters, non-satisfactory results lead us to take another dataset and we trained and tested our final model on modified Gunthercox dataset which gave us satisfactory results for an open domain chatbot or general domain chatbot.

Twitter tweets play an important role in every organisation. This project is based on analysing the English tweets and categorizing the tweets based on the sentiment and emotions of the user. The literature survey conducted showed promising results of using hybrid methodologies for sentiment and emotion analysis. Four different hybrid methodologies have been used for analysing the tweets belonging to various categories. A combination of classification and regression approaches using different deep learning models such as Bidirectional LSTM, LSTM and Convolutional neural network (CNN) are implemented to perform sentiment and behaviour analysis of the tweets. A novel approach of combining Vader and NRC lexicon is used to generate the sentiment and emotion polarity and categories. The evaluation metrics such as accuracy, mean absolute error and mean square error are used to test the performance of the model. The business use cases for the models applied here can be to understand the opinion of customers towards their business to improve their service. Contradictory to the suggestions of Google’s S/W ratio method, LSTM models performed better than using CNN models for categorical as well as regression problems.

This project is regarding Motion Planning for Autonomous Vehicle under uncertainty. There are various kinds of uncertainty for an autonomous vehicle/autonomous robot like: Uncertainty in system configuration, Uncertainty in the system model, Uncertainty in environmental situations​, Uncertainty in Future environment state.(Dynamic obstacle)​. So, as to deal with these uncertainties, motion planner should plan by incorporating these uncertainties. We majorly focus on dynamic obstacle uncertainty. After doing literature survey, we found that sampling-based approach scale well with respect to deal with uncertainty. So, we studied and explored following RRT variants which are use to deal with uncertainty: RRBT(Rapidly exploring Belief trees)​ Chance Constrained RRT(CC-RRT)​ RRT-FNDynamic()​ RRT-X We have successfully implemented RRT for Ackerman based vehicle for static obstacle.(You can see the implementation in the below video) We finally implemented RRT* as global planner and Dynamic Window approach algorithm as local planner. RRT* returned us a optimal path, we fed the path to our local planner. We simulated our dynamic moving obstacle in Pygame by incorporating random uncertain moving obstacles. Our local planner is successfully able to avoid uncertain moving dynamic obstacles and it reached it's goal destination without colliding to any of the obstacles. You can see the demo of our implementation in the below video which is at the right hand side and project presentation video at the left hand side.

My survey of literature on compilers for AI models

A template to create your own literature survey engine

Claude Code skill for citation-verified literature surveys in political science & computational social science

De-Anthropocentric Research Engine — AI-powered academic research automation with deep literature survey, gap analysis, idea generation, experiment design & execution. Combines iterative deep research, adversarial debate, evolutionary generation, and distributed GPU execution.

Literature survey of convex optimizers and optimisation methods for deep-learning; made especially for optimisation researchers with ❤️

There's a good amount of literature on shot boundary detection, and there's a multitude of techniques that work successfully for hard-cuts, and more modern techniques like neural networks for fades/dissolves, wipes, etc. In this repository, I survey and implement/use tools to detect cuts with different kinds of content.

