Tsinghua 347 Cognitive Intelligence
Machine Learning
Tsinghua 347
The Tsinghua 347 Cognitive Intelligence program is organized and coordinated by the Department of Psychology and Cognitive Science at Tsinghua University, supported by a strong foundation in “Psychology + Artificial Intelligence”. The target audience includes professionals from Internet, Computer Science, Clinical Consulting, and Intelligent Manufacturing or Psychology who wish to deeply integrate knowledge of psychology and artificial intelligence with a certain technical background.
In this psychology-oriented master’s program, the relevant knowledge of artificial intelligence is an unavoidable topic. Today, I, Sister Xiao Zhi, will discuss machine learning in artificial intelligence with everyone.
1
What is Machine Learning?
Machine learning is a type of artificial intelligence technology that aims to enable computers to learn from data without explicit programming. If the performance P of a computer program on task T improves with the accumulation of experience E, it is considered capable of learning from experience (i.e., its performance optimizes as experience accumulates).
Machine learning sits at the intersection of computer science and statistics. It automatically discovers patterns and rules in large datasets by analyzing them and uses these rules for prediction, classification, clustering, and other tasks.
2
Applications of Machine Learning
The applications of machine learning are vast, including virtual assistants, e-commerce, healthcare, and financial media. It has demonstrated strong potential and practical value across various fields. Here are some more specific application scenarios:
Virtual Assistants: Virtual assistants use natural language processing technology to understand user intentions and context, providing more accurate feedback. By analyzing user behavior and preferences, virtual assistants can offer personalized suggestions and services, such as managing schedules and recommending products.
E-commerce: By analyzing users’ historical behavior data (such as purchase records and browsing history), machine learning models can predict products that users may be interested in and make personalized recommendations, improving user conversion rates and satisfaction. By utilizing machine learning models, e-commerce companies can better target their customer base, develop personalized marketing strategies, enhance marketing effectiveness, and reduce costs.
Healthcare: Machine learning technologies can accelerate the discovery and development of new drugs by analyzing chemical structures and biological data to predict compound bioactivity. By analyzing patients’ lifestyles and genetic information, machine learning models can provide personalized health advice and assess the risk of future diseases.
Financial Media: Provide personalized investment advice and portfolio optimization based on clients’ risk preferences and return objectives. By analyzing client data, precise risk assessments and fraud detection can be performed.
3
Common Algorithms in Machine Learning
Linear Regression: Used for predicting continuous values, such as housing price predictions.
Logistic Regression: Used for classification problems, such as spam detection.
Support Vector Machine (SVM): Used for classification and regression tasks, especially suitable for high-dimensional data.
K-Nearest Neighbors (KNN): Used for classification and regression, predicting by calculating the distance to training samples.
Decision Trees: Used for classification and regression, making decisions through a tree structure.
Random Forest: An ensemble model consisting of multiple decision trees, known for its high accuracy and robustness.
Naive Bayes: A classification algorithm based on Bayes’ theorem, suitable for tasks like text classification.
K-Means Clustering: An unsupervised learning algorithm used to partition data into K clusters.
Principal Component Analysis (PCA): Used for dimensionality reduction and feature extraction.
Neural Networks: Includes Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), and Recurrent Neural Networks (RNN), widely used in fields like image recognition and speech recognition.
4
Classification of Machine Learning
1
Supervised Learning:
Supervised learning is a method of machine learning that trains models using labeled datasets for classification or prediction. In supervised learning, each training sample contains input features (feature values) and corresponding target outputs (labels). The model learns the relationship between these inputs and outputs to make predictions and classifications on new, unlabeled data.
2
Unsupervised Learning:
Unsupervised learning is a machine learning method that uses unlabeled data for training. Unlike supervised learning, the training data in unsupervised learning does not have pre-defined labels or outputs, only input features without corresponding labels or target outputs. The main goal is to discover hidden patterns or structures within the data.
3
Self-Supervised Learning:
Self-supervised learning is a machine learning method that utilizes unlabeled data to automatically generate labels through the design of auxiliary tasks for model training. This method does not rely on manually labeled data but generates supervisory signals through the characteristics of the data itself to train without explicitly labeled data.
4
Semi-Supervised Learning:
Semi-supervised learning is a machine learning method that combines a small amount of labeled data and a large amount of unlabeled data for training. This method lies between supervised and unsupervised learning, aiming to leverage unlabeled data to enhance the model’s learning ability and performance, reducing the costs of data collection and labeling while improving accuracy and interpretability.
5
Reinforcement Learning:
Reinforcement learning is a machine learning method aimed at learning optimal policies through interaction with the environment to achieve specific goals or maximize rewards. In reinforcement learning, the agent continuously optimizes its behavior through trial and error and delayed rewards. When the agent takes an action in the environment, the environment returns a reward signal indicating the quality of that action. The agent learns an optimal policy through interaction with the environment.
5
Challenges and Future
Since the inception of machine learning technology, it has gradually changed people’s lives with its efficient learning imitation and problem-solving capabilities, but it also faces some practical challenges:
Data Quality and Availability:
High-quality data is key to the success of machine learning models. However, data may contain noise, be incomplete, biased, or unrepresentative, all of which can affect the model’s performance and generalization ability.
Model Interpretability:
As models become increasingly complex, especially deep learning models, understanding the decision-making process of the model becomes more difficult. Model interpretability is critical for building user trust, conducting error analysis, and complying with regulations.
Overfitting and Generalization Ability:
Models may overly learn specific patterns in the training data, leading to poor performance on new, unseen data. Improving the generalization ability of models is an ongoing challenge.
While facing various challenges, machine learning also presents a unique future landscape:
Multimodal Large Models: Future AI will not be limited to processing language and visual data but will also integrate audio, 3D, and other modalities, enhancing AI efficiency and accuracy in practical applications.
Embodied Intelligence: Embodied intelligence will further mature, applying to more industrial scenarios and humanoid robots, promoting AI applications in the real world.
Synthetic Data: Synthetic data will become an important source to supplement training data, reducing data labeling costs and alleviating data privacy issues.
AI Agents: More general and autonomous agents will penetrate work and life scenarios, becoming an important application form of AI products.
Generative Search: AI will shift from traditional keyword searches to generating answers, improving information retrieval efficiency while adjusting the content production ecosystem.
6
Study Material Recommendations
As we all know, choosing the right reference books for the graduate entrance exam is crucial. Tsinghua 347 is a relatively novel academic direction, and the plethora of artificial intelligence reference books on the market can be overwhelming.
With this in mind, the Shui Mu Shang An teaching and research group has carefully compiled the Artificial Intelligence Lecture Notes (2nd Edition) for preparing and studying for the Tsinghua 347 Cognitive Intelligence Applied Psychology Master’s program, gathering the strengths of various reference books, printed in color, explaining artificial intelligence-related knowledge points in a visually rich format. Illustrations, text, and formulas work together to aid understanding, combined with comprehensive explanations to clarify complex knowledge points with the simplest language.
Moreover, cognitive intelligence is not only artificial intelligence but also psychology. This book explains artificial intelligence through analogies with psychological knowledge, accurately covering exam points.
Follow Sister Xiao Zhi for more information on Tsinghua 347
Previous Exciting Recommendations
