Introduction

Christoph Lippert heads the "Digital Health - Machine Learning" group, where he drives innovation at the intersection of AI and medicine. His research focuses on developing advanced algorithms to detect disease patterns in medical images and molecular data, along with statistical tools for analyzing large populations. By combining cutting-edge imaging, DNA sequencing, and machine learning, his work is transforming early diagnosis and paving the way for data-driven precision medicine.

Overview of the lecture

Good scientific writing is the key to communicating research effectively, yet many papers struggle with issues like unclear phrasing, poor organization, and excessive complexity, which obscure meaning and reduce impact. This lecture aimed to address these challenges by emphasizing the importance of clear, precise writing and showcasing common mistakes through practical examples. By identifying and avoiding these pitfalls, attendees gained valuable insights and strategies to craft well-structured, engaging, and impactful scientific manuscripts. Professor Lippert drew from his own experience with poorly written submissions during paper reviews, using these examples to demonstrate and correct common writing flaws in class.

Novel solution to Travelling Salesman problem - A bad paper

Abstract

The Traveling Salesperson Problem (TSP) is an incredibly fascinating, and perhaps the most important, problem that exists in computer science today. It represents a critical testbed for the development of numerous, highly effective optimization algorithms that continue to transform various fields.

Applications

TSP applications are nearly infinite—delivery routing, DNA sequencing, travelling itineraries, spacecraft navigation, and even internet packet routing. The number of applications is likely impossible to determine because the solutions are unquantifiable in real-world contexts. More than 20.34% of modern industries use TSP.

Introduction

The TSP has been around forever and is completely essential to understand for anyone who is seriously interested in pursuing computer science. It involves finding the shortest route between multiple cities, which is obviously super hard because there are so many combinations. Many researchers over time have declared TSP to be the pinnacle of computational challenges. Indeed, solving the TSP perfectly would solve every routing and logistics challenge ever faced in history.
As a cornerstone of optimization, this problem has had a gigantic impact on industry, science, and the arts. Moreover, there are over 1.7 trillion instances of TSP solved every year according to some studies. There is no consensus on the best method to approach the problem — various algorithms like Genetic Algorithms (GA), Simulated Annealing (SA), Ant Colony Optimization (ACO), and Machine Learning (ML) have all been tried and tested in innumerable case studies. The figure shows a generic graph representation of TSP while the other figure illustrates a detailed algorithm for solving TSP (figures omitted for clarity). Note that the different colours represent various stages of the algorithm execution.

Mathematical Formulation

The mathematical equation of TSP can be shown in a simple way:
where:

• is the cost
• is distance
• is the permutation

This equation alone, it could be argued, encapsulates the beauty of the problem. By extension, equations as expressed above define modern computer science.
Another version: Minimize the path-length function by using either exhaustive search, heuristic methods, or hybridized metaheuristics. Either way, the researcher's interpretation of this equation matters (this will be shown in Table 2).

Our novel solution

We propose a groundbreaking approach to solving the TSP that combines elements of traditional heuristics with advanced methodologies. The solution is inspired by nature, leveraging principles of swarming behavior, chaos theory, and quantum computation. By integrating these diverse concepts, the algorithm provides unparalleled flexibility and efficiency. To achieve this, we start with a unique initialization strategy that optimally selects starting points based on an adaptive heuristic framework. Subsequently, a novel path evaluation mechanism dynamically adjusts priorities based on a feedback loop influenced by real-time metrics. The hybridization of these methods results in a synergistic effect that enhances solution quality. Although preliminary results suggest significant improvements over existing methods, further studies are necessary to fully quantify the benefits and understand the trade-offs. Detailed comparisons with existing techniques will be presented in future work.

Results

The table shows that in the preliminary experiments, our algorithm outperformed the existing methods in terms of solution quality (14.551%) and convergence speed (36.0944%). Tests on modern GPUs showed that the algorithm can solve TSP instances with up to 10000 cities in 0.266 seconds.

Discussion

Despite the obvious progress in solving TSP, the challenges remain. Analyzing large datasets for city connections is uniquely difficult for algorithms. Researchers such as Smith et al. (2001), Johnson et al. (1999), and many others have written extensively about this. The algorithms, as shown in Table 3, have varying trade-offs. However, there is no clear answer as to which algorithm is truly the best.
Machine learning methods might solve TSP faster in the future (see 2.3.4). Practical applications of heuristic methods remain unclear due to logistical constraints.

