[ИАД, 2026] Предзащита магистерских работ

[ИАД, 2026] Предзащита магистерских работ

by Machine Learning – Intelligent Systems

Formal & Physical Sciences AI — Models & Research UYQArtificial intelligence (AI)UYQMMachine learning
🎙 Machine Learning – Intelligent Systems 👥 8K 📅 June 13, 2026 ⏱ 173 min 👁 330 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

mixture of expertssurrogate functionoptimal controlgenerative replaystochastic optimization

Summary

The video is a recording of a pre-defense session for master’s theses at a Russian institution, likely MIPT. Three students present their work. The first, Pyotr Galkin, presents a method for architecture search in mixture of experts using a surrogate function to jointly optimize routing and expert architectures. He proves a convergence theorem and shows experiments on synthetic and real data. The second, Andrey, presents work on optimal control in AI systems with feedback loops, including a system degeneration example and generative replay for continual learning, with theoretical bounds and experiments on Chest X-ray. The third student presents research on stochastic optimization methods, including Frank-Wolfe variants and decentralized algorithms. The committee provides detailed feedback on presentation structure, terminology, and theoretical clarity, emphasizing the need for simpler formulations and better slide design.

131 words

Critical Evaluation

Value of the Information & Strength of the Argument

The presentations offer original theoretical contributions, such as a new formulation for mixture of experts optimization and theoretical guarantees for control in feedback systems. The argumentation is generally solid, with proofs and experiments supporting claims. However, the presentations are dense and sometimes lack clarity, as noted by the committee. The value lies in the novel theoretical insights and potential applications, but the practical impact is not fully demonstrated.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is moderate. The students present original work, but the sources are not explicitly cited in the video. The committee’s feedback highlights issues with terminology and presentation clarity, which affect the overall rigor. The title accurately reflects the content, and the video is a legitimate academic event. No external sources are mentioned, so the reliability relies on the expertise of the presenters and committee.

149 words

Title / Content Match

The title accurately reflects the content: a pre-defense session of master's theses.

Quality & Reliability

7/10

The video is a pre-defense session of master's theses, featuring presentations by students and feedback from a committee. The content is largely theoretical, with presentations on mixture of experts, optimal control in AI systems, and stochastic optimization methods. The committee provides critical feedback on presentation clarity and theoretical rigor. The scientific quality is moderate to high, with original contributions presented, but the format is informal and not peer-reviewed.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Contribution & Novelties

The video presents original research contributions from master’s students, including a novel method for architecture search in mixture of experts, theoretical analysis of control in AI feedback loops, and new stochastic optimization algorithms. The main novelty is the theoretical grounding and proofs provided for these methods.

Pour aller plus loin :

82 words

Radar Profile

The radar profile shows high scores in quantity and technical level, indicating a dense and technical presentation. Quality and reliability are moderate, reflecting the informal setting and lack of external sources. The overall balance suggests a content-rich but not fully polished presentation.

Reliability 6/10

💬 No comments were provided for analysis.