BDP 05: CLUSTERING OF LARGE UNLABELED DATASETS OVERVIEW Real world data is frequently unlabeled and can seem completely random. In these sort of situations, unsupervised learning techniques are a great way to find underlying patterns. This project looks at one such algorithm, KMeans clustering, which searches for boundaries separating groups of points based on their differences in some features. The goal of the project is to implement an unsupervised clustering algorithm using a distributed computing platform. You will implement this algorithm on the stack overflow user base to find different ways the community can be divided, and investigate what causes these groupings. The clustering algorithm must be designed in a way that is appropriate for data intensive parallel computing frameworks. Spark would be the primary choice for this project, but it could also be implemented in Hadoop MapReduce. Algorithm implementations from external libraries such as Spark MLib may not be utilised; the code must be original from the students. However, once the algorithm is completed, a comparison between your own results and that generated by MLlib could be interesting and aid your investigation. Stack Overflow is the main dataset for this project, but alternative datasets can be adopted after consultation with the module organiser. Additionally, different clustering algorithms may be utilised, but this must be discussed and approved y the module organiser. DATASET The project will use the Stack Overflow dataset. This dataset is located in HDFS at /data/stackoverflow The dataset for StackOverflow is a set of files containing Posts, Users, Votes, Comments, PostHistory and PostLinks. Each file contains one XML record per line. For complete schema information: Click here In order to define the clustering use case, you must define what should be the features of each post that will be used to cluster the data. Have a look at the different fields to define your use case. ALGORITHM The project will implement the k-means algorithm for clustering. This algorithm iteratively recomputes the location of k centroids (k is the number of clusters, defined beforehand), that aim to classify the data. Points are labelled to the closest centroid, with each iteration updating the centroids location based on all the points labelled with that value. Spark and Map/Reduce can be utilised for implementing this problem. Spark is recommended for this task, due to its performance benefits in . However, note that the MLib extension of Spark is not allowed to be used as the primary implementation. The group must code its own original implementation of the algorithm. However, it is possible to also use the mllib implementation, in order to evaluate the results from each clustering implementation. Report Contents Brief literature survey on clustering algorithms, including the challenges on implementing them at scale for parallel frameworks. The report should then justify the chosen algorithm (if changed) and the implementation. Definition of the project use case, where the implemented project will be part of the solution. Implementation in MapReduce or Spark of a clustering algorithm(KMeans). Must take into account the potential enormous size of the dataset, and develop sensible code that will scale and efficiently use additional computing nodes. The code will also need to potentially convert the dataset from its storage format to an in-memory representation. Source code should not be included in the report. However, the algorithms should be explained in the report. Results section. Adequate figures and tables should be used to present the results. The effectiveness of the algorithm should also be shown, including performance indications. Not really sure if this can be done for clustering. Critical evaluation of the results should be provided. Experiments demonstrating the technique can successfully group users in the dataset. Representation of the results, and discussion of the findings in a critical manner. ASSESSMENT The project according to the specification has a base difficulty of 85/100. This means that a perfect implementation and report would get a 85. Additional technical features and experimentation would raise the difficulty in order to opt for a full 100/100 mark. Report presentation: 20% Appropriate motivation for the work. Lack of typos/grammar errors, adequate format. Clear flow and style. Related work section including adequate referencing. Technical merit: 50% Completeness of the implementation. [25%] Provided source code. Code is documented. [10%] Design rationale of the code is provided. [10%] Efficient, and appropriate implementation for the chosen platform. [5%] Results/Analysis: 30% Experiments have been carried out on the full dataset. [10%] Adequate plots/tables are provided, with captions. [10%] Results are not only presented but discussed appropriately. [10%] Additional project goals: Implementation of additional functions beyond the base specification can raise the base mark up to 100. A non-exhaustive list of expansion ideas include: Exploration and discussion of hyperparameter tuning (e.g. the number of k groups to cluster the data into) [up to 10 marks] Comparative evaluation of clustering technique with existing implementations (e.g. mllib) [up to 10 marks] Bringing in additional datasets from stackoverflow, such as user badges, to aid in clustering [up to 5 marks] Cluster additional datasets (such as posts) [up to 10 marks] LEAD DEMONSTRATOR For specific queries related to this coursework topic, please liaise with Mr/Ms TBD, who will be the lead demonstrator for this project, as well as with the module organiser. SUBMISSION GUIDELINES The report will have a maximum length of 8 pages, not counting cover page and table of contents. The report must include motivation of the problem, brief literature survey, explanation of the selected technique, implementation details and discussion of the obtained results, and references used in the work. Additionally, the source code must be included as a separate compressed file in the submission.

The data and code for our CHI2021 accessibility literature survey paper

This repository provides the protocols and tools necessary for AI Agents (such as GitHub Copilot, Cursor) to conduct fully automated, deep, and systematic literature surveys.

Literature survey of scene graphs

A comprehensive, categorized collection of hundreds of research papers and surveys in Large Language Models. This repository organizes papers by topics with direct links for easy access. Continuously updated to help researchers, students, and practitioners quickly find relevant literature across the LLM landscape.

AI驱动的系统性学术文献回顾工具 | AI-Powered Literature Survey Tool with 8-Phase Workflow & Agent Swarm Architecture

Malicious URL, a.k.a. malicious website, is a common and serious threat to cyber-security. Malicious URLs host unsolicited content (spam, phishing, drive-by downloads, etc.) and lure unsuspecting users to become victims of scams (monetary loss, theft of private information, and malware installation), and cause losses of billions of dollars every year. It is imperative to detect and act on such threats in a timely manner. Traditionally, this detection is done mostly through the usage of blacklists. However, blacklists cannot be exhaustive, and lack the ability to detect newly generated malicious URLs. To improve the generality of malicious URL detectors, machine learning techniques have been explored with increasing attention in recent years. This project aims to provide a comprehensive survey and a structural understanding of Malicious URL Detection techniques using machine learning. i present the formal formulation of Malicious URL Detection as a machine learning task, and categorize and review the contributions of literature studies that addresses different dimensions of this problem (feature representation, algorithm design, etc.).