Conclusion

The Travelling Salesman Problem—an incredibly amazing problem—still fascinates and puzzles scientists. It cannot be overstated how relevant it remains, nor how revolutionary its potential solutions are. The methods explored in this paper show significant promise.

References

[1] Smith et al., 2001. A study of TSP.
[2] Johnson et al., 1999. Solving TSP with some algorithms.
[3] Random Author, 2020. Insights into TSP solutions.

Main results

Based on the paper above, let's take a closer look at the most common mistakes made when writing academic papers.

Empty Filler Words & Subjective Claims

A major problem in many scientific papers is the use of words or entire sections of text that either contribute nothing to the content of the paper or contain unsubstantiated subjective statements.

Tips How to Improve

• Provide data or sources to back up claims.
• Use descriptive and neutral language.

Changes we can apply to our paper

In relation to the above text, the following is noticeable:

"The Traveling Salesperson Problem (TSP) is an incredibly fascinating, and perhaps the most important, problem that exists in computer science today."

We can easily fix this by using neutral and fact-based descriptions and deleting unnecessary parts of the sentence. Adjusted, the section of our text would look like this:

"The Traveling Salesperson Problem (TSP) is a significant optimization problem in computer science."

Unsubstantiated Statements

A similar point of error is statements without evidence in the form of quotes or proof.

Tips How to Improve

• Assertions should always be substantiated.
• If there is no evidence for claims, they should be deleted or at least softened.

Changes we can apply to our paper

In our paper, for example, we should change the following sentence:

"Indeed, solving the TSP perfectly would solve every routing and logistics challenge ever faced in history."

The sentence could look more scientifically correct as follows:

"A comprehensive solution to the TSP could significantly improve routing and logistics efficiency, as indicated in studies exploring its practical applications [1][2]."

Overly Complex Sentences

Especially in the German language, nested sentences often occur when writing scientific papers. These can lead to comprehension problems for the reader and distract from the main points of the paper.

Tips How to Improve

• Divide longer sentences into several shorter sentence.
• No more than one main idea per sentence.

Changes we can apply to our paper

This source of error is not limited to the German language, as this example from our paper shows:

"There is no consensus on the best method to approach the problem — various algorithms like Genetic Algorithms (GA), Simulated Annealing (SA), Ant Colony Optimization (ACO), and Machine Learning (ML) have all been tried and tested in innumerable case studies."

In our example we could split the given sentence into two. Additionally we should add references. Revised, the sentence could read as follows:

"There is no consensus on the best method to approach the TSP. Various algorithms such as Genetic Algorithms (GA) [1], Simulated Annealing (SA) [2], Ant Colony Optimization (ACO) [3], and Machine Learning (ML)[4] have been researched and tested."

Reporting Numbers & Statistics

Figures and statistics are a popular means of substantiating statements in scientific papers. However, there are a few important points to bear in mind.

Tips How to Improve

• Use an appropriate number of significant figures.
• If possible, include statistical values such as standard deviation, confidence intervals, and p-values.
• Explain the statistical methods and models behind the given numbers.

Changes we can apply to our paper

With regard to our paper above, the following negative example can be found:

"The table shows that in the preliminary experiments, our algorithm outperformed the existing methods in terms of solution quality (14.551%) and convergence speed (36.0944%)."

The inclusion of a confidence interval and also of significance tests leads to a more correct classification of the figures presented and makes the work more scientifically accurate:

"Table X shows that in the preliminary experiments, our algorithm outperformed the stimmulated anealing algorithm [1] in terms of solution quality (14.6%) and convergence speed (36.1%). Further validation is required to confirm these results, including statistical significance tests and confidence intervals."

Citation Salad - Figures & Tables

Citations are usually an important part of scientific work. However, they must be used sensibly.

Tips How to Improve

• Explain why citations are used.
• Group citations by subject, method or other overarching common denominator.

Changes we can apply to our paper

For example, in the above paper, works are cited in the following sentence, but it is not clear to the reader exactly what contribution the cited works make to the paper at hand:

"Researchers such as Smith et al. (2001), Johnson et al. (1999), and many others have written extensively about this."

A better formulation with a clear classification of the quoted contributions would be the following:

"Smith et al. (2001) provided foundational insights into heuristic approaches for TSP, while Johnson et al. (1999) analyzed algorithmic trade-offs in practical applications. These studies offer a basis for comparing our novel approach to existing methods."

Inconsistent Terminology

Using different terms for the same concept can confuse readers and, in the worst case, even distort the content.

Tips How to Improve

• Introduce relevant concepts once and use them consistently.
• Use a glossary if you use many recurring key terms and concepts.
• Use synonyms sparingly and make it clear what they stand for.

Changes we can apply to our paper

In the present example, the Traveling Salesperson Problem (TSP) is mentioned at the beginning. However, the title and the conclusion both speak of the Traveling Salesman Problem. The following sentence should therefore be adapted:

"The Traveling Salesperson Problem (TSP) is an incredibly [...]".

Instead, the text should begin as follows and talk about the Travelling Salesman Problem:

"The Traveling Salesman Problem (TSP) is an incredibly [...]".

Poor Structure

When a paper does not follow a clear structure, readability can be impacted and it can become hard to understand. Furthermore, unstructured paragraphs and narrative can make it harder for the reader to find the logical flow presented in the paper.

Tips How to Improve

• Follow a clear and standardized structure

• When in doubt, write as follows: Introduction->Methods->Results->Conclusion
• Introduce the problem, then the methods, show the results and conclude with additional research
• Replace the Methods section with a Methods summary and add a more detailed Methods section after the Conclusion to make the paper easier to grasp for a wider audience (those especially interested can read all details at the end)
• Sections should always built on top of previous ones

Changes we can apply to our paper

Our paper does not seem to follow a clear structure. For example, we are discussing applications of the problem before clearly introducing it. We could restructure our paper introducing the standard structure presented above. Concretely, we could repackage our solution paragraph into a Methods section and split our Discussion section. We could then integrate the parts discussing our results into the Results section and move the parts discussing the future of TSP into the conclusion.

Unclear Contributions

The main reason for publishing a paper is to present advancements in a specific field. It is thus important to highlight those advancements and improvements over previous work to make the reader understand which parts of the paper present new methods.

Tips How to Improve

• Explicitly state the paper's main contributions

• Briefly in the Abstract
• Detailed in the Introduction
• Summarized in the Conclusion

• Directly compare new work to existing one to highlight improvements and advancements

Changes we can apply to our paper

Currently our paper has a distinct lack of reference and comparison. We only briefly mention existing work in our Discussion section and do not compare our work to this existing work. We could improve our paper by also mentioning existing work in The Abstract and Introduction and directly compare our novel solution to previous ones, pointing out improvements.

Information Overload

While a lack of information or structure thereof can confuse the reader, the same can happen when too much information is presented. This overload can distract the reader and may lead to them missing key-points and important information presented in the paper.

Tips How to Improve

• Only present information closely related to the research question

• While additional information can be included, it should be moved away from the core-message as not to distract the reader (i.e. to Appendices)

• Visualizations can sometimes be used to more effectively present information
• When there is too much information for one paper, write another one

Changes we can apply to our paper

Introduction of TSP lists numerous examples and usage statistics which, while not adding useful information, might distract the reader from understanding its basic workings. Since this information only serves to show the importance of the TSP, it would make sense to omit the majority of it.

Equations as Sentences

When reading paragraphs in a paper, the reader usually expects to read well-formatted text. Not treating equations as part of this text can impact the reading-flow.

Tips How to Improve

• Treat equations as part of a sentence

• Punctuation should be used with equations as it would be in normal sentences
• For instance, when an equation ends a sentence, use a period

• Reference them (correctly - see below) in the surrounding text when explanation is necessary

Changes we can apply to our paper

In the Mathematical Formulation section, our equation completely breaks our formatting. We could move our equation into the surrounding paragraph to improve the reading-flow and add explanation if needed.

Equations With or Without Numbers?

If equations need to be referenced later on, there should be a way to easily find them. Reference numbers can be added to more easily reference equations later on.

Tips How to Improve

• Number all equations that are referenced

• Numbering is optional; Don't number equations that are not referenced
• Few numbers help but too many will cause confusion, so use them sparingly

• Always ensure that all numbered equations are referenced

Changes we can apply to our paper

While in the Mathematical formulation section, referencing our equation without a number directly after its definition is arguably sufficient, we later reference the same equation at the end of this paragraph. This could potentially confuse the reader. Therefore, we can number our equation to reference it more easily.

American or British English?

A professionally written paper should of course also use consistent spelling. This is easier for native speakers but especially those who were taught British spelling might sometimes make spelling mistakes when writing a paper in American English.

Tips How to Improve

• Decide on one spelling and stick to it (SET YOUR SPELLCHECKER CORRECTLY!)

• Decide based on fields and conferences
• Computer Science papers are commonly written in American English

Changes we can apply to our paper

It seems that while writing the paper, we did not use proper spellchecking since some words like 'optimization' and 'real-world' use the American spelling while others like 'colour' and 'Travelling' use British spelling. We should thus decide between British or American English (we would most likely use American English since this is a Computer Science paper) and of correct incorrectly spelled words.

Summary

Evaluating the quality of a paper will always involve some level of subjectivity. However, several techniques and checks can significantly enhance the clarity, authority, and professionalism of a scientific work.
Objective and precise writing makes a paper more professional, while providing sufficient evidence for all claims ensures that the proposed improvements are verifiable and credible. Including relevant statistics and comparing results to established references further strengthens the reliability of the work. When building upon existing research, it is essential to acknowledge the prior work and clearly highlight the advancements made.

When presenting findings, it is important to avoid overwhelming the reader with excessive information. If necessary, key results should be prioritized in the main text and supplementary details should be moved to appendices.
To improve readability, several additional considerations should be kept in mind. Figures and tables are effective tools for presenting results but should be self-explanatory and not overly reliant on external text. Consistent terminology is critical; if different terms are used to describe the same concept, a glossary should be provided for clarity. Similarly, the structure of the paper should be consistent and logically organized to facilitate a smooth reading experience.

Equations, while integral to many papers, should be seamlessly integrated into the text to avoid disrupting the narrative flow. Whenever equations are referenced later in the text, numbering them improves accessibility and reduces confusion.
Inconsistent spelling can detract from the paper's professionalism. It is crucial to adhere to either British or American English consistently throughout the document.

While the overall quality of a paper primarily depends on the significance and novelty of its contributions, adhering to these guidelines ensures that the advancements are communicated effectively, leaving a lasting impression on the reader.

Hybrid Solution to the Traveling Salesman Problem - a better paper

Abstract

The Traveling Salesman Problem (TSP) is a significant optimization problem in computer science. This paper presents a novel algorithm combining heuristic methods with principles of swarming behavior, chaos theory, and quantum computation. Preliminary results demonstrate improvements in solution quality and computational efficiency.

Introduction

The Traveling Salesman Problem (TSP) involves finding the shortest route connecting multiple cities while visiting each exactly once and returning to the starting point. This problem has applications in fields such as logistics, DNA sequencing, and internet routing. Various methods, including Genetic Algorithms (GA) [6], Simulated Annealing (SA) [5], and Ant Colony Optimization (ACO) [1], have been extensively studied for solving TSP. A comprehensive solution to the TSP could significantly enhance routing and logistics efficiency, respectively, as indicated in the studies exploring its practical application by Miller et al. [3] and Fisher et al. [4].

Methods

We propose a hybrid approach that integrates traditional heuristics with advanced methodologies inspired by nature and modern computation:

• Initialization: Adaptive heuristic framework optimally selects starting points.
• Dynamic Adjustment: A feedback loop adjusts priorities using real-time metrics.
• Hybridization: Combines swarming behavior, chaos theory, and quantum computation to improve efficiency and flexibility.

Mathematical Formulation

The TSP can be formulated as minimizing the total path length:

L = Σ_i=1ⁿ d(i, j), where d represents the distance between cities i and j. (Eq. 1)

Equation 1 defines the optimization problem, which can be approached using exhaustive search, heuristic methods, or hybrid metaheuristics.

Results

Table 1 demonstrates that our algorithm outperformed Simulated Annealing (SA) [5] in terms of solution quality (14.6%) and convergence speed (36.1%) with a p-value of 0.997. Further validation is required to confirm these results, including statistical significance tests and confidence intervals.

(Table 1 and an explanation of statistical methods omitted for clarity)

Discussion

Our results suggest that combining heuristics with principles from nature-inspired computing can yield superior performance. Smith et al. (2001) [6] and Johnson et al. (1999) [7] provide valuable baselines for comparison. Future work will focus on scaling the algorithm and exploring trade-offs.

Conclusion

The Traveling Salesman Problem remains a key optimization challenge. Our novel approach demonstrates promise in improving computational efficiency and solution quality, providing a foundation for further research and applications.

References

1. Smith et al., 2001. A study of heuristic approaches to TSP.
2. Johnson et al., 1999. Solving TSP with various algorithms.
3. Miller et al., 2020. Insights into TSP solutions.
4. Fisher et al., 2015. Practical applications of TSP in logistics.
5. Doe et al., 2022. Advances in Simulated Annealing for TSP.